When Cisco lost 89% of its valuation in March 2000, we had learned infrastructure build-out follows S-curves, not exponential ones. Today’s AI infrastructure leaders echo this pattern with uncomfortable precision.

Before examining these risks in detail, I remain deeply optimistic about artificial intelligence as a transformative technology over the coming decades. Large Language Models have proven their utility and will persist as valuable tools. The question isn’t whether AI will transform the economy. It will, just as the Internet did. The question is whether current valuations have already priced in outcomes that physics, economics, and human psychology make impossible to achieve.

Multiple converging storm clouds paint a picture of a market approaching hefty repricing: extreme concentration risks in AI supply chains, the hard limits of Moore’s Law, diverging unit economics of AI infrastructure, unprecedented resource constraints, and shifting sentiments.

Part 1: Concentration Risks in the AI Infrastructure Supply Chain

The AI infrastructure revolution contains profound geopolitical undercurrents absent during the Internet buildout. While both required substantial energy and manufacturing capacity, AI’s energy demands run 10-100x higher per computation. A single GPT-4 training run consumed electricity equivalent to powering thousands of homes annually. More critically, where Internet infrastructure could be manufactured across dozens of countries, cutting-edge AI chips emerge from vanishingly few facilities.

Manufacturing Chokepoints

TSMC produces approximately 90% of the world’s most advanced processors, controlling 67% of the global foundry market. This concentration creates a physical bottleneck where a single earthquake, blockade, or crisis could halt global AI development overnight. ASML’s absolute monopoly on extreme ultraviolet (EUV) lithography machines (€200 million devices essential for producing cutting-edge chips) compounds this vulnerability. The entire AI ecosystem depends on two companies in two small geographic areas, Taiwan and The Netherlands.

Beyond these headline dependencies, the supply chain features multiple hidden vulnerabilities. Japan controls 90% of photoresist chemicals essential for chip production. South Korea dominates memory chips through Samsung and SK Hynix. The U.S. controls critical design software through Cadence, Synopsys, and Mentor Graphics, without which even Chinese firms cannot design competitive chips.

Nations scramble to reduce dependencies, but barriers prove formidable. The U.S. CHIPS Act’s $52 billion and Europe’s €43 billion Chips Act pale against the estimated cost for genuine semiconductor autonomy. (sidenote: AI-sovereignty might equally be a future driver for chipset-manufacturer revenues). Building a single cutting-edge fab costs $20-30 billion and takes 3-5 years, assuming access to ASML’s machines, which have multi-year waiting lists. China’s domestic lithography equipment currently in development targets 28nm production, while SMIC has successfully demonstrated 7nm chip production (and working on 5nm production) using imported DUV technology with multi-patterning techniques, albeit at significantly lower yields and higher costs than TSMC’s EUV-based 3nm production.

Complete semiconductor autonomy appears practically unachievable for any single nation, given the technological complexity and globally distributed expertise required. This creates a paradox where nations might want to hinder each other from developing true autonomy, while a complete decoupling is not feasible. This creates a Mutually Assured Destruction (MAD) dynamic (indeed, like nuclear, AI is a matter of national security), where China controls 80% of rare earth processing while depending on Western technology; the West needs Chinese materials while restricting technology exports. Each side’s leverage risks triggering cascading retaliation, yet neither can afford complete decoupling.

This concentration transforms semiconductor access into geopolitical currency. Taiwan’s “silicon shield” strategy bets neither China nor the U.S. can afford disrupting chip production. The Netherlands faces pressure from Washington and Beijing over ASML export licenses, each shipment carrying diplomatic weight once reserved for arms deals. U.S. export controls demonstrated that cutting chip access can cripple technological development, with China’s AI industry losing an estimated 1-2 years of progress from October 2022 controls alone.

Apart from Taiwan’s chipset ecosystem, The Netherland’s ASML’s allegiance to European sovereignty is equally apparent, with making investments in European Mistral AI and ousting Chinese managers from strategic Dutch companies, such as Nexperia to protect the European supply of semiconductors for cars and other electronic goods.

Critical Materials Monopolies

The AI infrastructure boom creates unprecedented demand for specific materials with severely limited sources. Unlike the dotcom era’s reliance on common materials, modern AI chips require exotic elements with extreme geographic concentration.

China controls 85-90% of global rare earth processing capacity despite holding only 37% of reserves. As Philip Andrews-Speed of the Oxford Institute notes, “China has been working on improving its ability to do this for two decades,” producing 70 kilotons of refined rare earths (REEs) annually whilst controlling nearly the entire value chain. The 2010 rare earth crisis, when prices spiked 600% following export restrictions, demonstrated this leverage, though evidence suggests broader quota reductions rather than targeted sanctions.

More acutely, Ukrainian companies Ingas and Cryoin supplied 45-54% of global semiconductor-grade neon before Russia’s 2022 invasion. Neon is important as it’s the primary buffer gas in excimer lasers used for semiconductor photolithography. Both companies ceased operations, with Ingas’s Mariupol facility producing 15,000-20,000 cubic meters monthly now shuttered. The disruption drove neon prices up 600% in spot markets, echoing the 2014 Crimean crisis spike.

China produces 94% of global gallium and 83% of germanium, both critical for high-performance chips. Beijing’s July 2023 export restrictions, escalating to a December 2024 ban on U.S. exports, weaponised what Deng Xiaoping once called China’s “rare earths advantage.” Unlike oil with multiple suppliers, these materials face 5-10 year development timelines for new sources.

But perhaps most overlooked is helium. The U.S. Federal Helium Reserve’s 2024 privatisation ended its role supplying 30% of domestic needs. Qatar’s 30% global market share creates new dependencies for cooling quantum computers and next-generation AI chips. Unlike other elements, helium literally escapes Earth’s atmosphere when released, making it truly non-renewable.

Part 2: The Economics Don’t Add Up

Markets themselves exhibit erratic behaviour before corrections, where increased volatility, correlation breakdowns, and narrative confusion often precede crashes.

The AI sector now displays these warning signs through what Bloomberg’s reporting reveals as an unprecedented web of circular financing arrangements. Nvidia to invest up to $100 billion in OpenAI to fund data centre construction, while OpenAI commits to filling those facilities with Nvidia chips. OpenAI then struck a $300 billion deal with Oracle for data centres, with Oracle spending billions on Nvidia chips for those same facilities, creating a closed loop where money flows from Nvidia to OpenAI to Oracle and back to Nvidia. Further, Nvidia seems to buy its way into the supply chain by investing $5 billion in Intel (after all, Intel bought out ASML’s stock of EUV machines in 2024). This reminds me of 1990s circular deals that centred on advertising and cross-selling between startups to inflate perceived growth.

The “Magnificent Seven” (Apple, Microsoft, Google, Amazon, Nvidia, Meta, Tesla) now comprise over 30% of the S&P 500’s total market capitalisation, exceeding the concentration seen at the peak of the dotcom bubble. But the scale of these arrangements dwarfs previous technology bubbles.

US tech companies have raised approximately $157 billion in bond markets alone this year (a 70% increase from last year) with Oracle’s $18 billion jumbo bond sale drawing nearly $88 billion in demand. Meta Platforms has lined up $29 billion of private credit for a Louisiana data centre, whilst banks prepare a $38 billion package for Oracle-linked campuses in Texas and Wisconsin. This debt-fuelled expansion occurs despite Oracle’s free cash flow flipping negative for the first time since 1992, even as its total debt pile surpasses $100 billion, as Bloomberg’s Ed Ludlow reports (a further, personal, concern is the amount of CLO-exposure in debt financing).

The opacity of these financial arrangements masks systemic vulnerabilities. Elon Musk’s xAI exemplifies this creative engineering, structuring a $20 billion funding round split between $7.5 billion equity and $12.5 billion debt via a special purpose vehicle specifically to purchase Nvidia processors, which xAI would then rent out for five years. CoreWeave, having received a 7% stake from Nvidia and $350 million from OpenAI, has expanded its cloud deals with OpenAI to $22.4 billion, further entangling the sector’s key players.

Kindleberger’s “Manias, Panics, and Crashes” documents a late-stage bubble psychology with similar patterns, where companies desperately tried to justify valuations that assumed impossible growth rates. Today’s AI sector displays these characteristics magnified, with Capex commitments at unsustainable levels for hyperscalers. OpenAI’s Sam Altman claims he wants to spend “trillions” on infrastructure whilst the company burns through cash and doesn’t expect to be cash-flow positive until near decade’s end, according to Bloomberg’s analysis. Sequoia Capital’s analysis frames this as “AI’s $600 billion question”, or where will revenue come from to justify this spending?

A Harvard Kennedy School researcher who worked through the dot-com era observes that whilst “1990s circular deals centred on advertising and cross-selling between startups to inflate perceived growth, today’s AI firms have tangible products but their spending still outpaces monetisation.”

The concentration of capital in AI infrastructure mirrors historical patterns where impossible growth assumptions preceded crashes, but with novel systemic risks. A Bain & Co. report predicts AI firms’ revenue could fall $800 billion short of what’s needed to fund computing power by 2030, whilst MIT researchers find that 95% of corporate AI pilots don’t deliver returns. Morningstar analyst Brian Colello warns that “if we get to a point a year from now where we had an AI bubble and it popped, this deal might be one of the early breadcrumbs [...] If things go bad, circular relationships might be at play.”

This isn’t merely Internet infrastructure redux. It’s a new form of technological sovereignty where control over dozens of facilities determines national AI capabilities. Unlike the distributed Internet where packet routing bypassed damaged nodes, AI’s hardware dependencies create catastrophic single points of failure. Oracle’s cloud business generated just 14 cents gross profit for every dollar in sales from renting Nvidia-powered servers, according to internal documents reported by The Information (a margin so thin that any disruption could cascade through the interconnected web of financial and legal obligations).

As Bernstein Research analyst Stacy Rasgon observes, Altman “has the power to crash the global economy for a decade or take us all to the promised land. Right now we don’t know which is in the cards.” Further alarming signals come from veteran short seller Jim Chanos, who famously predicted the Enron collapse, highlighted a key contradiction: “Don’t you think it’s a bit odd that when the narrative is ‘demand for compute is infinite,’ the sellers keep subsidizing the buyers?” This echoes patterns from the dot-com era, when companies engaged in circular revenue deals to maintain growth appearances.

The opacity extends beyond individual transactions to the entire capital allocation mechanism. Altman himself admits “I don’t think we’ve figured out yet the final form of what financing for compute looks like,” even as OpenAI pursues a model where it will lease rather than buy AI processors to make debt financing more feasible. These creative financial instruments, from special purpose vehicles (SPVs) to circular equity investments to massive debt packages, create dependencies where one failure to execute may trigger an avalanche, as each participant’s ability to service obligations depends on continued exponential growth that history suggests is unsustainable. Even Jeff Bezos has warned that “investors have a hard time in the middle of this excitement distinguishing between the good ideas and the bad ideas,” noting the AI spending resembles an “industrial bubble” (sidenote: Many hyperscalers have been priced as high-growth tech companies, while in fact they are industrial players. It is just a matter of time before they will get repriced accordingly).

