Running AI products requires a lot of computing power. As the technology industry taps on the power of AI models, parallel races are underway to build the infrastructure that powers them. Nvidia CEO Jensen Huang estimated in his recent revenue call that it will be spent on AI infrastructure by the end of the decade, between $3 trillion and $4 trillion. Along the way, they put a heavy strain on the power grid and push the capacity of industry buildings to the limit.
Below, we have laid out everything we know about the biggest AI infrastructure projects, including major spending from Meta, Oracle, Microsoft, Google, Openai. It will continue to be updated as the boom continues and the numbers rise further.
Microsoft’s $1 billion investment in Openai
This is definitely a transaction that launched the entire modern AI boom. In 2019, Microsoft invested $1 billion in a buzzy nonprofit called Open-E, which is primarily known for its association with Elon Musk. Importantly, this agreement has made Microsoft an exclusive cloud provider for Openai. As model training demands became more intense, an increase in Microsoft’s investments began to occur in the form of Azure Cloud credits rather than cash.
That was a big deal for both parties. Microsoft could charge more Azure sales, and Openai got more money with the largest single cost. Over the next few years, Microsoft will build up investments of up to $14 billion. This is set to be extremely rewarding when Openai transforms into a for-profit company.
The partnership between the two companies has recently been lifted. In January, Openai announced that it would no longer use Microsoft’s cloud exclusively, and instead grants the company the right to initially reject future infrastructure requests, but announced that it would pursue others if Azure failed to meet its needs. More recently, Microsoft has explored other basic models to enhance its AI products and further established independence from the AI giants.
Openai’s arrangement with Microsoft was so successful that it became a common practice for AI services to sign on to specific cloud providers. Anthropic has received $8 billion in investment from Amazon, but is now making kernel-level changes to the company’s hardware to make it suitable for AI training. Google Cloud has signed small AI companies like Lovable and Windsurf as “primary computing partners,” but these transactions do not include investments. Even Openai returned to the wells, receiving a $100 billion investment from Nvidia in September, giving it the ability to buy more companies’ GPUs.
The rise of Oracle
On June 30, 2025, Oracle announced in a filing of the SEC that it had signed a $30 billion cloud service agreement with an unnamed partner. This exceeds all companies’ cloud revenues for the previous fiscal year. Openai was eventually revealed as a partner, and Oracle secured the spot along with Google as one of Openai’s post-microsoft hosting partners. Naturally, the company’s stocks were shot.
TechCrunch Events
San Francisco
|
October 27th-29th, 2025
A few months later, it happened again. On September 10th, Oracle revealed its five-year, $300 billion contract scheduled to begin in 2027. Oracle’s stocks have risen even higher, making founder Larry Ellison the richest man in the world. The scale of the contract is incredible. Openai doesn’t cost $300 billion. So this figure estimates great growth and a bit of faith for both companies.
But before a dollar is spent, the deal already solidifies Oracle as one of the leading AI infrastructure providers, taking into account financial forces.
We will build a hyperscale data center tomorrow
For companies like Meta, who already have important legacy infrastructure, the story is more complicated, but equally expensive. Mark Zuckerberg said Meta is planning to spend $600 billion on US infrastructure by the end of 2028.
In the first half of 2025, the company spent $30 billion more than the previous year, driven primarily by the increased company’s AI ambitions. A portion of that spending is directed towards big ticket cloud deals, like the recent $10 billion deal with Google Cloud, but more resources are pouring into two large new data centers.
Louisiana’s new 2,250-acre site, called Hyperion, costs an estimated $10 billion and offers an estimated 5 gigawatts of computing power. In particular, the site includes arrangements with local nuclear power plants to handle increased energy loads. A small Ohio site called Prometheus is expected to be online with natural gas in 2026.
Such build-outs come with actual environmental costs. Elon Musk’s Xai has built its own hybrid data center and power plant in Southern Memphis, Tennessee. The factory has become one of the county’s biggest emitters thanks to a series of natural gas turbines that experts say are violating the Clean Air Act.
Stargate Moon Shot
Two days after his second inauguration, President Trump announced a joint venture between Softbank, Openai and Oracle. Named “Stargate” after the 1994 film, the project arrived with an incredible amount of hype, with Trump saying “it’s the biggest AI infrastructure project in history. Sam Altman thinks it will be the most important project of this era.”
On the broad strokes, SoftBank was a plan to fund it, with SoftBank handling the buildouts with input from Openai. It’s Trump who oversees it all, and he has promised to clean up any regulatory hurdles that could slow down the build. However, there were doubts from the start, including Altman’s business rival Elon Musk, who claimed that the project had no available funds.
As hype fads faded, the project lost momentum. In August, Bloomberg reported that their partners were unable to reach the consensus. Nevertheless, the project is working on building eight data centers in Abilene, Texas, with construction to be completed on the final building set by the end of 2026.
This article was first published on September 22nd.
