Billion-dollar infrastructure deals power AI boom

Running an AI product requires a lot of computing power — and while the tech industry races to harness the power of AI models, there’s a parallel race underway to build the infrastructure that will power them. In a recent earnings call, Nvidia CEO Jensen Huang estimated that between $3 trillion and $4 trillion will be spent on AI infrastructure by the end of the decade — with much of that money coming from AI companies. Along the way, they are putting enormous pressure on power grids and pushing industry building capacity to its limits.

Below, we’ve laid out everything we know about the biggest AI infrastructure projects, including big spending from Meta, Oracle, Microsoft, Google, and OpenAI. We will keep it updated as the boom continues and the numbers grow.

Microsoft’s 2019 investment in OpenAI

This is arguably the deal that launched the entire contemporary AI boom: In 2019, Microsoft invested $1 billion in a buzzy non-profit called OpenAI, known mostly for its association with Elon Musk. Most importantly, the deal made Microsoft the exclusive cloud provider for OpenAI — and as model training requirements became more intense, more of Microsoft’s investment began to come in the form of Azure cloud credit rather than cash.

It was a great deal for both sides: Microsoft was able to claim more Azure sales, and OpenAI got more money for its largest single expense. In the years since, Microsoft has built its investment to nearly $14 billion — a move that is expected to pay off in a big way when OpenAI turns into a for-profit company.

The partnership between the two companies recently ended. Last year, OpenAI announced that it would no longer exclusively use Microsoft’s cloud, instead giving the company right of first refusal for future infrastructure requirements but pursuing others if Azure can’t meet their needs. Microsoft has also begun to explore other basic models for operating its AI products, leading to more independence from the AI ​​giant.

OpenAI’s arrangement with Microsoft has been so successful that it has become common practice for AI services to sign on with a particular cloud provider. Anthropic has received an $8 billion investment from Amazon, with kernel-level modifications to the company’s hardware to make it more suitable for AI training. Google Cloud has also contracted with smaller AI companies such as Lovable and Windsurf as “core compute partners,” although these deals did not involve any investment. Even OpenAI is back on its feet, receiving a $100 billion investment from Nvidia in September, giving it the ability to buy more of the company’s GPUs.

The rise of the oracle

On June 30, 2025, Oracle disclosed in a Securities and Exchange Commission filing that it had signed a $30 billion cloud services deal with an unnamed partner; This is more than the company’s cloud revenue for the entire previous fiscal year. OpenAI was eventually revealed as a partner, securing Oracle a place alongside Google as one of a series of post-Microsoft OpenAI hosting partners. Unsurprisingly, the company’s shares rose significantly.

A few months later, it happened again. On September 10, Oracle unveiled a five-year, $300 billion deal for computing power, set to begin in 2027. Oracle stock soared even higher, briefly making its founder Larry Ellison the richest man in the world. The sheer size of the deal is staggering: OpenAI doesn’t have $300 billion to spend, so the number assumes massive growth for both companies, and more than a little faith.

But before a single dollar has been spent, the deal has already cemented Oracle as one of the leading providers of AI infrastructure — and a financial force to be reckoned with.

Investment spree in Nvidia

As AI labs scramble to build infrastructure, they mostly buy GPUs from one company: Nvidia. The trade has left Nvidia flush with cash — and it’s investing that money back into the industry in increasingly unconventional ways. In September 2025, Nvidia bought a 4% stake in rival Intel for $5 billion — but what’s even more surprising are the deals with its customers. One week after the Intel deal was revealed, the company announced a $100 billion investment in OpenAI, driven by GPUs that will be used in OpenAI’s ongoing data center projects. Nvidia has since announced a similar deal with Elon Musk’s xAI, and OpenAI has launched a separate GPU-for-stock arrangement with AMD.

If this sounds circular, that’s because it is. Nvidia GPUs are valuable because they’re so rare — and by trading them directly into an ever-inflating data center chart, Nvidia is making sure they stay that way. You can say the same about privately held OpenAI stock, which is more valuable because it can’t be acquired through public markets. Right now, OpenAI and Nvidia are performing highly and no one seems too worried – but if momentum starts to wane, this type of arrangement will come under more scrutiny.

Building tomorrow’s hyperscale data centers

For companies like Meta that already have significant legacy infrastructure, the story is more complicated – although just as expensive. Meta CEO Mark Zuckerberg said the company plans to spend $600 billion on US infrastructure through the end of 2028.

In the first half of 2025, the company spent $30 billion more than the previous year, driven largely by the company’s growing AI ambitions. Some of that spending goes toward big cloud contracts, like a recent $10 billion deal with Google Cloud, but more resources are being pumped into two massive new data centers.

A new 2,250-acre site in Louisiana, called Hyperion, will cost an estimated $10 billion and provide an estimated 5 gigawatts of computing power. Notably, the site includes an arrangement with a local nuclear power plant to handle the increased power load. A smaller site in Ohio, called Prometheus, is expected to come online in 2026, powered by natural gas.

This type of construction comes with real environmental costs. Elon Musk’s xAI company has built its own hybrid data center and power plant in South Memphis, Tennessee. The plant quickly became one of the largest emitters of smog-producing chemicals in the county, thanks to a series of natural gas turbines that experts say violate the Clean Air Act.

Moon shot in the Stargate

Just two days after his second inauguration last January, President Trump announced a joint venture between SoftBank, OpenAI, and Oracle, aiming to spend $500 billion to build artificial intelligence infrastructure in the United States. Named “Stargate” after the 1994 film, the project has generated an incredible amount of hype, with Trump describing it as “the largest AI infrastructure project in history.” OpenAI’s Sam Altman seems to agree, saying: “I think this will be the most important project of this era.”

Overall, the plan was for SoftBank to provide the financing, with Oracle handling the build process with input from OpenAI. All of this was supervised by Trump, who promised to remove any regulatory obstacles that might slow down the construction process. But there were skepticism from the beginning, including from Elon Musk, Altman’s business rival, who claimed the project did not have the funds available.

As the hype subsided, the project lost some momentum. In August, Bloomberg reported that the partners had failed to reach a consensus. However, the project has moved forward with the construction of eight data centers in Abilene, Texas, with construction of the final building scheduled to be completed by the end of 2026.

Capital expenditure crisis

“Capex” is usually a very dry metric, referring to a company’s spending on physical assets. But with technology companies lining up to report their capital expenditure plans for 2026, the rush in data center spending has made the numbers more interesting — and much larger.

Amazon was the leader in capital spending, forecasting $200 billion in spending in 2026 (up from $131 billion in 2025), while Google came in second with an estimate of between $175 billion and $185 billion (up from $91 billion in 2025). Meta estimated $115 billion to $135 billion (versus $71 billion the previous year), although that number is a bit deceptive because many data center projects have been taken off its books entirely. All told, hyperscalers plan to spend nearly $700 billion on data center projects in 2026 alone.

This amount was enough to panic some investors. However, companies were mostly undeterred, explaining that AI infrastructure was vital to their companies’ future. A strange dynamic has been created. As you might expect, tech executives are more optimistic about AI than their counterparts on Wall Street — and the more tech companies spend, the more nervous bankers become. Add to that the massive amounts of debt that many companies are taking on to finance these construction operations, and you start to hear CFOs all over the Valley grinding their teeth.

This hasn’t hindered AI spending yet, but it will soon, unless superscalers show they can make these investments pay off.

This article was first published on September 22.