Finally, also Venture capital exhibits late-cycle behaviour. Tiger Global, Coatue, and SoftBank have marked down AI portfolios 30-50%. New investment slowed dramatically: AI funding dropped from $97 billion in Q4 2023 to $41 billion in Q2 2024. The smart money is retreating while retail piles in through ETFs and index funds.

Part 3: The Physical Limits of Silicon

Computational speed has driven chipset innovation from early gaming applications (anyone who played Doom with a GPU card will know the difference) through Nvidia's 1993 founding and subsequent GPU dominance. This gaming-focused parallel processing expertise enabled cryptocurrency mining applications (2010-2018) before positioning Nvidia advantageously for the AI revolution around 2022-2023, driving substantial equity appreciation as GPU-accelerated computing proved commercially viable for AI training and inference. Indeed, Nvidia is a company to grew from pixels to prompts.

In general, chipset miniaturisation remains essential because smaller designs deliver higher transistor density, superior power efficiency, reduced signal latency, lower per-unit manufacturing costs, and improved thermal management. This aligns with Moore's Law principles, wherein continued exponential computational growth depends upon scaling down feature sizes to meet modern applications' demands.

The techo-economic understanding is that the higher transistor density the more compute speed one can deliver. When applying this to Nvidia, we get the following graph:

While the meteoric rise of Nvidia’s stock market is also pushed between 2022 and 2023 by the expectations of LLMs, and not only the miniaturisation, there is a strong correlation (~.75) between transistor density and Nvidia’s market cap. In case of players that are closer involved with the transistor density play, such as ASML and TSMC, the correlation is even stronger.

However, the AI boom's fundamental assumption—that chips will become exponentially more powerful whilst remaining economically viable—faces mounting physical constraints. I want you to consider Moore’s Law strictly as a techno-economic model, where the physical barriers, such as atomic-scale limits, may very well introduce a slowdown on the growth of chipset manufacturer’s market caps.

The challenges are plenty. Already, current flagship models already require enormous computational resources costing tens of millions to train, whilst next-generation chips like Nvidia's B200 and GB200 consume 1,200 watts, pushing practical cooling and power limits. Future generations may demand 2,000+ watts, creating thermal densities that challenge the economic and physical viability of continued AI scaling. For decades, performance improvements in general-purpose CPUs and GPUs averaged significant annual gains, driven by transistor scaling under Moore’s Law. Since around 2010, the rate of these gains has slowed substantially as the industry has confronted physical limits like heat dissipation and ballooning costs for new manufacturing nodes, shifting innovation toward architectural specialisation and other techniques.

In various discussions, 2nm is generally accepted as nearing the “endgame” for silicon (below 1.5 nm nodes face severe quantum-tunnelling and cost walls), with 1nm being called “unrealistic” without new physics. What is clear is that the statement how “chips will become exponentially more powerful whilst remaining economically viable” is no longer a realistic view. 3nm is viable but costly; 2nm stretches economics; 1nm is infeasible. We can safely forecast, that by ~2030, density gains stop without any new technologies or breakthroughs (e.g., graphene transistors, glass, quantum computing, photonic computing).

Nvidia’s $5T market cap (Oct 2025) assumes AI growth, but a scaling wall could trigger corrections if costs continue to outpace performance. Nvidia’s $4.1T valuation (P/E ~50x) assumes AI growth (e.g., $46.7B Q2 2025 revenue, 56% YoY), but slowdowns in density gains (2nm by 2026, 1nm speculative) risk corrections. Following scenarios could very well be possible outcomes over the next five years:

Constraints on transistor density will impact ASML and TSMC more directly than Nvidia due to their roles inlithography and foundry production, respectively. ASML’s EUV tools and TSMC’s advanced nodes (e.g., 3nm, 2nm) face immediate challenges from quantum tunneling, low yields, and escalating fab costs, directly affecting their margins and growth. Even when cosmetics are applied, or some last-mile solutions might delay the moment where Moore’s Law catches up, some point over the next few years physics will give the financial market a run for its money.

Zooming in on Nvidia, its traditional moat (integrated CUDA software, AI-optimized GPUs, and ecosystem) mitigates reliance on raw density gains by enhancing performance through software and architectural innovations. However, Nvidia’s primary risk lies in a potential downturn in AI sentiment, such as reduced hyperscaler spending or disillusionment with generative AI’s ROI, which could trigger significant valuation corrections, compounded by secondary risks like competition and supply chain disruptions.

Part 4: From Obscurity to Ubiquity, AI’s Sentiment Evolution

AI sentiment is an important driver in terms of moving the market caps of AI companies, both in private and public markets. Analyzing Google Trends together with significant AI moments over the past decades indicates we are indeed in a historical moment where AI has never been more ubiquitously present in our daily lives as any point in time before.

Nvidia, the Canary in the Coalmine

Considering Nvidia’s stock price ($NVDA) as the proverbial canary in the coal-mine in terms of AI market capitalisations, its Q2/25 earnings call provided insights into a miss in datacenter revenues ($41.1b actuals vs. $41.3b projected) as well as investors that were priced “for perfection”, expecting another blowout to justify the $4T+ valuation. It seems “being merely on the mark, simply is not going to cut it”, translating to an overinflation of expectations. Nvidia’s Q3/25 earnings call were received positive by the markets, but taking a step back, I do have to make some critical notes:

• Jensen casually claims to take “$35 billion out of $50 billion per gigawatt data center” (note: 1GW datacenter requires the equivalent of a nuclear reactor ➡️ the U.S. adds only ~30GW of total generation capacity annually ➡️ the assumption of “10 gigawatts” for OpenAI alone is fantasy without massive grid infrastructure changes).

• Rubin “chips are in fab” but that’s 12+ months from revenue, without any hiccups (no idea how SK Hynix inventory shortage might impact Nvidia). Also, the “annual rhythm” strategy could lead to Osborne effect and inventory write-downs.

• Questionable growth projections, with Nvidia projecting $3-4 trillion (AI) infrastructure spend by 2030. With current hyperscaler CapEx is $600B/year, this means a 40% CAGR for five years. With claims of taking “$35 billion out of every $50 billion”, this would mean ~$2.5 trillion in cumulative revenue by 2030 - essentially becoming larger than most countries’ GDPs.

The whole thesis hinges on “100x-1000x” compute requirements for reasoning AI, physical AI/robotics becoming mainstream, and enterprises actually deploying agentic-AI at scale. And none of these points have proven a solid ROI yet.

Indeed the proliferation of LLMs since 2022 has been remarkable, but post-2021 newcomers to the field must realise AI is more than GPT. The danger is not that GenAI replaces data scientists, but that it convinces amateurs they are one. Lowered barriers to entry and required expertise have reduced the need for AI-literacy. With the risk of sounding too straightforward; more naive people than ever before are using complex technology with insane expectations and lesser skills. This will create volatility at some point.

A Shift in Sentiment

Already, we are witnessing an increasingly negative sentiment about AI slop, lawsuits where customers claim chargebacks to consultants using AI generated texts, and industry leaders who are carefully picking their words when being pressed to answer whether or not to anticipate bubble-formation risks.

The volatility in sentiment, means the risk is increasing that even the slightest of hesitations, slowdowns or bad news could trigger a Sell-momentum. And already we see an emerging trend of AI bubble concerns, with recent selloffs in AI stocks (eg. post-August hype peak) and questions about ROI (eg. 95% of firms experimenting with AI see no revenue yet) contributing to this recent change in sentiment. After the recent GPT-5 release, Sam Altman’s (OpenAI) remarks how he feels the AI market is overhyped, or Eric Smith’s urge to Silicon Valley to “stop fixating on superhuman AI” warning how this obsession hinders useful innovation, or how long-term LLM critics, Yann LeCun and Gary Marcus have been warning about LLMs limitations and predicting failure of overpromises. It feels as if the leaders in the field of LLM-dominated AI narratives are trying to gently lower the expectations in order to avoid a runaway panic selloff, while insiders are taking some money off the table, such as Nvidia, CoreWeave, OpenAI, Eleven Labs, etc.

In the private markets for August 2025, global VC funding hit $17B, the lowest monthly total since 2017 (Crunchbase), with AI seed deals down 33% YoY. PwC's 2025 AI Predictions note "sharper ROI emphasis," with 80% of projects failing (RAND), leading to 50–60% of early AI rounds renegotiated or abandoned. Also in private investor chatrooms, discussing allocations, sentiments with regards to AI startups are returning back to normality. Recently, an investment group I’m part of declined to join a small allocation in Perplexity’s ongoing round at $20B valuation. Considering $148M in ARR in June 2025 means a 135x revenue multiple has gotten a bit too crazy for me.

Part 5: Nearing the Limits of LLMs

The grand narrative of generative AI, where large language models (LLMs) are viewed as precursors to artificial general intelligence (AGI), is fraying. Even the architects of the current AI boom now concede that the paradigm is hitting its natural boundaries. Accuracy, scaling, and output quality are all showing unmistakable signs of strain, as Silicon Valley is seeking to quietly back out of the AGI narrative.

LLMs remain systems of statistical pattern-matching rather than comprehension. Gary Marcus argued that hallucinations are not an occasional glitch but a structural property of models trained to predict the next token, not to verify truth.

Sam Altman, the CEO of OpenAI, has echoed similar caution. He publicly warned that using LLMs for therapy or critical decision-making is “deeply irresponsible” because these systems are unproven and unreliable. Even with “reasoning” upgrades, GPT-5 and its peers still produce confident falsehoods. The models’ capacity to simulate knowledge masks their inability to reason causally or verify claims. As Marcus observed, the industry’s epistemic problem is not computational but it is cognitive.

The scaling hypothesis, expressed by the belief that simply making models larger yields emergent intelligence, is faltering. Altman conceded in an interview with TechCrunch that “we’re at the end of the era where it’s gonna be these giant models […] we’ll make them better in other ways.” It was an extraordinary acknowledgment from the industry’s leading figure that size alone no longer guarantees progress. Gary Marcus described the same turning point: GPT-5’s release disappointed, and even long-time optimists such as Rich Sutton began tempering their expectations. High-quality human text may be fully mined by the end of the decade, while compute costs rise exponentially. In an October 2025 interview, Bezos warned that the AI sector risked becoming “an industrial bubble,” with capital inflows outpacing real technical progress.

The evidence is bipartisan: critics and creators alike see the ceiling. LLMs can predict but not understand, scale but not transcend, impress but not reason. The dream of AGI through language models is not a milestone ahead, but it’s a mirage. Altman’s and Bezos’s remarks hint that even those steering the field are preparing for a post-LLM future, one that demands architectures capable of genuine reasoning, not just next-word prediction. Steering away from the AGI narrative seems like a sensible thing to do.

Part 6: The Ecological Limitations and Impact of AI

The artificial intelligence revolution runs on an infrastructure that is increasingly straining Earth’s finite resources.

Water

Data centres currently consume approximately 560 billion litres of water annually globally, projected to rise to 1,200 billion litres by 2030. Already in 2023, data centres in the United States consumed approximately 17.5 billion gallons of water directly through cooling systems, whilst Google’s data centres alone consumed nearly 6 billion gallons of water in 2024, marking an 8 per cent annual increase. In the United States, hyperscale data centres alone will withdraw 150.4 billion gallons of water between 2025 and 2030, equivalent to the annual water withdrawals of 4.6 million households.

