The huge demand for compute and energy from frontier AI developers like OpenAI, Anthropic, and Google is forcing those companies to buy and lease resources from their former competitors, creating a messy web of interdependence of von Bismarckian proportions. It has also resulted in a complex network of circular moneymaking mechanisms. The most common goes something like this: Company A buys AI chips from Company B, which then invests in Company A to boost its AI efforts, fund its data center expansion plans, and help meet demand for Company B’s chips. The carousel of capital keeps rotating from there.
For the past few years, Nvidia has been at the center of the ever-evolving web of circular dealmaking that has come to encompass the AI industry. Chipmakers’ graphics processing units (GPUs) have been and will continue to be the most valuable building blocks of the AI boom. It is also investing heavily in AI startups like OpenAI and Elon Musk’s XAI, which was acquired by SpaceX in February. So when AI labs buy AI chips from Nvidia, Nvidia provides funding to build the massive data centers that house the chips and ultimately power models like ChatGPT and Grok.
In November, it was announced Anthropic will pay $30 billion for expanded access to Microsoft’s cloud computing arm Azure, which is powered by graphics processing units (GPUs) made by Nvidia. In return, Nvidia and Microsoft said they would invest up to $10 billion and $5 billion, respectively, in Anthropic. (See circularity?) The extra compute helped Anthropic meet demand from businesses using its popular enterprise tools and developers using its popular coding assistant, Cloud Code. It took over OpenAI last month Highest Valued AI Startup In this world.
bloomberg published a supporting graphic January envisioned a web of circular deals being woven across the AI industry. But since that time the number of such agreements has been continuously increasing. For example, in March, Amazon pledged $50 billion to OpenAI as part of the ChatGate-maker’s most recent funding round, while OpenAI said it would invest an additional $100 billion in its existing partnership with Amazon Web Services (AWS), Amazon’s cloud computing business. Behind the scenes, Nvidia is selling AI chips to both Amazon and OpenAI, boosting the fortunes of all three companies.
you get the idea.
Sand foundation?
Technology leaders argue that increasing circularity throughout the industry is a virtuous cycle that benefits everyone. But skeptics see a noose that is tightening around the tech sector and perhaps even around the neck of the US economy.
The companies driving the AI boom, like OpenAI and Anthropic, have yet to become profitable, and the surge in investment across the industry in recent years has been a gamble that the technology will deliver the financial growth the companies are promising. The worry is that the more these companies become intertwined, moving money back and forth for technology that has not yet provided real value to businesses, the potential for financial ruin increases.
If demand for AI fails to grow the way people like Sam Altman, Dario Amodei, and Elon Musk have predicted, many of the huge data centers built over the past few years will suddenly become useless. AI laboratories will suddenly be unable to pay their bills to the companies that provide them with chips and computers; And since those same companies were the major backers of the rest of the industry, the entire edifice could collapse – skeptics warn – into the sand foundation on which it was built.
The economic impact of such a collapse could reach far beyond Silicon Valley. Increasing investment in tech companies driving the AI race in recent years has become a key financial driver for the US economy, helping boost the country’s GDP at a time when it is fighting the dual headwinds posed by inflation and tariffs. The financial security of millions of Americans through common financial instruments like 401(k) plans is also directly tied to the fortunes of the tech industry.
spacex ipo rocket
On Thursday, SpaceX—officially called Space Exploration Technologies Corp.—announced the finalization of its long-awaited initial public offering (IPO) priced at $135 per share. It is the largest IPO ever in history and officially makes its founder and CEO Musk the world’s first trillionaire.
SpaceX is perhaps best known as a maker of rockets and satellites (Starlink is a subsidiary), but it has also become a major force in the AI industry through its data center infrastructure. It is selling the computations created by these data centers to some of the same companies that are believed to be competing with xAI.
For example, anthropological announced Last month it agreed to pay $1.25 billion a month for access to SpaceX’s Colossus 1 data center based in Memphis, Tennessee, which SpaceX called “the world’s largest AI supercomputer.” Through the deal, Anthropic said it will gain access to 300 megawatts of power powered by 220,000 Nvidia GPUs. And last week, SpaceX said it had signed a multiyear deal with Google, through which Google will pay $920 million per month for compute to power its AI efforts. (The collaboration is expected to begin in October and extend through July 2029.)
Meanwhile, Anthropic and Google are deepening their own financial ties: In April, Google said it would Invest up to $40 billion At Anthropic, which recently teased the Mythos, a model was deemed too powerful to release publicly. And according to a report published on Thursday InformationAnthropic is also moving forward with plans to lease data centers from several US companies, with Google potentially stepping in as a guarantor.
Nvidia, Amazon, Google and other tech giants that are providing new AI startups with computing power are often referred to as the “picks and shovels” companies of the AI boom, referring to the industry that benefited from miners’ dreams of striking it rich during the California Gold Rush. Most miners did not extract gold, but those providing picks and shovels profited from a widespread frenzy of speculation and greed. By investing in its data center infrastructure and making lucrative deals with Google and Anthropic, SpaceX is clearly trying to become a major pick-and-shovel company in the AI race.
Much of the initial excitement from investors in the SpaceX IPO was driven by Musk’s promise to build data centers in space, powered by an unlimited supply of solar energy. It’s the kind of futuristic vision – along with Musk’s goal of colonizing Mars – that attracts many investors who (not unreasonably) believe in his ability to deliver on seemingly impossible promises. But it also remains to be seen whether data centers in space are technically available. With the future of the AI industry uncertain, in which SpaceX has hinged its fortunes, its stock is starting to look like a riskier investment than its initial market surge did.
Yes, SpaceX’s rocket under evaluation has twin engines from Google and Anthropic — two MVPs of the AI boom — powering it. But the circularity of the relationships between these companies makes all their situations more uncertain than they initially appear. If the rocket comes crashing back to Earth, it could bring with it more than just Google and Anthropic.
<a href