Google I/O, OpenAI io, and the Battle for a New World

The most important story of our time

I.

Sam Altman has announced—in a video way too long for the creatures of shrinking memory spans that dwell online—that OpenAI has acquired io, the new venture of former Apple design legend Jony Ive, for $6.5 billion. Being pre-product and pre-revenue (and even if it was post-), that’s a crazy number. Altman is confident, however, that it's the deal of the century, potentially adding $1 trillion in value to OpenAI, per the Wall Street Journal.

What is io exactly? An “AI companion” or “AI device” startup. That's the simple, factual description. But what is it to Sam Altman? The future. A brighter one. A world shaped on his terms, unlike the current one, which he’s forced to share with Google, Meta, and Elon Musk. But neither an overused label nor a grandiose dream fully explains io.

io is also, in a way, about Apple—Altman’s all-time favorite company, the ideal blend of daring innovation and exquisite design. Here’s the analogy I’ve come up with: What Apple, Steve Jobs, and the smartphone—understood, without grudge, as the digital platform on which we build, nurture, and spend nearly half our waking lives—have been to the first quarter of the 21st century, OpenAI, Sam Altman, and io’s product (undisclosed) aim to be for the rest. Though knowing Altman's unique mix of ambition and audacity, he’d probably say: the rest of human civilization, not just the century. To him, this is the beginning of infinity.

That’s io, the company. What about io, the (non-existent) product? (I will refer to the company and the product as io because I kinda like the name.) For that, we have to rely on the scarce details WSJ, Bloomberg, and other outlets have reported about what Altman told OpenAI employees. From WSJ:

The product will be capable of being fully aware of a user’s surroundings and life, will be unobtrusive, able to rest in one’s pocket or on one’s desk, and would be a third core device a person would put on their desk after a MacBook Pro and an iPhone.

Contrary to what one would naturally imagine from this, and much to my relief, the product is not a phone. It's also not a pair of glasses. They’ve even discarded screens altogether and Ive profoundly dislikes the idea of making a device to wear in the body (perhaps around it, as Ming-Chi Kuo, industry specialist, suggests). I will confess now: the prospect of relinquishing screens is so appealing that I'm tempted to lower my guard. I want presence back, people eye-contacting one another. I want to forget the word “scrolling” ever existed and recover enough attention span to watch a 9-minute launch clip in one sitting.

But this is Sam Altman we're talking about, the person trying to save humanity while being so untrustworthy that his right-hand man and left-hand woman arrived at the conclusion they had to push him out, separately. The person who would do anything—even face Musk (twice)—to control his destiny and that of his company. The person who has the nerve and focus to do just that. He already has. Is io about a brighter future or about navigating the dubious legality of turning OpenAI into a for-profit?

But hey, I’m not here to condemn before the facts are on the table. Let him cook, whatever he’s cooking. Then we judge. What I care about is how this affects the state of the technological chessboard—especially the part that concerns us, the ever-unwitting public.

Why io? Why now? What's next?

II.

I try not to mix competitors too much in the same post, but I cannot stop thinking about the secret, hidden connection between Google I/O and OpenAI’s io. Don't laugh, this is not a joke. Both I/O and io happened this week. It's confusing but makes for a good, catchy headline (couldn't figure it out, though).

Now, don't take either the naming or timing choices as juvenile pettiness. I've already seen people say that. That's lazy thinking. Altman, petty? Come on. (Besides, io sounds as much like I/O as it sounds like iOS; perhaps that is the inspiration.) So no, I don't think he intended for the io acquisition to steal Google’s spotlight. I mean, as if he could. Given Google's impressive suite of announcements—which I’ll briefly review in a moment—it would've seem like a lame failed attempt whose lameness would be multiplied by ten given that OpenAI already did it last year, launching GPT-4o the day before I/O 2024—and then again, naming its research agent “Deep Research” in February 2025 after Google released its own “Deep Research” in December 2024. Ok, maybe they are lame.

I don’t think it was to mock them either—nonchalantly copying the name just to get under Sundar and Demis’s skins. Altman loves to do this but he can’t hand-wave the over-production of the launch video. It's too self-serious. The mocking has taken place, though, if just as a fun meme at Altman's expense because he insists he doesn’t think “too much about competitors.” (Oh, Altman, always so effortlessly witty and charming.)

What matters, above spotlight theft, mockery, and copycatting, is sending a message. io—both in name and in purpose—is a statement of intent, Altman's style: “This is your world, big guy, for now.” Altman knows he can't topple Google's empire on their terms, but—in typical infinite-agency-human fashion—he believes he can create a new world where Google's power is rendered meaningless.

io is the seed of that world, and it couldn't have been planted at a better moment.

III.

Let me side-note a relevant piece of trivia.

As far as changing worlds go, the planetoid Io—the third of the Jovian moons, like “io” but capitalized because there aren't any Gen Z astronomers—is a good contender to host life if Earth were to become inhospitable. (I'm growing suspicious that too many billionaires seem obsessed with this topic.)

