What scares Sam Altman is a handful of companies deciding who gets to use AI - a rogue machine doesn't worry him nearly as much. His own company can't build a single model without three of them.
Speaking on David Senra's Relentless podcast in an episode that drew wide coverage on August 23 and 24, the OpenAI CEO said the defining question of this moment is stark: "Are we going to head to a world of AI authoritarianism or liberty?" Big question. Altman told Senra that one of his biggest fears is that the most powerful AI systems end up gatekept by a small number of companies or countries. That outcome, he said, would be "very, very bad," according to a report from Yahoo News.
Altman didn't name Microsoft, Amazon, Google, Meta or Nvidia directly in the authoritarianism remarks, but those are the five firms that, alongside OpenAI, control most of the compute, cloud capacity and chip supply the entire AI industry runs on. He framed the stakes in civilizational terms. "It may be a little messy, but every time humanity has traded off its liberty for safety, it's been a long-term net loss," he said, according to SiliconANGLE. His stated goal, he added, has long been "abundant intelligence" without "a weird power concentration and kind of a new authoritarianism."
The gatekeepers he depends on #
Here's the awkward part. Altman is warning about power concentrating in a handful of companies while running one of them, and while depending on two of the others just to keep the lights on.
OpenAI trains and serves every model on Microsoft's Azure data centers and Nvidia's chips. No exceptions. Nvidia agreed last year to invest up to $100 billion in OpenAI as the company committed to deploying at least 10 gigawatts of Nvidia systems, a deal that ties OpenAI's entire compute roadmap to one supplier's chip pipeline. Microsoft remains OpenAI's largest cloud partner and one of its biggest shareholders. So when Altman warns against a small number of gatekeepers, he's describing, in part, his own supply chain.
Nvidia is now bankrolling the company that buys its chips and the numbers are getting hard to ignore
Nvidia is in talks to guarantee up to $250 billion for OpenAI's data center buildout in Ohio, with separate discussions covering $350 billion more in chip financing. The deal reignites circular-financing concerns as Nvidia's total AI deal exposure crossed $750 billion last week. - nvidia financing openai infrastructure deals - gpu supplier funding data center buyers
OpenAI also isn't the open one in this story. Microsoft, Nvidia and Meta have all pushed model makers to release open-weight systems that outside researchers and governments can inspect and run on independent hardware. OpenAI has largely kept its frontier models closed. Anthropic has gone further than Altman in the opposite direction on policy: CEO Dario Amodei has openly called for a bigger government role in AI regulation, a position Altman rejects. Altman argues that heavier regulation is itself a liberty-for-safety trade, and that the fix for concentrated corporate power isn't concentrated government power either.
That leaves Altman's argument in an odd spot. He wants less regulation and less corporate gatekeeping at the same time, without saying who actually enforces that balance if not government and not the companies holding the compute.
A trillion-dollar timeline #
None of this is happening in a vacuum. OpenAI is privately valued at roughly $852 billion after its March 2026 funding round, and Altman has told investors he won't take the company public below a $1 trillion valuation, according to reporting from the Motley Fool. That IPO push is now colliding with a separate deadline: SoftBank, one of OpenAI's largest backers, faces a $40 billion loan repayment in 2027, adding pressure to a listing Altman originally hoped to complete this year. A trillion-dollar AI company arguing against power concentration is, at minimum, a striking image.
Altman also told Senra he'd gotten the pace of AI's economic disruption wrong. When GPT-4 launched in 2023, he expected sweeping change to follow almost immediately. It didn't, at least not on that timeline. That admission matters here too, because it undercuts the urgency of his own framing. If the disruption is slower than he predicted, the window to decide who controls it is longer than his authoritarianism-or-liberty framing suggests, not shorter.
Frankly, the more useful question isn't whether Altman is right that concentration is dangerous. He probably is. It's whether a company chasing a trillion-dollar valuation, backed by Nvidia's balance sheet and running on Microsoft's servers, is the one positioned to prevent it.
Also read: SEC Subpoenas Wall Street Banks Over AI Hedge Fund Situational Awareness • Chinese State Hackers Doubled Their Attack Volume Using DeepSeek AI • Pew Research Finds a Third of Web Pages Written Since ChatGPT Show AI Signs