It takes a lot of computing power to run an AI product – and as the tech industry races to power AI models, there’s a similar race going on to build the infrastructure to 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 — most of that money coming from AI companies. Along the way, they put a huge strain on power grids and push the industry’s capacity to build 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 you informed as the boom continues and the numbers go up.
Microsoft’s 2019 investment in OpenAI
This is the deal that started all modern AI developments: In 2019, Microsoft made a $1 billion investment in a non-profit company called OpenAI, which is well known for its association with Elon Musk. Worse, the deal made Microsoft the exclusive cloud provider for OpenAI — and as demands for model training grew stronger, most of Microsoft’s investment began to come in the form of Azure cloud credits instead of cash.
It was a win-win for both sides: Microsoft was able to claim more Azure sales, and OpenAI got more money at its biggest expense. In the following years, Microsoft will build its investment up to $14 billion – a move that will pay off big when OpenAI turns into a profitable company.
The partnership between the two companies has ended recently. Last year, OpenAI announced it would no longer use Microsoft’s cloud exclusively, instead giving the company right of first refusal on future infrastructure needs but pursuing others if Azure can’t meet their needs. Microsoft has also started to explore other basic models to power its AI products, establishing 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 up with a particular cloud provider. Anthropic received $8 billion in funding from Amazon, while making kernel-level changes to the company’s hardware to make it better suited for AI training. Google Cloud has also signed on small AI companies such as Lovable and Windsurf as “major computing partners,” although those agreements do not include any investment. And even OpenAI is back at the source, getting a $100 billion investment from Nvidia in September, giving it the power to buy more of the company’s GPUs.
The rise of the Oracle
On June 30, 2025, Oracle revealed in an SEC filing that it signed a $30 billion cloud services deal with an unnamed partner; this is more than the company’s revenue for the entire previous financial year. Finally OpenAI was revealed as a partner, securing Oracle a place alongside Google as one of OpenAI’s series of post-Microsoft hosting partners. Unusually, the company’s stock rose.
Techcrunch event
Boston, MA
|
June 9, 2026
A few months later, it happened again. On September 10, Oracle unveiled a five-year, $300 billion deal to acquire computing power, which will begin in 2027. Oracle’s stock soared, making founder Larry Ellison the richest man in the world. The overall size of the deal is surprising: OpenAI is not worth $300 billion, so the figure takes into account significant growth for both companies, and little faith.
But before a single dollar is spent, the deal already cements Oracle as one of the leading AI infrastructure providers — and a financial force to be reckoned with.
Nvidia investment
As AI labs try to build infrastructure, they mostly buy GPUs from one company: Nvidia. That trade left Nvidia flush with cash — and it’s been investing that money back into the industry in unconventional ways. In September 2025, Nvidia bought a total of 4% of its rival Intel for $ 5 billion – but the most surprising were the agreements with its customers. One week after the Intel deal was revealed, the company announced a $100 billion investment in OpenAI, paid for by GPUs to be used in OpenAI’s ongoing data center projects. Nvidia has announced a similar agreement with Elon Musk’s XAI, and OpenAI is launching a separate GPU-for-stock program with AMD.
If that seems circular, that’s because it is. Nvidia’s GPUs are important because they’re so rare – and by selling them directly into the ever-increasing data center scheme, Nvidia is making sure they stay that way. You can say the same about OpenAI’s private stock, which is very valuable because it cannot be acquired through public markets. At the moment, OpenAI and Nvidia are riding high and nobody seems too worried – but if momentum starts to flag, this kind of program will be under more scrutiny.
Building the hyperscale data centers of tomorrow
For companies like Meta that already have significant legacy infrastructure, the story is even more complicated – though equally 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 dollars more than last year, mainly driven by the growing AI ambitions of the company. Some of that spending is going toward big-ticket cloud contracts, like the recent $10 billion deal with Google Cloud, but more resources are being poured into two giant data centers.
The new 2,250-acre site in Louisiana, called Hyperion, will cost about $10 billion to build and provide an estimated 5 gigawatts of computing power. Notably, the site includes a facility with a local nuclear power plant to handle the additional energy load. A small site in Ohio, called Prometheus, is expected to be online in 2026, powered by natural gas.
That kind of construction comes with real environmental costs. Elon Musk’s xAI is building its own hybrid data center and power generation facility in South Memphis, Tennessee. The facility has become one of the largest smog-producing chemical plants in the state, thanks to a series of natural gas engines that experts say violate the Clean Air Act.
Stargate moon image
Just two days after his second inauguration last January, President Trump announced a partnership between SoftBank, OpenAI, and Oracle, which aimed to spend 500 billion dollars to build AI infrastructure in the United States. Named “Stargate” after the 1994 film, the project came with an incredible amount of hype, with Trump calling it “the largest AI infrastructure project in history.” OpenAI’s Sam Altman seemed to agree, saying, “I think this is going to be the most important project at this time.”
With broad contributions, the plan was for SoftBank to provide the funding, Oracle to handle the buildout with input from OpenAI. Overseeing it all was Trump, who promised to remove any regulatory hurdles that might slow construction. But there were skeptics from the beginning, including Elon Musk, Altman’s business rival, who said the project did not have the funds available.
As the hype died down, the project lost momentum. In August, Bloomberg reported that the partners were failing to reach an agreement. Nevertheless, the project has moved forward with the construction of eight data centers in Abilene, Texas, and the construction of the last building will be completed by the end of 2026.
The capex crunch
“Capital expenditure” is usually a very dry metric, referring to a company’s spending on physical assets. But as tech companies line up to report their capex plans for 2026, rapid data center spending made the numbers even more interesting — and even bigger.
Amazon was the capex leader, projecting $200 billion in spending by 2026 (up from $131 billion in 2025), while Google was a close second with an estimate of between $175 billion and $185 billion (up from $91 billion in 2025). Meta estimates it at $115 billion to $135 billion (from $71 billion last year), although that figure is a little misleading because many data center projects are kept in their books entirely. In total, hyperscalers plan to spend nearly $700 billion on data center projects in 2026 alone.
It was enough money to distract some investors. The companies did not give up, however, explaining that AI infrastructure is essential to the future of their companies. It is set to a random variable. As you might expect, tech executives are more bullish on AI than their Wall Street counterparts — and the more tech companies use, the more nervous their bankers are. Add in the huge amounts of debt many companies take on to fund those structures, and you start to hear the CFOs in the village gnashing their teeth.
That hasn’t put an end to the use of AI yet, but it will soon — unless, of course, hyperscalers show they can make those investments pay off.
This article was originally published on September 22.