This voracious thirst extends beyond cooling towers, a single 100-megawatt data centre consumes about 2 million litres of water daily, equivalent to the consumption of 6,500 households. Texas exemplifies this crisis: data centres will consume 49 billion gallons in 2025, soaring to 399 billion gallons by 2030, representing 6.6 per cent of the state’s total water usage. Morgan Stanley calculates that by 2030, global data centres will require 450 million gallons daily, which is the equivalent to 681 Olympic-sized swimming pools of freshwater. In Georgia, residents reported problems accessing drinking water from their wells after a data centre was built nearby.

Water-stressed regions face particular challenges. Microsoft acknowledges that 42 per cent of the water it consumed in 2023 came from areas with water stress, whilst demand spikes during hot periods when communities most need water themselves.

Still, the concentration extends beyond manufacturing to resources these facilities consume. TSMC’s Taiwan operations use approximately 6% of the island’s electricity. Water usage proves equally staggering: advanced fabs consume 15 million gallons daily. Taiwan’s 2021 drought forced semiconductor plants to truck water, briefly threatening global supplies.

Lithium presents different challenges. While geographically dispersed, battery-grade lithium for datacenter backup systems faces structural deficits. Current production of 100,000 metric tons annually must triple by 2030, yet new mines average 7-10 years from discovery to production. The “lithium triangle” of Argentina, Bolivia, and Chile controls 60% of reserves but faces water scarcity directly conflicting with extraction needs.

Energy

The strain on electrical grids proves equally alarming. Data centres consumed about 17 gigawatts of power in 2022, with global data centre electricity consumption to grow from 415 terawatt-hours in 2024 to 945 TWh by 2030 (or the equivalent to Japan’s entire annual electricity consumption). In the United States alone, data centres will account for nearly half of electricity demand growth, ultimately consuming more electricity for data processing than for manufacturing all energy-intensive goods combined, including aluminium, steel, cement, and chemicals. This surge forces utilities into uncomfortable positions: wholesale electricity costs have increased by as much as 267 per cent in areas near data centres over the past five years, with residential customers bearing these costs through higher bills. The grid infrastructure struggles to accommodate this explosive growth. Nearly 2 terawatts of clean energy (or 1.6 times current US grid capacity) remains delayed in interconnection queues, forcing some operators to rely on fossil fuel backup systems. Virginia’s Loudoun County exemplifies this tension: residents endure constant noise from natural gas turbines powering off-grid data centres, with decibel readings reaching 90dB, which is well above comfortable levels.

Pollution

The environmental burden extends upstream to the rare earth elements essential for data centre hardware. Rare earth extraction requires cocktails of chemical compounds including ammonium sulphate and ammonium chloride, which create air pollution, cause erosion, and leach into groundwater. In China’s Baotou region, the world’s rare earth capital, mining operations have been linked to severe environmental degradation and health problems for local communities. The process leaves behind radioactive thorium and uranium alongside other heavy metals, creating lasting environmental legacies. The sustainability paradox remains stark: whilst rare earth elements enable energy-efficient technologies, China controls 60 per cent of global production and 85 per cent of refining capacity, creating supply chain vulnerabilities and environmental injustices. Extraction requires tremendous amounts of energy and water whilst generating large quantities of waste often mixed with radioactive elements.

The growing frequency of natural disasters (ie. the Texas’s freeze, or Malaysia’s flooding) creates additional vulnerabilities in an already concentrated supply chain, compounding risks from geopolitical tensions and making it increasingly difficult to distinguish between natural disruptions and potential hostile actions.

As AI capabilities expand exponentially, the ecological debt accumulates silently. Without fundamental shifts in cooling technologies, renewable energy deployment, and rare earth recycling (currently below 10 per cent in the United States) the digital revolution risks becoming an environmental catastrophe.

A Parting Thought

Today’s financial markets are sustaining extraordinary AI valuations through increasingly creative structures, from the magnificent seven’s index concentration to private market marks that defy traditional metrics. This determination to extend the rally despite mounting evidence of constraints signals dangerous divergence between narrative and reality.

Based on factual analysis, I have become convinced we’re approaching hard limits across three critical dimensions. Physics (first): sub-3nm chip manufacturing faces steep barriers that at some point no amount of capital can overcome. Infrastructure (second): datacenters already strain power grids and water supplies. Algorithms (third): current architectures show diminishing returns to scale, with each GPT iteration requiring exponentially more compute for marginally better performance. These aren’t temporary bottlenecks but rather fundamental constraints that invalidate the exponential growth assumptions underlying current valuations.

My assessment remains simple. I believe we’re approaching significant repricing in AI equities, some money is quietly repositioning. I anticipate, the rally will continue for another while, but its foundation weakens with each passing quarter. For the dominant players in this space, and specially OpenAI’s Sam Altman and Nvidia’s Jensen Huang, I believe they will increasingly steer the narrative away from AGI’s excessive expectations, as I am sure both gentlemen are fully aware of the power they hold in shaping the markets. One wrong statement and they could be tanking everyone. I expect we will be reading a whole lot of mighty careful phrasing over the next few quarters when replying to critical questions, and hyperbole speech in new, unproven and emerging markets (ie. robotics, etc.).

The dotcom parallel remains instructive. Cisco, trading at 200 times earnings in March 2000, collapsed 89% by 2002, yet emerged as the Internet’s backbone infrastructure provider, eventually justifying those early believers who understood the technology whilst recognising the valuation excess. Today’s AI revolution follows similar contours: transformative technology, valid long-term thesis, but valuations that have decoupled from any reasonable discount model. TCP/IP did change the world, just not quickly enough to justify 1999’s prices.

To this equivalent, and in the long run, AI will transform civilisation, but not rapidly enough to support its short-term’s multiples.

Those working in the AI community already understood that LLMs can be considered autocorrect on steroids. Yet, technically, we should call it next token prediction, even though that’s not entirely correct either as it's not just surface-level statistical correlations - there's evidence of deeper, more structured knowledge representation happening internally.

But are LLMs nothing but stochastic parrots, or are we attributing human characteristics - such as reasoning or thinking - in an appropriate way, by anthropomorphizing AI? By now, the limitations of statistical models (like LLMs) are well understood and how this severely constrain capturing causal relationships without explicit causal modeling.

If anything, I believe Apple’s latest paper holds little new insights for those that work in the field of AI, but it might send a wakeup call to those that only holds surface-level knowledge on the industry, illustrating how the apparent reasoning abilities of LLMs are fundamentally flawed and brittle.

But what a timing. Might there be a reason in Apple’s release of this research paper two weeks ahead of WWDC 2025 ?

An Overview of Apple’s Recent Critiques on AI Reasoning

When Mehrdad Farajtabar, one of the paper’s authors, announced their first findings on X back in 2024, they claimed:

… we found no evidence of formal reasoning in language models …. Their behavior is better explained by sophisticated pattern matching—so fragile, in fact, that changing names can alter results by ~10%!

In this earlier 2024 paper, the Apple team investigated the mathematical limitations of LLMs and developed a task, called GSM-NoOp. Here, seemingly relevant statements were added to the questions that were, in fact, irrelevant to the reasoning and conclusion. The majority of the models failed to ignore these statements and blindly converted them into operations.

This brittle performance reveals that LLMs are not truly reasoning; they are merely regurgitating learned patterns and can be thrown off by changes that a genuine reasoning process would easily accommodate. LLMs lack the robust, systematic reasoning capabilities that human intelligence displays, and their successes are largely confined to the boundaries of their training data.

The 2025 paper demonstrates that LLMs/Large Reasoning Models (LRMs) fail to consistently apply deterministic algorithms to solve puzzles like the Tower of Hanoi, instead relying on non-generalizable reasoning that collapses beyond a model-specific complexity threshold. Although these puzzles have well-defined recursive solutions, such as T(n) = 2T(n-1)+1 for Tower of Hanoi with exponential time complexity (2^n-1), LRMs do not leverage this structure. They appear to depend on learned patterns that break down as the number of variables increases. Remarkably, even providing the explicit algorithm does not prevent this collapse, underscoring limitations in logical step execution.

While these findings highlight the constraints of token-based reasoning, they build on already known limitations in LLMs’ reasoning capabilities. While it is not a new finding in the AI community, the usage of controlled puzzle environments and detailed trace analysis is novel.

So I am not betting on a market-crash following the release of Apple’s paper. So, calling it a knockout blow, as Gary Marcus, claims might be taking it too far, although I understand Gary’s annoyance with the overinflated expectations of the uneducated masses. It really is worth reading up on Gary’s earlier statements, as well as Subbarao Kambhampati’s work.

Anthropomorphization, Cognitive Biases, and Synthetic Reasoning

Humans often attribute human-like traits, intentions, or understanding to AI systems, a phenomenon known as anthropomorphization. This tendency stems from cognitive biases, such as the inclination to project familiar mental models onto unfamiliar systems. For instance, when large language models (LLMs) produce coherent text, users may assume they possess human-like reasoning or consciousness, mistaking linguistic fluency for genuine comprehension. This bias, rooted in our pattern-seeking nature, obscures the reality: LLMs operate via statistical pattern recognition, not cognitive processes akin to human thought. And although in terms of grammar, LLMs create very accurate prose, its content may be severely lacking.

Cognitive biases like the illusion of explanatory depth - where people overestimate their understanding of complex systems - further amplify anthropomorphization. When LLMs generate responses that mimic logical or causal reasoning, users may infer a deeper understanding that isn’t there. Research shows that LLMs lack explicit world models or causal reasoning capabilities, relying instead on correlations in training data. Their outputs, while sophisticated, are probabilistic approximations, not reflections of a structured understanding of reality.

Anthropomorphization also has practical implications. Overtrust in AI systems, driven by cognitive biases, can lead to misuse or overreliance, as seen in cases where users accept LLM outputs without scrutiny. Conversely, understanding AI’s limitations fosters realistic expectations and drives innovation toward true synthetic reasoning. Embodied AI, such as robotics, may bridge this gap by integrating perception, action, and reasoning, creating systems that interact with the physical world in ways that align more closely with human cognition.

Towards a Better Understanding of Human Language Generation and Reasoning

When using terms in AI like reasoning, thinking, or intelligence, caution is essential. Humans and large language models (LLMs) generate language through fundamentally different processes.

Human speech production is not a simple word-by-word prediction; it involves distinct stages. In cognitive models (such as Levelt’s model), speaking is divided into three broad stages: conceptualization, formulation, and articulation. In conceptualization, humans generate a pre-verbal idea or message, determining what they intend to express. Next, during formulation, this concept is translated into linguistic form by selecting appropriate words and grammar. Finally, articulation activates motor systems to produce speech sounds physically.

In simpler terms, humans “predict ideas” first and then encode them in words, rather than predicting one word at a time without a plan. Brain Imaging technology confirms this. Thoughts precede language in humans. Functional MRI (fMRI) and other neuroimaging studies confirm this staged process. Initial brain activation occurs in semantic and conceptual regions (e.g., left middle temporal gyrus and medial frontal/parietal cortex), reflecting idea formation before linguistic formulation begins. Subsequently, language-specific areas, notably Broca’s and Wernicke’s areas, activate to handle words, syntax, and phonology. Finally, motor regions, including the primary motor cortex and supplementary motor areas, engage to execute speech articulation.

