WritingResources
Back to all posts

OpenAI’s unchecked expansion

Ludivine Siau

In short

No summer break for OpenAI. The company has spent the last three months moving at a breathtaking pace, dropping a flood of product releases and infrastructure deals in a calculated all-at-once push to build a vertically integrated empire. From chips and data centres to frontier research to consumer apps and a developer ecosystem: the OpenAI platform vision is here. And it raises critical questions about market control, safety and what the ultimate endgame is for the AI industry.

Quick breakdown

Fort Mason, San Francisco, 25th of March 2015. Mark Zuckerberg announces on stage at Facebook’s annual F8 developer conference that the company is opening up its Messenger app to app developers and businesses. One more step to expand on Facebook’s platform vision.

Ten years later, Sam Altman stands on the same stage for OpenAI’s DevDay, and the echo with Facebook’s old ambition is uncanny. OpenAI is making its own platform play, and the event is only the latest drop in a torrent of moves from the company. To say they've been busy is the understatement of the year.

Here’s a look at their last three months:

21 Jul: Deal with the UK Government to use OpenAI in public services

22 Jul: Agreement with Oracle to build an additional 4.5GW data centre capacity in the U.S. as apart of Stargate

29 Jul: Release of Study Mode in ChatGPT

31 Jul: Announcing Stargate Norway, OpenAI’s first AI data center initiative in Europe, with a plan to initially deliver 230MW of data centre capacity

6 Aug: Partnership with the US General Services Administration to offer access to all federal agencies to ChatGPT for $1 for a year

7 Aug: GPT-5 release

25 Aug: Silicon Valley AI leaders and investors launch a new super-PAC focused on advocating against AI regulation, Leading the Future. Greg Brockman, OpenAI’s President, is backing the PAC.

26 Aug: Family of teen who died by suicide sues OpenAI, blaming the company’s chat for harmful responses

2 Sep: Announcing product improvements to better respond to signs of users’ mental and emotional distress

2 Sep: Acquisition of Statsig, experimentation platform; Founder Vijaye Raji appointed CTO of Applications.

4 Sep: Announcing development of the OpenAI Jobs Platform, a hiring platform for AI talent, and the OpenAI Certifications with the commitment to certify 10 million Americans by 2030

8 Sep: $50M People-First AI Fund opens for applications “to support frontline and mission-focused organizations working at the intersection of innovation and public good.”

15 Sep: GPT-5-Codex becomes available in research access

16 Sep: Announcing Stargate UK, an AI infrastructure partnership with NVIDIA and Nscale to grow compute capabilities in the UK

22 Sep: Partnership with NVIDIA to build and deploy 10GW of NVIDIA-equipped AI datacenters

23 Sep: Five new U.S. sites announced for Stargate’s datacenter buildout

24 Sep: Partnership with SAP to develop a sovereign “OpenAI for Germany”

25 Sep: ChatGPT Pulse release, a home page with personalised prompts and updates

29 Sep: Instant Checkout and Agentic Commerce Protocol release, allowing purchases directly in ChatGPT

29 Sep: Introduction of Parental Controls and age prediction capabilities in ChatGPT

30 Sep: Sora 2 video and audio generation model released alongside a new social app, Sora

Oct 1: Partnership with Samsung and SK to supply advanced memory chips

Oct 2: OpenAI becomes the most valuable private company at an estimated $500 billion

Oct 6: Partnership with AMD to deploy 6GW of AMD chips

Oct 6: DevDay 2025:

  • AgentKit release, including a drag-and-drop AgentBuilder tool
  • Codex public release
  • Introducing apps in ChatGPT and a new Apps SDK

Oct 7: The innocent announcement that OpenAI is using custom AI tools internally causes several tech companies’ stock value to dip

Why you should care

We’ve seen tech giants before, but OpenAI’s ambition is on another scale. They want to become the foundational layer and the home for everything that people have to do with AI, what Ben Thompson from Stratechery calls "The Windows of AI". In a recent interview, Sam Altman made the goal clear: to create a "continuity of experience" where one AI service powers everything. That's a vision for 24/7, lifelong user engagement that would make even the social media titans blush.

From OpenAI’s perspective, this relentless push makes sense: the compute, the research and the consumer demand must all be built at once. In Sam Altman’s head, there’s no feasible, slower sequence of action that can get AI where it needs to go. It’s all or nothing.

This creates a precarious ecosystem. Billions are flowing between a handful of players (NVIDIA, Microsoft, SoftBank, Oracle, OpenAI) in a high-stakes bet that almost feels circular. The entire AI industry is becoming dangerously dependent on one company’s success. If OpenAI fails to deliver on any front, all the deals come crashing, bringing down chipmakers, app developers, and investors.

The influence campaign is also deeply political. OpenAI is actively courting governments across the world to shape the regulatory landscape to its advantage and get their products embedded in the running of public services.

Most concerning is the persistent "release and see what happens" ethos. The timeline above shows a familiar pattern: launch a powerful new technology, wait for the impact to manifest (sometimes tragic…), and then roll out reactive measures to somewhat correct the situation. Altman maintains a public persona of innocence, suggesting that ChatGPT is just "genuinely trying to be helpful" and that OpenAI never anticipated its market-shaking impact. For anyone with a background in tech, that sounds either deeply naive or deliberately disingenuous. We've seen this play out before with social media. The public is once again the testing ground for a powerful and unpredictable technology, and the consequences are (barely) being dealt with after the fact.

The AI industry is at a crossroads. As Kanjun Qiu, CEO of Imbue, noted on the Decoder podcast, the future could go one of two ways: an ecosystem controlled by a single platform on a scale never seen before, or one that puts power back in the hands of users. OpenAI is clearly building for the former. Can we even nudge this the other way?

Sources