A look at OpenAI’s tangled web of dealmaking

OpenAI CEO Sam Altman is everywhere.

His artificial intelligence startup, now valued at $500 billion, has been inking deals valued in the tens to hundreds of billions of dollars with infrastructure partners, even as it continues to burn mounds of cash.

Those expenditures are driving the market.

The Nasdaq and S&P 500 rose to record highs this week after Nvidia agreed to invest up to $100 billion in OpenAI. That followed a $300 billion deal between OpenAI and Oracle in July as part of the the Stargate program, a $500 billion infrastructure project that’s also being funded by SoftBank.

Its commitments don’t stop there. CoreWeave on Thursday said it’s agreed to provide OpenAI up to $22.4 billion in AI infrastructure, an increase from the $11.9 billion it initially announced in March. Earlier this month, chipmaker Broadcom said it had secured a new $10 billion customer, and analysts were quick to point to OpenAI. 

While OpenAI says that scaling is key to driving innovation and future AI breakthroughs, investors and analysts are beginning to raise their eyebrows over the mindboggling sums, as well as OpenAI’s reliance on an increasingly interconnected web of infrastructure partners. 

OpenAI took a $350 million stake in CoreWeave ahead of its IPO in March, for instance. Nvidia formalized its financial stake in OpenAI by participating in a $6.6 billion funding round in October. Oracle is spending about $40 billion on Nvidia chips to power one of OpenAI’s Stargate data centers, according to a May report from the Financial Times. Earlier this month, CoreWeave disclosed an order worth at least $6.3 billion from Nvidia. 

And through its $100 billion investment in OpenAI, Nvidia will get equity in the startup and earn revenue at the same time.

OpenAI is only expected to generate $13 billion in revenue this year, according to the company’s CFO Sarah Friar. She told CNBC that technology booms require bold bets on infrastructure. 

“When the internet was getting started, people kept feeling like, ‘Oh, we’re over-building, there’s too much,'” Friar said. “Look where we are today, right?”

Altman told CNBC in August that he’s willing to run the company at a loss in order to prioritize growth and its investments. 

But some analysts are raising red flags, arguing that OpenAI’s deal with Nvidia is reminiscent of vendor financing patterns that helped burst the dot-com bubble in the early 2000s.

Nvidia has been the biggest winner of the AI boom so far because it produces the graphics processing units (GPUs) that are necessary to train models and run large AI workloads. Nvidia’s investment in OpenAI, which will be paid out in installments over several years, will help the startup build out data centers that are based around its GPUs. 

“You don’t have to be a skeptic about AI technology’s promise in general to see this announcement as a troubling signal about how self-referential the entire space has become,” Bespoke Investment Group wrote in a note to clients on Tuesday. “If NVDA has to provide the capital that becomes its revenues in order to maintain growth, the whole ecosystem may be unsustainable.” 

Peter Boockvar, chief investment officer at One Point BFG Wealth Partners, said names of companies from the late 1990′s were ringing in his ears after the OpenAI-Nvidia deal was announced. 

A key difference, however, is that this transaction is “so much bigger in terms of dollars,” he wrote in a note.

“For this whole massive experiment to work without causing large losses, OpenAI and its peers now have got to generate huge revenues and profits to pay for all the obligations they are signing up for and at the same time provide a return to its investors,” Boockvar said.

An OpenAI spokesperson referred CNBC to comments from Altman and Friar this week, adding that the company is pursuing “a once-in-a-century opportunity that demands ambition equal to the moment.”

The total amount of demand for compute could reach a staggering 200 gigawatts by 2030, according to Bain & Company’s 2025 Technology Report. Building enough data centers to meet this anticipated demand would cost about $500 billion a year, meaning AI companies would have to generate a combined $2 trillion in annual revenue to cover those costs.

Even if companies throw their whole weight behind investing in the cloud and data centers, “the amount would still fall $800 billion short of the revenue needed to fund the full investment,” Bain said.

There’s a clear uphill battle ahead, but OpenAI’s Altman brushed off concerns on Tuesday, rejecting the idea that the infrastructure spending spree is overkill.

“This is what it takes to deliver AI,” Altman told CNBC. “Unlike previous technological revolutions or previous versions of the internet, there’s so much infrastructure that’s required, and this is a small sample of it.”

–CNBC’s Yun Li and MacKenzie Sigalos contributed to this report

WATCH: OpenAI’s Sam Altman defends Stargate expansion as demand for AI soars