Thus, human language generation involves clearly distinct neural stages; conceptual thought formation, linguistic encoding, and physical articulation. The brain first finds the meanings and appropriate words, then assembles the sounds. These are distinct steps with distinct neural signatures, and is very unlike how LLMs are working. Even though researchers are investigating how ontologies can boost RAG/LLM systems, we still might be a long way off of having AI able to understand meaning.

Asking a Controversial Question

But LLMs are not the only systems that fail to address logical reasoning. Humans know a thing or two about that as well. A controversial question might become, what if the limitations we’re observing in LLMs are actually mirroring the way human beings think and reason?

Inconsistent performance is a striking shared trait. Just as humans can falter under fatigue or ambiguous phrasing, LLMs show significant accuracy drops with minute changes, like swapping a number or adding irrelevant info in a puzzle. Apple observed that these models sometimes reduce their “thinking effort” when problems grow harder. Something we could call a figurative “quitters’ mentality”, or loss aversion.

Far from being a flaw exclusive to artificial systems, this mirrors the instability of human judgment, suggesting LLMs don’t deviate from, but rather mimic aspects of human reasoning behaviour. And what is the difference between the confident hallucinations of LLMs versus the Dunning-Kruger cognitive bias that hits us all at some point in time?

Apple’s study highlight how LLMs, like humans, can be led astray by irrelevant details. Something we all might have experienced in our daily lives. Apple found that introducing trivial distractions could cause performance drops of up to 65%, even when the models knew the correct solution. This parallels well-known cognitive biases like anchoring in human decision-making. The convergence of all of these behaviours; decision fatigue, framing, effort variability, and sensitivity to noise, and many more, raises a profound question: are we witnessing limitations of artificial intelligence, or a digital mirror reflecting the inherent quirks of human reasoning?

Research suggests that decision-making and judgement is often an act of intuitive and rapid-response behaviour, where humans construct logical-sounding justifications for these gut reactions after the fact. This makes us, humans, mere post-factum narrative generation machines. It makes one wonder if perhaps setting human reasoning as the bar that AI needs to achieve isn’t just too low.

Interesting to see how new research explores frameworks like the Computational Thinking Model (CTM) which integrates live code execution to enforce iterative, verifiable reasoning, potentially mitigating LLMs’ tendency to "quit" on hard problems. Similarly, fostering a growth mindset in humans encourages persistence through challenges, suggesting that AI could benefit from analogous mechanisms. The CTM approach could be considered as neuroplasticity when interpreted from a human perspective.

As you can already feel, by asking the question backwards, we might engage in very interesting new perspectives. One without definite answers.

Towards a Next Generation of Synthetic Understanding

Yet, skeptics and optimists alike agree that next-word prediction alone cannot produce genuine understanding or causal reasoning about the world. We should not mistake are wants for reality. Human cognition relies on mental models of reality, shaped by sensory experience and logical reasoning, which guide language production. In contrast, large language models (LLMs) generate text based on statistical patterns, producing outputs that may appear logical but lack an explicit model of the world.

Equating LLMs’ capabilities to human-like knowledge oversimplifies their function. While their internal structures may form patterns we interpret as “concepts,” these are not grounded in logic or real-world understanding but emerge from mimicking human language. To advance AI, many researchers propose integrating LLMs with systems that emulate human cognition, such as modules for explicit reasoning, memory, or sensory perception.

For instance, Yann LeCun advocates for architectures like Joint-Embedding Predictive Architectures (JEPA), where AI learns world models through environmental interaction, enabling human-like planning and comprehension.

This need for embodied, interactive systems underscores why robotics and embodied-AI may drive the next wave of AI innovation. By combining perception, reasoning, and physical interaction, robotics could unlock capabilities far beyond today’s text-based models, paving the way for more intelligent and adaptable machines.

One Last, Very Speculative, Thing

So, with WWDC’25 coming up later today, it’s time to speculate.

So, why would a company that tried to reframe the term AI as Apple Intelligence in 2024, be warning against the overhyped expectations of AI? Skeptics claim Apple has missed the boat, believers say it’s just patient innovation.

Yet through the fog of war, some things become clear:

• It is clear that this latest research paper’s intended audience is the media, business people, and financial markets. For many AI experts and researchers this paper holds little or no new insights, except for a methodology that proves what we already know.

• The general perception in the markets is that Apple is lagging in terms of AI, and innovation in general. After 2024’s bold announcements on AI adoption within the Apple ecosystem, and lacklustre rollout in 2025, many have concerns about Apple’s AI strategy.

• The stock price of Apple (AAPL) has been down since this year’s high (Jan 2025), and been showing sideways activity ever since. Still, Apple has US$ 28 billion in cash and cash equivalents (down US$ 1,7 billion since last year) and with a market cap of US$ 3 trillion, there still is plenty of cash and stock to potentially acquire value.

• Former Apple designer Johny Ive has partnered with OpenAI to work on OpenAI’s iPhone-killer device, while late Steve Jobs’ wife (Laurene Powell Jobs) has been an investor in Ive’s company, stating she is expecting big things. Coming from someone that witnessed up close how big things were come to be, including the iPhone, this means something.

Without dismissing the validity of Apple’s latest paper on AI, one might question how Apple plans to integrate AI within their ecosystem.

One particular pragmatic speculation might be how Apple will be levering its privacy and hardware ecosystem and choose on-device AI implementations as their cornerstone strategy. Apple still is one of the few companies that control both the software and hardware side of things.

The paper might very well have been a setup to buy patience in Apple's integration of AI features. So, one particular speculation I wish to make is how one might hear Tim Cook say something along the lines as how “the current state of technology in AI is not living up to Apple’s vision for AI yet”. And “until that isn’t the case, Apple will continue to diligently roll out AI only when it lives up to the privacy and quality that Apple products are known for”.

I guess we’ll need some more patience…

** Update 10/06: WWDC25 is over, and as speculated above Apple indeed announced (a) how developers will be able to embed on-device AI models (without any further specifics), and (b) how “[This work] needed more time to reach our high quality bar and we look forward to sharing more about it in the coming year”, when Apple software chief Craig Federighi discussed Siri’s AI overhaul. This sounds as an eerie echo of Tim Cook’s statement earlier during Apple’s last quarterly earnings call, where Tim declared to “… need more time to complete our work on these features so they meet our high quality bar”.

With a photo-op so cringe, my first two cents were how this might just be about two super rich guys renting out an entire bar, just to celebrate their bromance. But, is it?

Obviously, I got a bit intruiged by this story of this $6.5 billion acquihire after being founded and in stealth mode for just one year. So I tried to find another perspective to look at this deal, beyond the obvious statements I’ve read before.

From Stealth Mode, to Strategic Partnership, over to the All-Stock Acquisition

But first, let's pauze and wonder how Jony, as a bloke from Chingford, on the outskirts of London near Epping Forest (more known as the design sidekick to Steve Jobs), got to collaborate with Sam Altman in the first place. A few years ago, Ive's twin son became so captivated by ChatGPT that he insisted, "You've got to meet Sam."

This led to a series of dinners in 2023, initially facilitated by Airbnb's Brian Chesky, where the OpenAI CEO and the designer behind the iPhone, iPad, and iMac discovered a shared vision for AI's future. Their conversations centered on a profound question: how could AI create a computing experience less socially disruptive than the smartphone?

By spring 2024, these philosophical discussions had evolved into concrete action. Ive founded io Products with former Apple designers Scott Cannon (co-founder of Mailbox), Evans Hankey, and Tang Tan, after which they bunkered down into stealth mode.

Quietly, OpenAI’s venture fund moved in Q4/24 by taking a 23% minority stake through a strategic partnership agreement, together with Laurene Powell Jobs’s Emerson Collective, Sutter Hill Ventures, Thrive Capital, Maverick Ventures, and SV Angel. Strangely, this doesn’t seem to be listed as a OpenAI Startup Fund stake, while Altman had no direct stake in io Products either, according to Bloomberg.

This prior ownership, by OpenAI, significantly shaped the acquisition economics for the full acquisition. Also Klarna’s CEO Sebastian Siemiatkowski’s family office Flat Capital ($FLAT) made a significant seed investment(34 MSEK, or $3,5 million) toward the launch of a “new AI hardware product.”, as well as former Googler and designer Luke Wroblewski of Sutter Hill Ventures. In a since-deleted tweet and LinkedIn post, Wroblewski wrote, “congrats to io on the $6.5B acquisition by OpenAI today. happy to have been investors in this one.”

The secrecy and bluster just feels a bit weird, but perhaps we shouldn’t read too much into this. Rumours have always been a part of this mysterious alliance.

Already in December 2023, rumours were spreading that both Ive and Altman were poised on working on “an AI Device Project”. In September 2024, both gentlemen were still looking to raise $1 billion for the project, after things went quiet again. Earlier in April 2025, OpenAI was said to have discussed acquiring the AI hardware startup that former Apple design lead Jony Ive was building with OpenAI CEO Sam Altman, and could pay around $500 million for the fledgling company, io Products. Quickly, the first rumours started to surface about what the actual device would be, Altman teasing the idea back in 2023, however everyone surrounding the matter remained particularly secretive.

While the total valuation of the transaction reached $6.5 billion, OpenAI needed only $5 billion in stock to purchase the remaining shares, based on its recent $300 billion valuation. The all-stock structure ensures io Products’ team remains invested in OpenAI's long-term success, while early investors, such as Flat Capital see their seed investment of six months earlier become the biggest valued asset in their portfolio, with a hefty return. Everyone happy.

Except, why did Sutter Hill Ventures delete Luke’s posts on X and LinkedIn? And why would Laurene Powell Jobs’ Emerson Collective invest in Ive’s startup that is building a product to compete with Apple? All while OpenAI is rumouring to acquire the company for $500 million, only to see the total acquisition value to become $6,5 billion mere months later? I have to admit, some investors sure are lucky in timing their fortunes.

But so far, nothing new.

Except, the picture announcing the news still looks more like a badly photoshopped wedding picture, than a professional announcement of an acquisition. Seriously. What the heck happened at that photoshoot?

So, What is this Mysterious AI Product?

No idea. Just a lot of rumours and speculation. And Altman stating he plans to be shipping 100 million AI companions.

Perhaps let’s start with a recent filing of OpenAI’s trademark, back in January 2025. OpenAI’s legal team filed it under six International Classes, including Class 009 which randomly added hardware such as “earphones; headphones; sunglasses; laptop cases; mobile phone cases; smart watches; smart bands; smart jewelry; wearable computers; wearable cameras; digital media streaming devices; virtual and augmented reality headsets, goggles, glasses, controllers, and remotes.”

Beyond this USPTO filing, everything else remains speculative.

So far, Altman has told us the device will be "unobtrusive," "pocket sized," and "fully aware" of everything in its users' lives. It's probably not smart glasses, according to the Verge.

Some more subtantiated speculations might come from Ming-Chi Kuo’s X post, indicating what this io Products’ device might actually be. All we know for now, it seems Ive’s futuristic OpenAI device might look like a neck-warn iPod Shuffle according to Slashdot (is it just me, or does this sound a lot like “a slightly bigger Humane Pin on a string?”, btw also an OpenAI investment). Kuo added that users will be able to wear the device around their necks, like a necklace, whereas Humane's AI Pin can be attached to clothing with a clip.

