In a single day, OpenAI laid out the 2 pillars of its subsequent empire: first, it signed a sweeping deal with AMD to safe at least six gigawatts of GPU compute, an settlement that might give it as much as a ten% stake in AMD if sure milestones are met. Then, on stage at DevDay, it unveiled a brand new layer of “mini-apps” that dwell inside ChatGPT, turning the chatbot into one thing a lot greater: not a product, however a platform.
Collectively, these strikes outline OpenAI’s ambition with excellent readability: management the ability and management the interface.
Energy, actually
The AMD deal is more than a supply contract: it’s a sign. Six gigawatts of GPU compute by 2026, the primary one-gigawatt plant in development, and stock warrants worth up to 160 million shares at a cent apiece if efficiency targets are hit.
That’s not procurement: it’s vertical integration through financial engineering. By embedding itself in AMD’s roadmap for the next-generation MI450 chips, OpenAI is locking in compute capacity at a planetary scale. It’s additionally shopping for affect: the correct to co-design, the flexibility to form pricing, and a hedge towards Nvidia’s dominance.
Compute has change into the brand new oil, and OpenAI simply secured drilling rights.
From app to ecosystem
Then got here DevDay. On stage, Sam Altman introduced “mini-apps” from Spotify, Canva, Expedia, Zillow, and others, micro-interfaces that live inside ChatGPT. The objective: let customers work together with third-party providers with out ever leaving the chat, OpenAI’s bid to make ChatGPT your conversational operating system.
Consider it because the app retailer with out the shop. No icons, no screens, simply dialog. You ask ChatGPT to plan a visit, it calls Expedia; you ask about housing, it queries Zillow; you design a emblem, and Canva seems, seamlessly. The interface disappears. The agent decides.
This isn’t a super-app within the Asian sense. It’s one thing deeper: an orchestration layer that sits above each different digital service, turning pure language into the default management floor in your digital life.
If it really works, ChatGPT stops being a chatbot and becomes the front end of the internet.
We’ve been right here earlier than
Anybody who has watched the historical past of Silicon Valley is aware of how this story goes. Platforms start as enablers and finish as gatekeepers. Within the Eighties, Microsoft used Home windows to regulate distribution. Within the 2000s, Google turned search into an public sale for consideration. Within the 2010s, Apple and Meta constructed app shops and advert ecosystems that extracted rents from the whole lot that handed by way of them.
Now, the interface itself, the dialog, turns into the platform. And the sample is repeating.
When ChatGPT “suggests” which app to make use of, who decides which of them seem? Zillow proudly claims to be the unique real-estate companion inside ChatGPT as we speak. However what occurs when rivals arrive, and everyone knows they’ll? Will placement depend upon benefit, or on bidding? Will we see a market the place firms pay for his or her slot within the agent’s suggestions, as search engine optimisation for AI conversations?
Historical past suggests we are going to. The distinction is that, this time, there’s no search outcomes web page to scrutinize. The choice occurs invisibly, within the circulate of a chat.
The phantasm of company
For customers, the promise is fairly seductive and sounds apparently very effectively. You not must juggle tabs or apps, the agent does all of it, it even starts the conversations. However the value of comfort is asymmetry. If you ask ChatGPT to “discover one of the best flight,” you’re not looking out, you’re delegating. And everyone knows that delegation with out transparency results in dependence.
Who audits the logic behind your agent’s decisions? What information informs them? What financial incentives bias them? The extra the interface simplifies, the extra opaque the underlying course of turns into.
We’ve spent 20 years complaining about algorithmic black containers in search and social media. Now we’re about to construct one round each digital choice we make.
Compute as a barrier, distribution as seize
The AMD alliance and the mini-apps announcement are two halves of the identical technique. Compute is the barrier to entry, distribution is the mechanism of seize.
By securing huge power and chip capability, OpenAI ensures that no competitor can simply match its scale. By embedding itself because the interface to different apps, it ensures that even when rivals exist, they’ll should undergo its ecosystem to achieve customers. It’s the basic Silicon Valley playbook, executed with breathtaking velocity and a layer of AI pixie mud.
Altman discovered from one of the best. He watched Apple, Google, and Fb flip management of interfaces into management of economies. Now he’s making use of the lesson to the age of brokers: personal the dialog, and also you personal the person.
The power query
The AMD deal additionally underscores an uncomfortable reality: large-scale AI is energy-intensive by design. Six gigawatts is roughly the output of six nuclear reactors. Coaching and operating superior fashions already devour staggering quantities of energy. What occurs when the world’s hottest interface can be considered one of its greatest electrical energy patrons?
OpenAI isn’t just constructing software program: it’s constructing infrastructure with a carbon footprint and geopolitical penalties. When a non-public firm begins locking up gigawatts of technology capability, regulators ought to deal with it not as a startup, however as a utility.
The governance hole
Each platform shift creates governance lags: guidelines arrive years after dominance is established. That’s how we ended up with app-store monopolies, ad-tech cartels, and search markets price trillions, however accountable to nobody.
ChatGPT’s platformization is occurring sooner than any earlier transition. And regulators, distracted by content material moderation and copyright disputes, appear utterly unprepared.
The dangers should not theoretical. As soon as an agent acts in your behalf (reserving journey, recommending purchases, even making hiring selections) it is going to be inconceivable to disentangle comfort from manipulation. The extra we outsource judgment to machines, the better it turns into for individuals who personal the machines to form our habits.
What occurs subsequent
The momentum is simple. OpenAI is shopping for computing, embedding companions, and positioning ChatGPT because the entrance finish of the whole lot. The monetary press reads it as a triumph of execution. The tech business reads it because the daybreak of agentic computing. Each could also be proper.
However beneath the joy, there’s a warning written within the footnotes of tech historical past. Each time a platform guarantees frictionless integration, it finally ends up centralizing energy. Each time we predict “this one can be totally different,” it isn’t.
I’m not yet another European obsessive about regulating the whole lot, I’m simply sufficiently old to recollect a number of earlier experiences akin to this one. The world doesn’t want one other working system that mediates entry to the whole lot: it wants transparency, interoperability, and competitors. If we don’t insist on them now, we could discover ourselves dwelling inside probably the most highly effective black field ever constructed: one which doesn’t simply reply our questions, however quietly decides which of them we’re allowed to ask. Be warned.
