Artificial Intelligence & Machine Learning
,
Next-Generation Technologies & Secure Development
AWS to Build Server Clusters, Nvidia to Supply Chips for 7 Years
Loss-making OpenAI added to a string of deals with a $38 billion commitment on Monday to using compute resources provided by Amazon Web Services.
See Also: OnDemand | Navigate the threat of AI-powered cyberattacks
The artificial intelligence giant said AWS will build out server clusters using Nvidia flagship Blackwell chips for use by OpenAI over the next seven years.
Shares of Nvidia and Amazon went up during Monday morning trading, gaining roughly 3% and 5%, respectively. OpenAI, which completed a recapitalization and corporate restructuring in late October, is reportedly laying the groundwork for an initial public offering that could value the company up to $1 trillion (see: OpenAI Restructures, Nonprofit Foundation Retains Control).
The AWS deal comes just days after the firm eliminated a right of first refusal by Microsoft to supply compute services. That provision itself was a downgrade of what earlier was an exclusive relationship requiring OpenAI to use the Azure cloud. Microsoft has pledged $13 billion to the AI company, of which it funded $11.6 billion as of Sept. 30, it told investors in a quarterly filing.
OpenAI spent the summer and fall signing a slew of deals The Financial Times calculates are worth up to $1.5 trillion, including a commitment to spend as much as $350 billion on deploying 10 gigawatts of Nvidia chips in exchange for a $100 billion investment. It signed a $300 billion deal with Oracle to buy 4.5 gigawatts of cloud compute from Oracle and commitment to buy 6 gigawatts worth of chips from AMD in a transaction letting it purchase up 160 million shares of the chipmaker’s stock for one cent a share. Earlier this month, it said it will collaborate with Broadcom on manufacturing 10 gigawatts of custom AI chips.
The frenzy of deal making has provoked concerns that the AI boom amounts to a circular economy blurring the lines between consumers and investors of AI infrastructure. “Nvidia invests $100 billion in OpenAI, which then OpenAI turns back and gives it back to Nvidia,” a hedge fund executive told CNBC in September.
OpenAI itself continues to lose more money than it makes each year, with The Information reporting that the firm posted an operating loss of $7.8 billion and a net loss of $13.5 billion during the first half of this year. Analysis of Microsoft earnings suggested that OpenAI may have racked up a $12 billion loss in just the third quarter.
Worries that AI is caught up in a speculative boom fueled by self-perpetuating feedback loops likely weren’t eased by an August assessment of generative AI published by The Massachusetts Institute of Technology concluding that only 5% of AI pilot programs produce returns. The finding tracks with other reports that corporate executives are losing faith in AI. One survey of global chief technology officers found that confidence among C-Suite leaders has dropped 11 percentage points in the past year – down to 58% in 2025 from 69% in 2024.
OpenAI CEO Sam Altman has appeared unperturbed by mounting costs and skeptics. “Whether we burn $500 million a year or $5 billion or $50 billion a year, I don’t care. I genuinely don’t,” he told Stanford University students in May 2024. “It’s going to be expensive. It’s totally worth it.”