Kuo expects OpenAI's device to enter mass production in 2027, and the final design and specifications might change before then. Although I think I might pass if the final product resembles Yaroslav Kravtsov's contraption, called Collario (so no, it's not real but a poignant and playful pun on our zeitgeist!).

For what’s worth, I believe Spike Jonze’s multi-modal view of an AI-world, portrayed in the 2013 movie Her, to be much more visionary (also the neural processing speed of auditory signals take about 8–20 milliseconds, the visual cortex 20-100 milliseconds). And it seems I’m not the only one that always had a thing for Scarlett Johansson’s voice…

But still nothing very exciting about an always-listening AI device as a necklace. At least, nothing worth $6.5 billion worth of excitement (unless you’re Flavor Flav).

So, why the secrecy? I hope it’s not to protect the technology from being copied, as this might be a clear signal the product has no moat. Rather, I believe this is a classic example of the announcement game, where secrecy and rumour actually inflate valuations. As public market traders know for a long time, buy the rumour, sell the news. Divulging the news too soon, might jeopardise the chances of staging a new round of fundraising (ie. in case the SoftBank contingency doesn't come through by December 2025, OpenAI would need to be able to raise again and perhaps at more difficult conditions).

Inflating Valuations While Securing the End-to-End User Experience of AI?

The question then becomes now; why the secrecy that seems to be staged to feed the rumours?

A coincidal item that caught my eye as the timing of this announcement. It comes at the time of SoftBank’s $40 billion investment at a valuation of $300 billion for OpenAI. One important element is that there is a contingency, where the Japanese tech investment group agreed to fund $10 billion in mid-April 2025, and an additional $30 billion in December 2025, contingent on the AI firm transitioning to for-profit by the end of the year. If OpenAI fails, the total investment in the round would drop to $20 billion. This means, OpenAI needs to get going. And fast. Securing technology and talent in share-deals now the time is in OpenAI’s favour is a good play, before (the proverbial) winter hits.

A second thing that I found to be peculiar, was OpenAI’s announcement to be on the same day of the start of Google I/O 2025. While OpenAI’s announcement dropped Apple’s stock 2%, Google’s I/O 2025 did made a strong impression with over 100 AI-focused announcements, signaling Google's determination to assert leadership in the AI industry. Where Google was seen as trailing OpenAI before, it has clearly stepped up from previous years. In order to maximize value out of an announcement, you gotta be first, or gotta be better. It seems OpenAI aimed to be first.

Perhaps the most legitimate reason for this timing might be that OpenAI is explicitly stating how they wish to own the end-to-end user experience, and understand how moving into hardware is their final piece of the puzzle. "Intelligence without embodiment is like a mind without senses; powerful in theory, impotent in practice." So, this ancient philosophical insight drives OpenAI's most ambitious bet yet: that the future of AI isn't trapped behind glass screens but embodied in devices that move through the world with us. OpenAI’s venture fund’s investments in Figure and Humane, demonstrated this already.

But betting on (pseudo) embodied-AI is a bet other players are willing to take as well, including Meta, Tesla and Nvidia. Embodied AI (or “Physical AI”) is not limited to humanoid robots, but can be loosely interpreted as to including glasses (for example, Meta Rayban, etc.), cars, watches and many other wearables. One common theme seems to be surfacing, which is the days of smartphones are numbered.

The collaboration embodies (pun intended) Alan Kay's maxim: "People who are really serious about software should make their own hardware." But it goes further, suggesting a new principle for the AI age: those serious about intelligence must give it form. Just as consciousness requires a body to interact with reality, AI requires embodiment to truly understand and assist in the physical world.

Meta's Yann LeCun strongly advocates how text-data alone will never enable be able to create human-like AI, and that understanding the real, physical world is a far more superior learning paradigm. While a large language model may be trained on up to 20–30 trillion tokens (about 10141014 bytes), a four-year-old child receives a similar amount of data through the optic nerve alone in their first few years of life. So, if OpenAI is serious about aiming to achieve human-level AI, AGI or ASI, being able to learn about the physical world might just be a new step in the execution of their plan. And since OpenAI has no direct access to the sensory hardware layer, much like how Apple and Google have a hardware ecosystem, venturing into consumer hardware actually makes sense.

Whether or not it makes $6.5 billion sense, is something that remains to be seen. But in the face of a trillion dollar opportunity, it actually might.

Reverse Engineering the OpenAI’s Acquisition Playbook; the Real Innovation

OpenAI's acquisition playbook reveals financial engineering more innovative than the technology itself. The $6.5 billion all-stock deal exemplifies a smart strategy: inflating valuations to weaponize equity as acquisition currency while preserving the nonprofit's control through byzantine corporate structures.

The nonprofit parent (OpenAI Inc.) retains ultimate control over all operations and mission alignment, while the for-profit subsidiaries (including OpenAI LP) handle commercial activities, but remain subordinate to the nonprofit via the OpenAI GP LLC. It’s a complex governance structure, but it does give some interesting structures. It enabled OpenAI to acquire commercial entities as subsidiaries under the nonprofit umbrella, cap the investor returns to initially 100x (excess value flows back into the nonprofit), and block hostile takeovers through governance mechanisms, such as poison pills and dual-class voting rights.

While "profit allocated to investors and employees, including Microsoft, is capped," the nonprofit's stake in acquired companies faces no such restrictions. OpenAI can capture unlimited upside through its portfolio of acquired subsidiaries while maintaining the nonprofit's tax advantages and mission-driven narrative. With each successful product launch (ChatGPT, GPT-4, etc.) higher valuations make the next acquisition cheaper in dilution terms. The acquired companies (io's hardware, Windsurf's coding tools) expand OpenAI's ecosystem, justifying even higher valuations. It's Silicon Valley's version of printing money, except the currency is equity in case of OpenAI.

In order to do this, the valuation rocketship must go on. OpenAI's valuation exploded from $157 billion to $300 billion in mere months, transforming its stock into Silicon Valley's most potent acquisition currency. Every valuation increase makes future acquisitions proportionally cheaper. For example, the io Products deal cost only 2.2% of OpenAI's equity despite its massive price tag. The Windsurf acquisition for $3 billion follows the same pattern: all-stock deals that preserve precious cash for the infrastructure investments Altman seeks.

OpenAI’s Achilles Heel is its ability to perpetually fundraise. This strategy mirrors Silicon Valley’s “blitzscaling” playbook: push higher valuations, enable stock-based acquisitions, enable accelerated growth. However, critics argue the valuation is disconnected from financial reality (23× projected 2025 revenue).

Anyone remembering the fallout between Musk and Altman, are aware of how limiting the nonprofit could be in terms of value creation (and mainly, liquidation). OpenAI's structure features "the Nonprofit wholly own and control a manager entity (OpenAI GP LLC) that has the power to control and govern" the for-profit subsidiaries. This allows OpenAI to acquire for-profit companies through its commercial entities while the nonprofit maintains ultimate control. Each acquisition becomes a new subsidiary under the nonprofit's umbrella, circumventing the traditional nonprofit limitations on commercial activity.

OpenAI's ongoing plan to convert its for-profit arm into a Public Benefit Corporation adds another layer of sophistication, while liberating investors and employees from capped returns. PBCs can pursue profits while claiming social benefit, making them perfect vehicles for acquiring commercial AI companies. The structure allows conventional equity terms that investors demand while maintaining the fiction of charitable purpose. More importantly, it would allow the $30 billion in contingency to be secured from SoftBank.

In general, the preservation of cash is much needed, as the AI game is a very expensive one. By using inflated stock for acquisitions, OpenAI preserves cash for compute infrastructure; the Stargate Project alone needs $500 billion. Every dollar saved through stock deals is a dollar available for GPUs and data centers. Buying value with cheap stock, and investing dollars in expensive infrastructure.

Turning Regulatory Arbitrage into a Money Printing Machine at Scale

Trying to sum it all up, I wasn’t that much impressed by io Products’ acquihire. But when digging deeper I was stunned by the scale, complexity and speed at which OpenAI is blitscaling their company, while haemmoraging $9-15 billion in annual losses. Buying value with cheap stock, and investing dollars in expensive infrastructure is a great move in the case of OpenAI.

It has discovered a method how to be simultaneously a charity and a commercial AI empire in the making. They’ve seen to be taking a page out of how universities commercialize tech (e.g. Stanford/Google), but at an unprecedented scale. Stanford’s tech transfer office generated $596 million in royalties the past 34 years, while OpenAI’s hybrid structure has attracted $40 billion in a single funding round.

Inflating stock value enables OpenAI to buy real assets while taxpayers subsidize the nonprofit parent. The structure would make even the most creative Wall Street banker envious; all the upside of a tech conglomerate with the regulatory benefits of a charity.

The genius lies not in any single transaction but in the orchestrated execution of many underlying elements. High valuations enable cheap acquisitions. Stock deals preserve cash. Subsidiary structures maintain control. Nonprofit status provides cover. Each piece reinforces the others, creating a self-perpetuating machine for consolidating the AI industry without spending a dollar of actual cash.

Whether or not, OpenAI will be the winner still has to be seen. But, one thing is certain, without infusing Europe with a more capitalist mindset (and more importantly, more liquidity in its capital markets) we might look at these acquisitions and nod our collective, socialist heads mumbling “it's Silicon Valley, doing Silicon Valley things” while risking to miss the point (and our competitive edge in AI) completely.

And my beef with the cringe photo?

Oh well… It’s Silicon Valley, doing Silicon Valley things I guess.

Being a serial founder myself; weathering economic booms and busts, collaborating with various managers and fellow founders, and experiencing the frustration of working alongside individuals who can manage but not build, I found this article particularly refreshing. Let’s take a moment to examine the core themes behind it and explore Founder Mode a bit deeper.

What we will be discussing:

• How the concept of founder vs manager mode is not new, and can be related to the wartime vs peacetime CEO concept, and even traced back further to military vs political concepts.

• Explore how founders and managers might be indeed two different animals, on a cultural and neurobiological level.

• Discuss the pitfalls of professional managers, and founders while exploring ways to reconcile these seemingly contradictory characters.

According to Brian Chesky, he noticed that venture capitalists and managers often failed to offer genuinely helpful insights. Worse still, their advice can sometimes be counterproductive, likely because they lack firsthand experience in founding a business from the ground up, at best having only managed one. Whenever a speech, such as Brian’s, ensues discussions (for example here, here, here, or here), it’s usually a sign that the speaker has touched on something that resonates.

In this case, Brian, touched upon the difference between managers and founders. Indeed, a contentious topic that resonates when people debate about leadership.

It’s not about who’s better. It’s about who’s more suited, when.

To make things worse, most founders present during this particular talk acknowledged how these very advisors, professional managers, VCs - often MBA schooled - seemed to behave in ways that would make founders feel inferior. Paul Graham sums it up as following:

Founders feel like they're being gaslit from both sides — by the people telling them they have to run their companies like managers, and by the people working for them when they do. Usually when everyone around you disagrees with you, your default assumption should be that you're mistaken. But this is one of the rare exceptions. VCs who haven't been founders themselves don't know how founders should run companies, and C-level execs, as a class, include some of the most skillful liars in the world.