Io would be to us a new physical home, like Altman wants io to be a new digital home.

Wait, no. A quick search tells me Jupiter's Io is actually the one moon that's undoubtedly inhospitable. It's the most volcanically active body in the solar system due to tidal heating from the gaseous giant's gravity. Its ground is covered by a mix of sulfur and lava. It has almost no water, intense radiation, and surface temperatures that go from icy heart to hellish overnight.

So if Earth became inhospitable, Io would be among the worst places to go. I guess Altman didn't think the name through. Or perhaps he did.

Anyway.

Dantean and bizarre as all this io business sounds—AI companions, a battle of tech giants, a moon you don’t want to visit—I believe it deserves our attention as a focal point: whatever happens next, eventually we’ll look back and see, with the absolute clarity that only hindsight confers, that the timeline forked at this very moment.

Much like it did in 2007, when Jobs, in his usual attire, took the stage to present the most unusual of his company's inventions, one that would change the world—perhaps, dare I say, more than he would have wanted.

And much like in 2007, I'm yet to decide if the forking path we ought to take in this garden we call Earth is the one path we will take.

---

---

IV.

I said above that Google I/O was amazing. But that choice of adjective is underplaying it. It was amazing on top of amazing because Google and DeepMind have been winning for a while now. It felt less like a newsworthy moment where the underdog surprises us and we all cheer and switch sides, and more like a reaffirmation of its hegemony.

Hot take: Google has always been ahead. The past four years of struggle—headlines claiming the death of the search behemoth, analysts underscoring the dominance of OpenAI and ChatGPT, and investors wondering aloud “who's going to be the next Google”—have been, at most, a hiatus. The time it takes the beast to awaken from hibernation. A global-scale coordination effort to manifest that what doesn't kill you makes you stronger. Turns out, Google is the next Google.

Google and DeepMind’s positions are so unbelievably strong right now—and will remain that way as long as they solve their deficient marketing tactics and finally give UX the priority it deserves—that I don't need to convince you with arguments. Actually, not even with facts. The examples I’m using below to illustrate Google’s leadership are less useful if you read them than if you just let the sheer size of the list make my point.

So here it goes, as succinct as Google's streak of successes allows me to be:

• LLM performance: Gemini 2.5 Pro (with Deep Think) either leads or co-leads—OpenAI o3 and Anthropic Claude 4 are quite good as well—across reasoning, math, coding, and high-taste tester vibes. Besides, users can now control how much AI thinks with a “thinking budget”.

• LLM price-per-token: Gemini 2.5 Flash offers top-tier quality at a fraction of the cost, making it the best price/performance model. (Besides, Google offers plenty of free use.)

• Pareto frontier (price vs performance): The Gemini 2.5 family defines the current Pareto frontier, letting users choose between extreme performance and extreme efficiency without leaving Google’s ecosystem. (Google guys, if you're reading, stop using the LMArena to get your point across.)

• Low-latency inference: Gemini 2.5 Pro and Flash are fast compared to their counterparts, but Gemini Diffusion—a new concept—is effectively instantaneous (5x faster than Google’s fastest model, 2.0 Flash-Lite).

• Memory and long-context: Gemini supports extended context windows (1M tokens) and memory management (Google and OpenAI launched this feature on the same day, but for some reason no one talks about Google’s).

• User growth: While not dominant in mindshare—ChatGPT will probably always hold that honor—Google models have seen the fastest acceleration recently, among users and developers.

• Agents: The next frontier is agents and Google, unlike its competitors, has one of each kind.

  • Project Mariner: Browser agent (akin to OpenAI Operator).

  • Project Astra: Universal AI assistant.

  • Jules: Asynchronous coding agent (akin to OpenAI Codex).

  • Deep Research: Included in Gemini.

  • Protocols: Agent2Agent (A2A) and Model Context Protocol (MCP) integration.

• Open source: Gemma 3 and Gemma 3n are the best small-scale open LLMs.

• Audiovisual models: Google leads without much opposition across all major modalities.

  • Flow: AI-powered filmmaking.

  • NotebookLM: AI-assisted research with audio and video overviews.

  • Veo 3: Best-in-class video generation model (by far).

  • Imagen 4: High-fidelity text-to-image generation (Midjourney level).

  • Lyria 2: A dedicated music generation model (Suno/Udio level).

  • Chirp 3: Industry-leading speech recognition and synthesis.

• World modeling: Genie 2.

• AI-powered search: Google Search still holds the majority of global market share, and integrates Gemini for AI Overviews (expanded with AI Mode).

• Video training data and generation: Google owns the largest labeled video dataset in history (YouTube). That dominance directly supports its leadership with Veo and Flow.

• Science: Breaking boundaries in biology, mathematics, and algorithm design, setting the standard for AI-driven discovery.

