It all began, as many things do, with Elon Musk. In the early 2010s he realized that AI was on a track to become perhaps the most powerful technology of all time. But he had deep suspicion that if it were to fall under the control of powerful profit-driven forces, humanity would suffer. Musk had been an early investor in DeepMind, the UK-based lab that was ahead of the pack in pursuing artificial general intelligence. After Google bought DeepMind in 2014, Musk cut ties with the research organization. He felt it was essential to create a counterforce incentivized by human benefit, not profits. So he helped create OpenAI. When I interviewed Musk and Sam Altman at the company’s unveiling in 2015, they were adamant that shareholder profit would not be a factor in their decisions.

Fast-forward to today. OpenAI is worth a half a trillion dollars, or maybe $750 billion, and its for-profit arm has become a public benefit corporation. Musk, the world’s richest human, runs his own for-profit AI company, xAI. So much for nonprofit labs leading the way. But even the most panicked Cassandra of a decade ago likely didn’t imagine that advanced AI would be controlled by a single, interlocking, money-seeking behemoth.

That’s what we have today. Even more concerning, this interlinked complex is funded in part by overseas powers and supported by the US government, which seems to prioritize winning over safety. This rococo collection of partnerships, mergers, funding arrangements, government initiatives, and strategic investments links the fate of virtually every big player in the AI-o-sphere. I call this entity the Blob.

Blob’s Black Box

A full description of the interlocking connections of these entities would take me way beyond my word limit here. Even compiling a streamlined list requires the use of—you guessed it—AI. Reader, I confess. I went to GPT-5 to help me get the full picture. ”My head is spinning,” I wrote, swallowing my pride to ask this smug stochastic parrot for a comprehensive list of cloud deals, investments, partnerships, and government arrangements. It took two minutes and 35 seconds before the normally speedy LLM came back with some answers. “You’re not wrong that it’s dizzying,” said the bot, ever the sycophant. “It’s basically one giant circular money-and-compute machine.” Note to GPT: You do not get to write the display text for this essay. Leave the editorializing to me. In any case, once it stopped its punditry, GPT-5 proceeded to churn out several thousand words, with flow charts, arrows, and cross-references to dozens of back-scratching arrangements like the iconic Stargate initiative that binds OpenAI, Oracle, Nvidia, Softbank, and an Abu Dhabi investment firm, with the US government’s support.

This week provided a fresher case in point: a complicated deal involving Nvidia, Microsoft, and Anthropic. Microsoft’s press release sums it up in three lines, like a subpar Allen Ginsberg verse: “Anthropic to scale Claude on Azure / Anthropic to adopt NVIDIA architecture / NVIDIA and Microsoft to invest in Anthropic.” The deal has the earmarks of what critics call a circular arrangement, whereby money goes back and forth between companies before a single customer is involved. Microsoft is investing at least $5 billion in Anthropic—a direct rival of Microsoft’s key partner OpenAI—and Anthropic has committed to buying $30 billion worth of compute from Microsoft’s cloud. Meanwhile, Nvidia invests in Anthropic, which commits to develop its technology on Nvidia chips. Poof! Nvidia gets deeper into the business of its customers. Microsoft gets a hedge from its earlier reliance on OpenAI. And Anthropic’s valuation jumps to $350 billion. (Only two months ago, it was valued at $183 billion.)

Anthropic didn’t comment on the deal beyond a press release, pointing journalists to a video where the three CEOs explain the deal. The hyperscale bosses participate remotely; these deals are so routine, it’s apparently not worth the bother to get on a plane to announce them in person. In the video, Microsoft’s Satya Nadella appears in the middle, beaming like the Cheshire cat as he invokes what might be the Blob’s slogan: “We are increasingly going to be customers of each other.” As he lays out the particulars, the others give bobblehead nods. On the left is Anthropic CEO Dario Amodei. Anthropic doesn’t have its own cloud or a non-AI-related revenue flow like Google, Microsoft, or Meta, so it’s now added Microsoft to its previous stock-for-compute deals with Amazon and Google. Hat trick!

Nvidia’s Jensen Huang, in his signature leather jacket, calls the deal “a dream come true,” elaborating that he’s been eyeing Anthropic for some time and is thrilled to add the company to his bulging deal book. “We’re in every enterprise in every single country,” he says. “Now this partnership of the three of us will be able to bring AI, bring Claude, to every enterprise, to every industry, around the world.”

Instead of unleashing a pack of rabid trustbusters on this crowd, the current administration is cheering it on. Even when hundreds of billions of dollars are on the line, the president isn’t asking “How might this harm the public?” as much as “How can I get a piece of that?” Now, Nvidia gets to sell chips to China. (Kicking back a piece of the profits to the US government.) The Saudis—who have funded numerous AI efforts in the US—agreed to guardrails to buy chips to create their own version of an AI Blob. Eventually, if things go well for them, the Saudi firms will compete with US companies.

Speaking of things going well or not, if the financial AI bubble pops or deflates, the existence of the Blob means that everyone goes down. As Google CEO Sundar Pichai put it this week, “no company is going to be immune, including us.”

Sympathy for the Developers

In a way you can’t blame these big companies. They would probably prefer to do their business without becoming clots in the Blob. But the pursuit of hyperscaled AI is too big for even the biggest firms to go it alone. “I don't think any of us realized how real the scaling laws would end up being, and therefore how much compute these things would require,” one executive at a key AI entity tells me. As a result, he says, “We've all turned into infrastructure and construction companies.” The massive capital cost of building miles of data centers requires partnerships. It also puts the companies that train and operate LLMs in a polyamorous situation with cloud providers.

So don’t think of the Blob as a classic cartel. There doesn't seem to be skulduggery such as price-fixing or similar forms of collusion. The rivalry among LLM creators is fierce. This week, Google rolled out its new and improved Gemini model, and seems to have nudged into the lead, at least until ChatGPT, Claude, or Grok gets a refresh. Nvidia’s motives are understandable as well: It is desperate to avoid commoditization, and locking in customers is simply good sense.

Still, consider the spectacle this week as some of the AI illuminati gathered at White House events to honor Mohammed bin Salman, the ruthless Saudi crown prince who is acting as the AI world’s Daddy Warbucks. Do those leaders trek to Washington for this and other Trump-led showcases to rep their individual companies? Of course. But they also attend as charter members of The Blob.

This is an edition ofSteven Levy’sBackchannel newsletter. Read previous newslettershere.