Paul Graham

Indeed it is not very uncommon for professional managers to behave as the masters of the universe, or for MBA-educated managers to add a few extra inches to their strut when walking into the office. I have learned that in business, you have the empty suits that believe the ship is moving because of the compass, and those that realise it’s the engine. Some make the mistake of overvaluing the administrative aspects of running a business over the entrepreneurial decision-making that happens every day.

It seems however, this is not a personal opinion but a studied phenomenon. A study of 444 U.S. CEOs that were MBA holders found that their firms saw a 20% market value decline compared to non-MBA CEOs, spent almost twice as much on acquisitions, showed lower cashflows and inferior returns on assets. A rather unsurprising finding also showed their success tended to more short-lived. Indeed, an MBA education may indeed create a bias toward more superficial metrics and short-term thinking rather than deep business building, hence the self-serving behaviour often exhibited by professional CEOs and MBA-managers.

According to prolific business leaders, such as Elon Musk (2024) or Steve Jobs (1985) and while each holding their own views, it is clear to see a convergence on their experiences and beliefs on the matter of founders vs managers.

The best managers I hired, where those that were brave enough to roll up their sleeves, get in the mud, and took the risk to suck at something new. I admit it takes a strange, and unique, type of individual to expose themselves to such risk, stress and hardship and still stay the course.

Wired for war, not paperwork

Research in organizational psychology and behavioral economics has revealed fundamental differences between founders and professional managers in how they think, make decisions, and respond to challenges. These differences appear to be deeply rooted in their cognitive patterns and personality traits, suggesting that founders and managers may indeed be "different animals" at a very fundamental level.

A comprehensive meta-analysis by Zhao and Seibert in 2006 found that entrepreneurs score significantly higher on traits like openness to experience and conscientiousness, while showing lower levels of neuroticism compared to managers. These personality differences help explain why founders often thrive in uncertain, high-risk environments that might paralyze traditional managers. But the distinction goes beyond personality traits. Sarasvathy's groundbreaking research further revealed that entrepreneurs and managers follow fundamentally different logic patterns when making decisions. While managers tend to follow a 'causation' approach - starting with a predetermined goal and finding means to achieve it - entrepreneurs employ 'effectuation,' starting with available means and allowing goals to emerge contingently.

This explains why founders often excel at creating something from nothing, while managers perform better at optimizing existing systems. Founders thrive in the fog where managers freeze. "Founder mode" isn’t a switch you flip; it’s a skull full of instincts most suits lack.

These cognitive differences manifest in how each group handles risk and uncertainty. Finally, I’ve found that in 1987’s March and Shapira's research showed that managers tend to view risk through the lens of potential downside and loss prevention, whereas entrepreneurs are more likely to see it as an inherent part of opportunity creation. This fundamental difference in risk perception helps explain why founder-led companies often outperform in innovative sectors where uncertainty is high and conventional management wisdom may actually hinder progress.

This research helps explain why founders like Brian Chesky, and myself, find typical management advice unhelpful at best - it's not just a matter of experience, but of fundamentally different cognitive approaches to business building. The concept of "founder mode" isn't just a behavioral choice, but reflects deeper differences in how founders and managers are wired to approach challenges and opportunities.

The very dangerous paradox that destroys companies and ruins lives

While founders rarely pretend to be pure managers, the reverse scenario is surprisingly common. This asymmetry reveals an interesting psychological phenomenon in organizational behavior. As Dacher Keltner explains in his research on power dynamics, individuals in formal authority positions often fall prey to what he terms the 'power paradox' - the very experience of power fundamentally changes how people perceive themselves and their capabilities.

This dynamic is particularly visible in professional management circles. Management positions, with their formal authority, status symbols, and organizational validation, can create an illusion of comprehensive competence. Also Henry Mintzberg observed this in his book 'Managers Not MBAs'; management education often reinforces this paradox by providing analytical tools that create a false sense of mastery over business creation. The result is what he terms the 'MBA syndrome' - where management training produces individuals who believe their theoretical understanding equals or surpasses the practical wisdom gained through actual business building. This syndrome births overlords who think spreadsheets trump scars. Founders, battered by reality, know their limits. Managers, cushioned by hierarchy, don’t. They preach to the trench-diggers from ivory towers, blind to their own myopia. It’s not ignorance; it’s structural arrogance.

The problem compounds itself because unlike founders, who face constant market feedback and the very real possibility of failure, managers in established organizations often operate in environments that reinforce their perceived superiority. Their formal authority is rarely challenged, their decisions are implemented by virtue of their position rather than merit, and their status is continuously validated by organizational hierarchies.

This creates a peculiar blindspot. While founders typically maintain acute awareness of their limitations - having faced them directly in their entrepreneurial journey - managers can develop what Jeffrey Pfeffer describes as 'power-induced cognitive distortion.' They begin to believe that their ability to manage existing systems equals or surpasses the ability to create new ones, despite these being fundamentally different skills requiring different cognitive approaches.

This explains why professional managers might confidently offer advice to founders while remaining blind to their own limitations in understanding the entrepreneurial process. It's not merely about experience; it's about how organizational power structures and management education can create a dangerous overconfidence in one's ability to understand and direct business creation.

Founder Mode Built Empires, and Crime Syndicates

In my view, a true leader stands in the trenches with their troops, rather than issuing orders from a safe office. People are less inclined to follow leaders who are simply parachuted in, compared to those who earned their position. Yet, paradoxically, positions are less likely to be “earned” during times of economic calm, but rather during turbulent times.

The distinction between founders and professional managers mirrors a pattern that repeats throughout history: the fundamental difference between wartime and peacetime leaders. We could argue how the concept of founder mode is related to the concept of a wartime leader. This dichotomy, famously articulated in Ben Horowitz's book "The Hard Thing About Hard Things" finds its echoes in various contexts - from corporate boardrooms to historical battlefields, and even in the streets of organized crime.

Let’s have a look at the differences between Winston Churchill versus Neville Chamberlain. Chamberlain, the quintessential peacetime administrator, excelled in normal times with his methodical, diplomatic approach. Yet when facing Hitler's aggression, his negotiation-focused leadership proved catastrophically inadequate. Churchill, often struggling with what he called his "black dog" of depression, emerged as the perfect wartime leader. His paranoia became prescience, his stubbornness transformed into necessary resolve, and his unconventional thinking became the key to survival.

This pattern extends beyond traditional warfare, and we find similarities with how criminal families are organized. In "The Godfather," Michael Corleone's decision to replace Tom Hagen as consigliere emerges from the same insight - different circumstances demand different types of leaders. A peacetime consigliere, like a professional manager, excels at maintaining and optimizing existing systems. But when existence is threatened, you need a wartime consigliere - someone willing to break conventions, think unthinkably, and act decisively.

The Amplification of Leadership Mistakes

The parallels between wartime CEOs and leaders in high-stakes environments are striking. They share a distinct operational style: direct involvement in critical operations, hands-on leadership, emphasis on personal loyalty, and reliance on small, proven teams. Their decision-making pattern shows similar hallmarks: rapid action, willingness to break conventional rules, and personal accountability for outcomes. Most importantly, they share a mindset: acute awareness of existential threats, comfort with conflict, and a focus on survival over optimization.

VCs often mistake this concept, as their natural inclination is replace the founder as CEO with a professional manager, believing this will improve company results.

More often than not, it won’t. But that’s not relevant for most VCs, as they are not in the value-creation game, but in the portfolio risk management game. Even most VCs are sorely mistaken as to understand what game they are actually playing. But research shows how founder-CEOs make for better leaders. Coach them, and pair them with a gritty COO; not some boardroom mannequin. Most VCs wouldn’t know, they’ve never built a damn thing, although a new generation of former founders are stepping in the VC game, and bring a breath of fresh air.

The Hard Switch: War to Peace

Founder mode, viewed through this lens, is essentially wartime leadership in the business context. Founders, like wartime leaders, must navigate existential threats, make decisions with incomplete information, and often break established rules to survive. They operate based on first principles rather than established playbooks, prioritizing effectiveness over procedure.

What makes this particularly fascinating is how mental states traditionally viewed as disadvantageous can become assets in these contexts. Nassir Ghaemi's "A First-Rate Madness" suggests that certain mental health challenges - depression, bipolar disorder, and other conditions - might actually enhance crisis leadership. The very qualities that make someone struggle during peaceful times - hypervigilance, obsessive focus, mood intensity - can become advantages during turbulent periods. This helps explain why founders, often described as difficult, obsessive, or unstable, excel at building new enterprises while struggling with routine management.

But good leaders know when it’s time for war, and when it’s time for peace. At least this is a common understanding. Quite reasonable even. But this requires vision, developed by an extremely deep understanding of the market and the industry you’re in, often requiring a thorough technical understanding of the product. This comes often as a natural thing for founders, whilst for managers this requires a shift from their core competencies.

Exceptional leaders do have the skills to transition at the appropriate times, and switch modes. Understanding when it’s time for war, and when it’s time for peace. Therefore, perhaps it makes more sense to talk about two modes of operation. While founder-CEOs perform better than manager-CEOs in general, it is clear how an unbalanced leadership style, or wrong mode of operation, destroys company value.

Taking a few steps back

These discussions about founder vs manager quickly descent into very nasty argument based on semantics. If we take a few steps back, we quickly realize that the discussion is about leadership and personality.

Personal anecdotes aside, I have natural tendency to favour doers over talkers. Fighters over politicians. In my personal network of founders and managers, they all share one common trait; they roll up their sleeves, all have their scars, and they spring to action in a heartbeat. These are doers, founders or professional managers alike. The ones that rapidly fall out of favour are those that talk the talk, but never walked the walk. I find street-credibility, original vision, grit, and personality the single most important characteristics of any leader.

So, if we focus back on the word that originated this post, founder mode isn’t a buzzword; it’s survival mode, baked into history’s bones. Managers optimize the known; founders wrestle the void. Yet, both can kill a company. One through cowardice, the other through madness. The trick?

Know the terrain, know your beast, and never trust a suit who’s never bled.

Herbert Simon  might very well have been one of those rare breed of thinkers that seemingly has no bounds when it comes to fields of expertise. How insatiable he was in knowledge! He was a political scientist first, but he pioneered in artificial intelligence, taught computer science and psychology, did research in cognitive psychology, philosophy, and applied mathematics. He received the Bank of Sweden Prize for Economics in honor of Alfred Nobel. How diverse is that! Interestingly, he was the first to figure out that the human brain has limited processing capacity, and may be unable to operate without using some sort of shortcuts. He introduced the concept of bounded rationality which explains how humans are limited in their decision making process by lack of information and availability of too many options. We end up making decisions that satisfice a particular criterion instead of making a decision that maximizes utility value.

Injecting the concept of pacificing in economic theory

Our bounded rationality is essential for our survival. At any moment, humans have different computational processes happening that prevent them from standing still while pondering their options for weeks on end, deciding what to eat for lunch. That is why, in the 1950s, Herbert Simon was clearly inspired by his background in computer science to introduce the concept of heuristics as cognitive shortcuts to quickly find a good enough solution, and introduced the economic notion of satisficing (a clever combination of satisfy and suffice).