  • AlphaFold 3: Accurate structure prediction of biomolecular interactions.

  • AlphaDev: Discovered better sorting algorithms.

  • AlphaTensor: Discovered better algorithms for matrix multiplication.

  • AlphaGeometry: Solved complex Euclidean geometry problems.

  • AlphaChip: Improved the design of Google’s in-house hardware.

  • AlphaEvolve: LLM-powered evolutionary coding agent for general-purpose algorithm discovery and optimization. (I wrote about AlphaEvolve recently.)

• GCP and TPUs (Ironwood): Google owns its datacenters and designs its chips.

• AI-integrated smartphone: The Pixel 9 series features Gemini Nano, enabling on-device summarization, live translation, Recorder, Call Assist, Magic Editor, and the new Gemini Live with camera and screen sharing.

• Generalist robotics: RT-2 and RT-X combine vision, language, and action.

• LLM-integrated robotics: Through PaLM-SayCan and RT-series, Google pioneered using LLMs to plan and execute tasks in physical environments.

• Self-driving: Waymo, under Alphabet, is the most advanced commercial self-driving platform, with millions of miles and real-world autonomy in Phoenix, SF, and LA. (10 million robotaxi trips so far.)

• AI safety and interpretability research: DeepMind publishes more cutting-edge work on safety, robustness, and interpretability than any other lab. (I’m sorry, Anthropic, blog posts don’t count.)

• Cited AI papers and conferences: Google and DeepMind consistently dominate NeurIPS, ICML, ICLR, and Nature AI.

• Multilingual LLM performance at scale: Gemini 2.5 outperforms in non-English evaluations, particularly in Indian, Southeast Asian, and African languages, giving it a global performance edge.

• AI-enhanced productivity suite: Google Workspace (Docs, Sheets, Slides) offers the most seamless and contextually aware AI tools for document editing and real-time collaboration.

• Concentration of talent: DeepMind + Google Research houses many of the top minds in ML, reinforcement learning, theory, and alignment. (OpenAI, Meta, and others used to compete on this, but the quality of their talent pools has diluted over time.)

• Full-stack AI control: Google is the only company in the world that controls chips, datacenters, operating systems, browsers, productivity tools, and AI models.

V.

Wow. Ok. Let’s zoom out. I want to see the full extent of Google’s power, in the shape of a random skyscraper (Sam Altman for scale).

What's that? I can’t hear you from up here.

Let me zoom in one moment.

Not quite there yet.

A bit more.

VI.

io is Sam Altman’s attempt to trigger the quake that brings down Google’s towering stack of successes. How does he plan to achieve such a feat? By changing the substrate on which everything digital takes place. Like many others, he wants to build the next computing platform. Meta and Apple make glasses. Others have tried as well; Jony Ive had the generosity to dedicate them a few words that I paraphrase: “you made trash”.

Meta, Apple, and Google are not invulnerable. The unnamed others are not strong. Unlike all of them, Altman and OpenAI possess a strength yet lack a weakness.

ChatGPT is already the web’s interface for five or six or more hundred million people. That’s the strength. Its continued growth rests on the shoulders of the computing platforms, those that own the devices we use every day, but what would Altman risk if he went down to their level? Not much. That’s the lack of weakness. He has all to win, nothing to lose.

Incumbents can’t afford to tinker at the higher level of AI interfaces for everything without risking it all. Google boldly jeopardized its future by integrating generative AI into search, which provides the majority of its revenue, but it would never go further. How would they make money from ads without a screen? (Please, do not mention that Black Mirror chapter, I'm still having nightmares.)

Digital tech incumbents, like all incumbents, have one big weakness: their empires need a solid, stable, uncontested substrate—phones, screens, social media, computers—on which to stand. They can’t afford to lose it; they can’t afford to have it changed. Analysts thought Google would face the innovator’s dilemma in the shape of ChatGPT. But Altman had another plan; you don't find earthquakes in shallow digs.

He knew trying to kill search was thinking backward. He knew staying in the app layer was thinking small. When he sees Google’s empire, he thinks big and forward: Everything you’ve built only works as long as the ground beneath it holds. So he looks up, unafraid, already standing on a different plate.

It's hypnotic to imagine the chain of thought of an infinite-agency person; how he juggles the mundane and the otherworldly. If he can't fight Google in this world, he will invent a new one in between breakfasts in San Francisco’s coffee shops and taking care of his little kid. He wants to be like Jobs but achieve 10x his impact and influence. I don't know if he will reach that far—I'm not sure whether that strikes me more as extra ambitious or extra presumptuous—but he will try.

This might all sound a little crazy, so let me ground it in history: when you look back at 2007, the golden days of Apple, and think about Jobs’ legacy, can you deny he was a decisive fork in our path?

As for the Google vs. OpenAI battle for a new world, my bets have long been on Google and DeepMind. I’ll have my popcorn ready anyway, for this promises to be the biggest story of our time.

---