Scratching the Surface of the OpenAI / IO Products Transaction
OpenAI's acquisition of io Products reveals sophisticated financial engineering: using inflated stock as currency while preserving ultimate governance in a bid for understanding the physical world
On May 21, 2025, OpenAI announced its largest acquisition ever: the $6.5 billion all-stock purchase of io Products, Inc, a hardware startup spun out from LoveForm and led by Jony Ive and several former Apple employees. The deal, brokered by Wachtell, Lipton, Rosen & Katz is expected to close in summer 2025, represents a defining moment in OpenAI's evolution from software to physical devices.
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.