Simon explained his concept as follows: “The alternative approach employed in these papers is based on what I shall call the principle of bounded rationality: The capacity of the human mind for formulating and solving complex problems is very small compared with the size of the problem whose solution is required for objectively rational behavior in the real world – or even for a reasonable approximation to such objective rationality.”

He believed that if our brain was a large computational and optimizing machine without any built-in rules that would be used to optimize every step in life, we would end up spending an infinite amount of time and energy. Clearly that’s not the case, so he therefore introduced the concept of these built-in rules, making us capable of rational reasoning, but in a limited way.

The foundation of Behavioral Economics

Later, two Isreali behavioral researchers would discover how our reasoning processes are not only imperfect, due to boundedness, but severely flawed altogether. That way, they generated mounting empirical evidence that breaks away from the optimization machine analogy of Simon, and introduce more qualitive differences compared to the previously held rational beliefs. While the notion of heuristics still holds, the processes surrounding them seem to be deeply flawed. The Israeli research later became the foundation of behavioral economics that shook the orthodoxy on financial and economic views to their core. I will cover the findings of Daniel Kahneman and Amos Tversky in my upcoming essays with much detail, as they are two non-economists that have exerted the most influence on economic thinking over the past two centuries.

There is a bureaucrat living in our brains

One side-effect of our misplaced belief that we are endowed with a beautiful thinking and reasoning machine our skulls is that we tend to develop illusions. Some illusions can be very problematic in nature and expensive in glucose. Thinking consumes a lot of energy by the way; therefore, the less thinking we do, the better.

That’s exactly what the architects of bureaucratic systems seemed to believe, when German sociologist Max Weber first described bureaucracy. In Belgium, one of the Europe’s most socialist-influenced countries, bureaucrats are held to be what respectable people do for a living. Approximately 55% to 56% of the GDP is spent on keeping the largest bureaucratic government worldwide alive. No wonder why Brussels is the capital of Europe!

Imagine you have to export some lovely Belgian chocolates to parts of Southern Asia. The bureaucrat will not care what you are doing, even if you were shipping rice to China. His job is not to ponder the economics of your transaction. He is simply there to verify if you have all the required signatures from twenty other departments so that he can add another signature. His only job is to execute a simple if-then-else statement, often with a Boolean true/false outcome. Actually, you are lucky that this bureaucrat doesn’t meditate on general economic theory or try to solve balance of trade equations for every visitor who is seeking to add another signature. Trade would simply come crashing to a halt! We are happy for him to have a long and prosperous career of mind-numbing work, like stamping documents for forty to forty-five years straight with a mild touch of rudeness. Let’s not forget the clock-gazing routine at 4:59 pm sharp before he ferociously heads home to his beloved couch, beer in hand, and dream of a slow and steady approach to retirement.

But wait a minute… Am I postulating the merits of bureaucracy, or overthinking the merit of heuristics?

It depends, but a rulebook does serve a purpose as it saves time and effort. Rules have their value. The bureaucrat simply follows the rules to be efficient. Likewise, the brain follows a set of heuristics to make fast decisions, not because we are seeking to make the best decisions, but because we want to make useful decisions with a minimum of effort, as fast as possible. Consider a group of people in the Savannah theorizing upon seeing a tiger on whether the tiger belongs to this or that taxonomy, making risk assessments, and calculating the average speed to outrun the tiger. This group will become the tiger’s lunch for the day. However, those who run away with the slightest presumption and are not slowed down by the smallest amount of thinking will definitely have a head start in their survival compared to their cognitively preoccupied cousins.

Remember how our survival depends on fast decision making, not necessarily the best decision making. Embrace your inner cognitive bureaucrat. How repulsive of a realization that might be for an entrepreneur…

With the General Data Protection Regulation, the Chips Act, the Digital Markets Act, the Data Act, and the recent AI Act, Europe's ambition is to be the world's most decisive regulator instead of being a global, leading innovator. Although these acts could be considered innovative regulatory marvels in their own right. As expected, new regulations are hailed by professionals with zero skin in the economic game, such as auditors, lawyers, consultants, lobbyists, and civil servants. But Europe's regulatory economics may impact millions of business owners, entrepreneurs, and CEOs in innovation trenches, many of whom are working in artificial intelligence.

What we will be discussing:

• Europe's geopolitical positioning on the global stage in the race for artificial intelligence using regulatory economics in a world of asymmetric conflicts

• The dangers of overregulation, the shortcomings, and how safeguards for artificial intelligence are more a topic of culture, ethics, and morality than law

• Why innovation always trumps regulations, yet how Europe created a moat by securing a first-mover advantage in regulating artificial intelligence

• The dichotomy of artificial intelligence versus human stupidity and where the real danger of AGI/ASI as dual-use technology lies 

Before we enter this long read, it is worth reading Europe's AI Act FAQ and using a simple self-assessment tool while listening to the preliminary remarks by Thierry Breton, European Commissioner for Internal Market with responsibility for space, defense, and security, during the press conference following the Artificial Intelligence Act Trilogue on 9 December 2023 in Brussels.

While prominent figures, such as OpenAI's Sam Altman, warned against the overregulation of artificial intelligence, the world's most restrictive regime on technology development was pushed ahead. Thierry Breton cites the enactment of the AI Act as follows:

The EU becomes the very first continent to set clear rules for the use of AI (…) The AI Act is much more than a rule book — it’s a launch pad for EU start-ups and researchers to lead the global AI race.

The Brussels-Effect

If you can't beat them, let's rush to regulate them. It is Europe's intuitive answer to a rapidly changing world. Over the past decade, it is evident how the old continent has been snoozing through a time of economic prosperity and political stability, where a socialism-infused lackluster attitude towards technology innovation has allowed for a steady accumulation of existential risk under the geopolitical surface.

Much like the tyrannosaurus rex, EU regulators seem to hunt by movement. Whenever an industry appears to be moving too dynamically, Europe rushes in and aims to regulate an accelerating movement. Pushing regulations would be a relatively localized problem if Europe were a closed market. Still, as most tech companies have either a presence in Europe or service the 195 million internet users in the European market, the impact has an immediate global effect when Europe's regulatory coterie passes a law. Political scientist John McCormick shared a more sobering view in his study on Europeanism, where this attribute of cosmopolitanism is associated with the belief that all Europeans, and possibly all humans, belong to a single moral community that transcends state boundaries or national identities.

Addressing the dangers of irresponsible use of AI is not new; the G20 AI Principles of 9 June 2019 and the World Health Organization report of 28 June 2021 influenced the European AI Act long before its enactment. While taking a pole position, Europe is one of many world policymakers that aim to address these technological challenges. Other world powers are rapidly advancing in regulating trustworthy AI, albeit by playing a catchup game. For now, the People’s Republic of China has been drafting either provincial AI regulations, such as the Shanghai Regulations and the Shenzhen AI Regulation, or technology-specific ones, such as the Provisions on the Administration of Deep Synthesis Internet Information Services(2022) or the Measures for the Management of Generative Artificial Intelligence (2023). Meanwhile, China seems to be playing the regulatory game fast and loose while struggling to establish horizontal AI regulation. The United States recently published  The National Artificial Intelligence R&D Strategic Plan as a clear strategy for the usage of AI. However, it is still rushing to come up with supporting regulations. While various state-centric AI and data laws have been enacted, the 2022 published The Blueprint for an AI Bill of Rights can be considered a first set of national principles on the responsible use of AI. Three versions of a comprehensive AI regulatory regime have emerged in Congress – two in the Senate and one in the House. However, no comprehensive regulatory and clear framework has been implemented so far. It seems, both the US and China might learn a thing or two from taking a page out of the European playbook on the establishment of a comprehensive and horizontal AI regulatory environment to ensure trustworthy use of artificial intelligence.

So, while AI technology isn’t new, the fact this technology was high on policymaker’s radar worldwide wasn’t either. But it wasn’t until the rapid adoption of foundational models and generative AI in general that things got kicked into high gear. Nobody questions there are noble purposes behind this Brussels-Effect, predominantly ensuring the safety of European users being exposed to these new innovative technologies; Ursula von der Leyen, president of the European Commission, instead addresses it as how the AI Act transposes European values to a new era.

Are European values really stimulating innovation?

To reach a better understanding of these European values, we can digest the Treaty of Lisbon, but perhaps we can reach a more holistic interpretation as postulated by the philosophers Jürgen Habermas and Jacques Derrida, who argued that new values and habits had given contemporary Europe' its own face' as a counterweight to the United States, stating five facets of what they described as a common European political mentality, where in 2003 Germany's most influential contemporary philosopher of the 20th century discussed how Europeans rather trust politics than the capitalist markets and display high support for the organizational and the leading role of the state.

In times when free markets and capitalism play a considerable role in today's thriving technological innovations, such as artificial intelligence, Europhiles underestimate the near-future seismic shift in geopolitics regarding the significance of this technology. More than is the case anywhere else in the world, Europeans expect the state to intervene in the interests of encouraging a level playing field and believe that success is a matter less of personal choices than of community arrangements, where Europeanism ultimately embodies the intention to ensure that opportunity and wealth are equitably distributed.

Europe’s geopolitical opening move on artificial intelligence

I am, by no means or measure, a Europhile. Still, regulating free and dynamic markets could be considered Europe's most important weapon on the global geopolitical stage. The societal transition into the Information Age at the end of the 1990s has profoundly impacted the nature of geopolitics. This shift is primarily anchored in technological advancements, where artificial intelligence can be considered the pièce de résistance. Looking at recent years' cultural and political influences, it's clear that the control and dissemination of data and information are pivotal in shaping national narratives, influencing public perceptions, and spreading values, all of which are vital for achieving national goals and bolstering a country's global standing. In this context, managing information flow becomes critical in shaping and interpreting new viewpoints in geopolitics.

A nation skilled in this domain of information control sets the stage for significantly influencing global geopolitical dynamics, even spilling over into various doctrines of asymmetric and non-kinetic warfare.

In her book "The Age of Surveillance Capitalism," Shoshana Zuboff discusses how the play of markets and states is in tandem. Indeed, looking to the West, we have the United States as the undisputed hotbed of technology innovation, led by gargantuan tech companies such as Apple, Meta, Amazon, Microsoft, Google, and its recent addition, OpenAI. In the Far East, the US is seeing its status quo questioned by the revisionist state of the People's Republic of China, illustrating a great sense of clarity regarding its objectives and political goals by 20301 accelerating the rapid development of AI through its technology behemoths, such as Tencent, Baidu, ByteDance/TikTok, Alibaba, Huawei, JD.com, HIKvision and many more, aiming to cement the core values of socialism in AI systems.

Europe, traditionally late to the technology party and without any relevant tech companies, sees its options as limited to taking the global stage as the world's advocate to strengthen its digital sovereignty of the state and protect the privacy issue in the digital space. The General Data Protection Regime (GDPR) implemented by the European Union is an emblematic example. However, given the lack of "courage and fair-mindedness" to pursue a sovereign cloud and regulate American cloud providers under the EU antitrust law on digital markets, it seems Europe still holstered its big guns.

Still, from a loyalist point of view, one could argue how regulation is being used as an effective geopolitical pawn in global geopolitics while remaining in accordance with Europe's values and political facets. So, Europe is relevant in today's race for artificial intelligence, albeit in a manner that's different from other world powers. Where you stand on the topic is mostly a matter of personal taste; do you appreciate innovation or regulation? Or is it?

