Amazon landed a massive AI infrastructure deal. OpenAI will spend $38 billion to access hundreds of thousands of Nvidia AI chips through Amazon Web Services.
Under the partnership, OpenAI runs its AI workloads on Nvidia GPUs housed within Amazon's AI servers. The companies said OpenAI would immediately start using AWS infrastructure, scaling to full capacity by end of 2026.
Amazon stock jumped roughly 5% Monday morning on the news.
This is just the latest in OpenAI's string of massive infrastructure deals. The AI company also has:
OpenAI is committing staggering sums to secure computing capacity for training and running its AI models. The scale of spending raises serious questions about the company's financial sustainability.
These circular AI deals are fueling fears of a market bubble.
Here's the circular flow: Tech companies invest in AI startups. Those startups spend billions buying cloud services from the same tech companies. The tech companies then point to surging demand as justification for massive AI infrastructure spending.
OpenAI's costs are projected to surpass $1 trillion by the end of the decade. Meanwhile, its revenue remains far below that level. How does a company with costs approaching $1 trillion and revenue nowhere close to that figure pay for all these infrastructure commitments?
OpenAI announced last week it's transforming into a for-profit public benefit corporation, paving the way to go public. The company is reportedly laying groundwork for an IPO that would value it at $1 trillion, according to Reuters.
A $1 trillion valuation would need to be justified by massive future profits. But with costs outpacing revenue by enormous margins, that valuation looks optimistic at best.
Amazon reported strong Q3 earnings last week that pushed its stock to all-time highs. The OpenAI deal adds to that momentum.
AWS also announced last Wednesday it completed a massive AI data center project and will provide OpenAI rival Anthropic with 1 million custom AI chips by end of 2025.
Amazon is positioning itself as the infrastructure provider for the AI boom, whether that boom is real or inflated.
Amazon's $38 billion OpenAI deal looks great for AWS - massive guaranteed revenue from a high-profile customer.
But the broader picture raises red flags. OpenAI is committing hundreds of billions in infrastructure spending across multiple providers. Its costs are heading toward $1 trillion while revenue lags far behind.
The circular nature of AI deals creates an echo chamber. Microsoft invests billions in OpenAI. OpenAI spends billions with Microsoft, Amazon, Oracle, and others for cloud services. Those cloud providers tout surging AI demand. Investors pile into AI stocks. Repeat.
This works beautifully as long as AI adoption justifies the spending. If AI revenue disappoints, the whole structure becomes precarious.
OpenAI planning a $1 trillion IPO while burning cash at unprecedented rates feels reminiscent of dot-com era excess. High valuations based on future promise rather than current profitability.
For Amazon, this is mostly upside. The company gets $38 billion in guaranteed revenue whether OpenAI's business model works or not. AWS benefits from the AI infrastructure buildout regardless of whether AI delivers promised returns.
The 5% stock pop shows investors approve. Amazon is selling shovels during a gold rush - a historically profitable position.
But for the broader AI ecosystem, OpenAI's spending spree highlights how capital-intensive this technology has become. The company needs over $1 trillion in infrastructure spending to compete. That's an enormous bet on AI delivering transformative returns.
Whether that bet pays off determines if this is the foundation of a new technological era or an expensive lesson in irrational exuberance.