It takes a lot of computing energy to run an AI product – and as a racing to tap the power of the AI models with the technology industry, a parallel nation is underway to create infrastructure that will give their strength. At the call of recent earnings, Nvidia CEO Jensen Huang estimated that $ 3 trillion to $ 4 trillion will be spent on AI infrastructure in the late decade – most money will come from AI companies. On the way, they are putting a lot of pressure on the power grids and pushing the industry to its limit.
Below, we have kept everything we know about the largest AI infrastructure projects, including the big expenses of Meta, Oracle, Microsoft, Google and Openai. We will keep it updated as the boom continues and the numbers rise higher.
Investing $ 1 billion in Microsoft’s OpenAI
This is the logic of the rationalist that kicked the entire contemporary AI Boom: In 2019A rumor called Microsoft OpenAI has invested $ 1 billion as non -profit, known for relationships with most Elon masks. Seriously, this agreement has turned Microsoft an exclusive cloud supplier of Openai – and Model training claims become more intense when Microsoft’s further investment begins to come In the form of Credit Credit in Azor Rather in cash.
It was a great thing for both sides: Microsoft was able to claim sales in more azur and Opina had more money for its largest single expenditure. In the following years, Microsoft will make his investment up to $ 14 billion-a step that is ready to pay a lot when the Open is converted into a profitable company.
The partnership between the two companies has been more uninterrupted recently. In January, the Open was announced that it would no longer use Microsoft’s cloud exclusively, instead gave the company the right to reject the future infrastructure demand, but Azur could not fulfill their needs if others could not meet others. Recently, Microsoft has begun exploring other foundation models to strengthen its AI products, establishing more independence from the AI giant.
The opening system with Microsoft was so successful that it has become a general practice signing with a specific cloud provider for AI services. Anthropic has received $ 8 billion investment from Amazon, while Kernel The company’s hardware to make it more suitable for AI training. Google has signed as “primary computing partners” in small AI companies such as Google Cloud and Windsorf, though these agreements were not involved in any investment. Even the OPEND has received $ 100 billion investment from Nvidia, has returned to the well In SeptemberIt provides the company’s GPU more ability to buy more.
Oracle
On June 5, 2021, Oracle filed a SEC and revealed that it signed a $ 30 billion cloud service deal with an anonymous partner; This is more than the company’s cloud revenue for everyone in the previous fiscal year. Opena was finally published as partner, securing Oracle Google as well as a spot Openai Post-Microsoft is one of the strings of hosting partners. Surprisingly, the company’s stock shooting began.
TechCrunch event
San Francisco
|
October 27-29, 2025
A few months later, it happened again. On September 10, Oracle published a five -year, $ 300 billion contract to start in 2027. Oracle’s stock rose to a higher higher, briefly made by Founder Larry Ellison as the world’s richest man. The perfect scale of the deal is stunning: There is no $ 300 billion to spend on opening, so the image seems to be a lot of growth for both companies and more than trust.
However, before spending a single dollar, the contract has already cement Oracle as the top AI infrastructure supplier – and will be considered a financial power.
Nvidia investment spree
As the AI labs scramble to create infrastructure, they mostly buy GPU from Nvidia. This trade flushed NVEDIA with cash – and it invests in the industry in a growing way. In September 2025, the company bought 4% share of the rival Intel for $ 5 billion – but even more surprising is an agreement with its own customers. A week after Intel is published in the contract, the company has announced Invested a 100 billion dollars in OpenAIThe GPU has been paid with which the Openai will be used on the ongoing data center projects. Nvidia has since announced a similar deal with Elon Mask’s Jai and launched OpenAI An individual GPU-Stock format Including AMD.
If it seems round then it should be. Nvidia’s GPUs are valuable because they are so scarce-and they are directly trading on the growing data center scheme to ensure that they are there that way. You can say the same thing about Openai’s personally held stock, which is more valuable because it is not available through public markets. Suddenly, Openai and Nvidia are climbing higher and no one seems to be too worried – but if the speed starts to flagging, this kind of system will investigate many more.
Tomorrow’s HyperCel Data Centers Building
Already for companies like Meta Important legacy infrastructureThe story is more complicated – though equally expensive. Mark Zuckerberg says Meta has planned to spend $ 600 billion in US infrastructure At the end of 2028The
In the first half of 2021, the company spent $ 30 billion more than the previous year, which was originally driven by the agency’s growing AI ambition. Some of those expenditures go to the cloud agreement like some of the most recent times $ 10 Billion Deal with Google CloudHowever, more resources are being provided in two huge new data centers.
A new 2,250 acre site in Louisiana, Hyperion dubApproximately 5 gigwatts will spend approximately $ 10 billion to create and supply power. Significantly, the site includes an arrangement with the local nuclear power plant to manage the enhanced energy burden. A small site of Ohio, known as Prometheus, is expected to come online by natural gas in 2026.
This type of buildout brings real environmental expenses. Elon Mask created its own hybrid data center and power generation plant in South Memphis, Tennessee. Thanks to a string of natural gas turbines the plant has become one of the largest emitters of chemicals producing fast smoke Experts say Clean Air Act is violatedThe
Stargate Munshot
Just two days after his second inauguration, President Trump announced a joint venture between Softbank, OpenAI and Oracle, which means $ 500 billion in the US infrastructure building in the United States. Named “Stargate” after the 5th film, the project brought an incredible amount of hype, Trump called it “the largest AI infrastructure in history”. Sam Ultman seemed to agree, “I think this would be the most important project of this era.”
On broad strokes, the Oracle Buildout with input from OpenAI was planned to provide softbank funds as well as operating the Oracle Buildout. They were the supervision of everyone who was supervised, who promised to remove any regulatory obstacles that could slow down this build. However, from the very beginning there was doubt that Altman’s business rival Elon Mask, who claimed that the project had no funds available.
The project has lost some speed as the hype dies. In AugustBloomberg says the partners are failing to reach the Sens Karo. Nevertheless, the project has gone ahead with construction Eight data centers in Abelin in TexasWill end at the end of 2026 with the construction of the final building.
This article was first published on September 22.