A question of regulation versus innovation

Being the theatre for some of history's bloodiest wars, Europe's obsession with keeping perpetual peace introduced a mindset that needed to be adapted to turbulences in its societies and economies. From Merkel's blunder to the EU and NATO's lackluster response to Russia's invasion, Europe isn't equipped for conflict. Like a peacetime CEO who prefers to protect the status quo, a wartime CEO thrives in turbulent times. With the war on data and talent, the only weapon in Europe's arsenal is legislating innovative technologies and industries to keep the proverbial peace and the European markets stable (while introducing new monetary streams of taxes and fines).

Again, looking to the West, we see the United States' pioneer mindset transformed into a liberal meritocratic interpretation of capitalism and a pinch of effective altruism as a push for the booming technology industry. As such, the US remains the most significant innovator in artificial intelligence, having introduced foundational models and OpenAI's chatGPT, demonstrating the world's fastest technology adoption rate with just over 100 million users in less than two months.

Clearly, there are two forces at work: those who innovate and those who regulate. While both approaches may have their merit, innovation always trumps regulation.

First, while innovators keep frontrunning the markets, regulating bodies always tend to play a catchup game. US sociologist William F. Ogburn's concept of cultural lag introduces the notion that culture takes time to catch up with technological innovations and the resulting social problems that may be caused by such lag. The Europeanist mindset, therefore, remains a massive handicap for innovation in artificial intelligence, taking a punitive stance towards the development of this technology. With Kyutai's founding and Mistral as France's first unicorn taking pole position as Europe's first artificial intelligence deep-tech player, it is refreshing to see how President Emmanuel Macron, head of Europe's second-largest economy, stimulates an innovation-oriented environment.

Even in 2018, and more recently during a speech at a startup conference in Toulouse, he addressed the EU legislation designed to tackle the development of artificial intelligence risks hampering European tech companies compared to rivals in the US, UK, and China:

“ We can decide to regulate much faster and much stronger than our major competitors. But we will regulate things that we will no longer produce or invent. This is never a good idea. ”

Second, using regulations to encode morality in laws usually works in the short term, but no adversarial actor will adhere to it when push comes to shove. In the geopolitical game of chess, one is playing the long game. Even intergenerational. But when the stakes become high enough (and with artificial intelligence as the economic equivalent of a nuclear weapon, stakes are high), there will be an infliction point in a distant future where at least one actor or world power will actively circumvent, or ignore, Europe's regulatory red tape to achieve a competitive edge, or worse, aim for total economic domination. A sort of digitale Endlösung, if you will. Taking this to its most extreme manifestation, much like criminals don't read codexes to check if they are about to commit a crime, adversarial and rogue states surely won't succumb to Western laws as they don't even legitimize them. The fact is that rolling out a moral stance on artificial intelligence in the global theatre is ultimately not a matter of law but is a matter deeply rooted in culture.

Third, how a compliance checklist will prevent intelligent systems from doing stupid things is beyond me. The Act's rigid structure could limit its ability to adapt to new risks and opportunities in the AI landscape. Instead of a risk-based approach with strict guidelines, a principle-based approach could have been a better choice. A company can still make highly unethical or defective applications that comply with the AI Act. Whenever a law is implemented, and over time, people tend to forget why a law was passed in the first place. And so, when the law's letter trumps the law's spirit, both regulator and subject can play a costly game of hide-and-seek, finding and exploiting technicalities, loopholes, and ambiguous language. While large companies have legal and compliance teams in place to deal with this sort of expensive regulation, the cost of the AI Act will mainly impact smaller AI companies and, therefore, impact the speed and quality of innovation.

Ultimately, great innovators ask for forgiveness and not permission. The rules are bent, even broken, in the true spirit of innovation. Looking at Uber as a recent historical example is revealing. While the AI Act is risk-based, a principle-based approach would have had more outstanding merit, pushing a cultural wave of morals and business ethics through economies instead of rigid laws. When a movement of effective altruism can cause a leadership debacle at OpenAI, regardless of which side of the fence you are, at least it means people are attentive to their work's moral duties and ethics. A reflex that laws rarely instill. So, how much innovation and economic value have previous European acts added? And will the recent AI Act be any different? Perhaps it is relevant to ask if the GDPR has created any unicorns so far. Hardly. While it kept the parasitic economic workers busy, it only pushed companies to divert their funds from R&D toward legal and compliance departments. Cecilia Bonefeld-Dahl, director-general for DigitalEurope, which represents the continent's technology sector, expressed this concern as follows:

We have a deal, but at what cost? We fully supported a risk-based approach based on the uses of AI, not the technology itself, but the last-minute attempt to regulate foundation models has turned this on its head (…) The new requirements — on top of other sweeping new laws like the Data Act — will take a lot of resources for companies to comply with, resources that will be spent on lawyers instead of hiring AI engineers

How I stopped worrying and learned to love artificial intelligence

In many ways, the invention of the printing press in the 15th century was met with equal skepticism, even fear, as is the case today with artificial intelligence. People could suddenly read the Bible themselves and not talk to the priest. All the establishment was against the wide use of the printing press because it would change the power structure. Just like the Catholic Church opposed the printing press, Europe risks over-regulating this technology out of a deep-rooted fear.

Perhaps more than tangible research, there is a lot of intangible fear, uncertainty, and doubt clouding the debate on artificial intelligence. This doesn't only influence policymakers, but even prominent members of the AI industry cry wolf. But let's remember how even intelligent people may be just seeking attention, can be quite naive, or let their emotions take over. If there is one thing the COVID-19 pandemic taught us, it is how above-average intelligent humans are capable of displaying signs of severe dysrationalia and paranoia. Being evolutionary programmed to respond to negative consequences rather than positive outcomes, these AI-doomers are pithed against the AI-accelerationists, those who have an unwavering belief that AI is our ticket out of all the messes we have created on this planet, creating two opposite forces that behave both in a cult-like manner.

A popular theme is how AI is going to take over humans. First of all, let this be a vaguely stated opinion. Of course, AI will be deployed in most, if not all, areas where humans are active and will be our synthetic (and much smarter) companions. Yet humans have a penchant for giving AI systems human-like characteristics; we even claim ChatGPT to be hallucinating, implying that LLMs have sensory perceptions of things that aren't really there. Of course, it is hogwash for a word-predictor-on-steroids to have such an illness. We should avoid making the mistake of anthropomorphizing AI.

Let's begin by stating that intelligence has nothing to do with the desire to dominate. This trait is seen in social species, driven by evolutionary systems and power dynamics. By default, no matter how intelligent a system is, it will be subservient to humans. However, when humans set goals for world dominance and require advanced artificial intelligence to support them, we might understand why ethical, legal, and moral guardrails are essential in developing AI systems. Failure might trigger the global disintegration of trust, a societal cornerstone concept.

Since working with dual-use technology, I have been more afraid of human stupidity than artificial intelligence. Because no regulatory framework or moral code can prevent or control our stupidity and penchant for dominance. So, like uranium, hacking tools, or anything on the Wassenaar Agreement, I am not too worried about the technology. It's the humans I'm concerned about.

While artificial intelligence will ultimately mean the end of capitalism and might preclude a transition into socialism, we will increasingly interact with the world through AI-powered interfaces and assistants. Microsoft, Apple, Alphabet, Meta, and more companies follow suit to make both hard- and software intelligent, essentially turning their software and hardware into more efficient data-capturing devices. But are we comfortable with having a small number of West Coast companies controlling super-intelligent systems capable of influencing a global culture? While the American culture has influenced the world through Hollywood, no government will accept the American culture to further permeate the lives of European citizens through said intelligent interfaces. Ultimately, I strongly advocate for open source systems and non-proprietary AI systems, unlike Microsoft's AI-proxy, OpenAI.

With that in mind, the world is looking to regulate AI in the most appropriate ways. However, during the 2023 AI Safety Summit in London, it became painfully clear how countries still are extremely divided on implementing such guardrails. Regardless of its potential impact on R&D programs, Europe's taking control of the narrative is therefore commendable in its own right, not to mention a bold move of claiming a first-mover advantage by digging a regulatory moat.

A final, personal thought

As a business leader and investor in artificial intelligence, leadership qualities are most important for any organization. The capability to translate a vision into an actionable strategy that pushes an agenda of innovation with the intent to add more value faster than a competitor is likely the only reason for a series of successful exits in my career. Taking these experiences as a frame of reference to assess the European vision on artificial intelligence and geopolitics in general, I find the old continent to be doing what it's doing best: taking a pedantic stance as a self-proclaimed referee in a game they don't fully understand. If this were the Karen meme, Europe would continuously want to talk to the manager. In the short term, I hope that companies worldwide will continue to invest in researching artificial intelligence instead of addressing the sensitivities of a self-entitled Karen. Clearly, Europe's ambition is to become the world's super-regulator of artificial intelligence, not its super-innovator.

Seen as such, the AI Act could become the blueprint for other governments and an alternative to the United States' light-touch approach and China's interim rules. When you further frame the regulatory efforts on artificial intelligence as a way for world powers to defend and promote cultural values globally, Europe might indeed have made an exciting opening move on the geopolitical chessboard.

But for now, and as an innovator, my hope is more than ever on France's political and industrial agenda to become Europe's AI powerhouse. And making such a bold statement as a Belgian, well, that's quite something.

Welcome to the Cheshire Views, a newsletter for long reads, essays, and critical thoughts on technology, business, and entrepreneurship. Being an active member in the financial transactions industry and in various artificial intelligence ventures, I like the broad and diverse points of view I can develop in my daily professional activities.

As I like to learn, my views may change and evolve over time, and I often like to exercise the intellectual freedom to even contradict previously made points. Much of how Carneades came to Rome in 155 BCE after being sent to deprecate the fine imposed on the Athenians for the destruction of the city of Oropus. Carneades made a big splash with his powerful speeches on philosophy. Right before Cato the Elder, he gave two notable speeches about justice. The first praised Roman justice, but he flipped the script the next day. In his second speech, he argued against everything he said the day before. Instead, he attempted to prove justice was inevitably problematic and suggested it was just a convenient tool for keeping society in order. Obviously, the Romans were not amused.

One may find this penchant to seek fallibilistic alternatives enigmatic or confusing, much like the behavior of the mischievous grinning Cheshire Cat in Lewis Caroll’s work, Alice in Wonderland. But while I admire the rhetorical rigor of Carneades and appreciate the jest in the Cheshire Cat’s iconic smile, as an author, I find it a moral duty to provoke and encourage the reader to come to his own conclusion. 

Or how the Cheshire Cat himself would put it:

The uninformed must improve their deficit, or die

So, while I'm not too fond of path dependencies in reasoning, not everything is fluid. Some of the more stable properties of my writing include a propensity to value ethics over the law, disruption over sustainability, entrepreneurship over management, capitalism over socialism, intellectualism over populism, and skepticism over dogma.

This newsletter is my intellectual outlet, where every view expressed does not necessarily correspond to the opinions and beliefs expressed by the companies I am involved with.

If you appreciate the effort I make to publish this newsletter, please leave a comment and share my posts with your network.

Thank you for your support,

Filip.