It takes a lot of computing power to run an AI product -and when the tech industry runs to press the power of AI models, there is a parallel run to build the infrastructure that will drive them. On a recent revenue call, Nvidia CEO Jensen Huang estimated that between $ 3 trillion and $ 4 trillion will be spent on AI infrastructure at the end of the decade -with much of this money coming from AI companies. Along the way, they put a huge load on the power network and push the industry’s building capacity to its boundary.
Below we have put everything we know about the largest AI infrastructure projects, including greater expenses from Meta, Oracle, Microsoft, Google and Openai. We keep it up to date when the boom continues and the numbers climb even higher.
Microsoft’s invest of $ 1 billion in Openai
This is arguably the deal that started the entire modern AI Boom: In 2019, Microsoft made an investment of $ 1 billion in a buzzy non-profit called Openai, best known for its association with Elon Musk. Of crucial importance made the agreement Microsoft the exclusive cloud provider for Openai – and as the requirements for model training became more intense, several of Microsoft’s investment began to get in the form of Azure Cloud Credit rather than cash.
It was a lot for both sides: Microsoft was able to demand more Azure sales, and Openai got more money for his biggest single expense. In the years that followed, Microsoft would build its investments up to nearly $ 14 billion steps set to pay tremendously when Openai converts to a pre-profit company.
The partnership between the two companies has recently been recently. In January, Openai announced that it would no longer use Microsoft’s cloud exclusively, rather than giving the company a right to the first rejection of future infrastructure requirements, but to pursue others if Azure could not meet their needs. Recently, Microsoft began exploring other foundation models to operate its AI products and establish even more independence from the AI ​​giant.
Openai’s arrangement with Microsoft was so successful that it has become a common practice for AI services to log in with a particular cloud provider. Anthropic has received $ 8 billion in investing from Amazon while making changes at core level on the company’s hardware to make it better suited for AI training. Google Cloud has also signed smaller AI companies such as Lovable and Windsurf as “primary computer partners”, although these offers did not involve any investment. And even Openai is back to the well and received an investment of $ 100 billion from Nvidia in September, giving it the capacity to buy even more of the company’s GPUs.
The emergence of Oracle
On June 30, 2025, Oracle revealed in a SEC archiving that it had signed a cloud service of $ 30 billion with a named partner; This is more than the company’s cloud income for the entire previous financial year. Openai was eventually revealed as a partner and secured Oracle somewhere with Google as one of Openai’s string of post-microsoft hosting partners. It was not surprising that the company’s stock shot up.
TechCrunch -event
San Francisco
|
27-29. October 2025
A few months later it happened again. On September 10, Oracle revealed a five-year-old $ 300 billion agreement for computational power to begin in 2027. Oracle’s share rose even higher, which briefly made founder Larry Ellison the richest man in the world. The clean scale of the deal is great: Openai does not have $ 300 billion to spend, so the figure supposes tremendous growth for both companies and more than a small faith.
But before a single dollar is spent, the agreement has already cemented Oracle as one of the leading AI infrastructure providers – and a financial strength to be counted.
Nvidia’s Investment Spree
As AI Labs is encrypted to build infrastructure, they are mostly all who buy GPUs from a company: NVIDIA. This trade has led Nvidia to flush cash – and it has invested that cash back in the industry in increasingly unconventional ways. In September 2025, the company bought a 4% share in rival Intel for $ 5 billion – but even more surprising has been the agreements with its own customers. A week after the Intel agreement was revealed, the company announced an investment of $ 100 billion in Openai, paid for with GPUs that would be used in Openai’s ongoing data center projects. Nvidia has since announced a similar deal with Elon Musk’s Xai, and Openai launched a separate GPU-for storage event with AMD.
If it works circular, it should. Nvidia’s GPUs are valuable because they are so scarce and by acting directly to an ever-inflated data center scheme, Nvidia makes sure to stay that way. You can say the same about Openai’s privately owned warehouse, which is all the more valuable because it cannot be achieved through public markets. For the time being, Openai and Nvidia are driving high and no one seems too concerned – but if the momentum starts to mark, this kind of arrangement will gain much more control.
Building of tomorrow’s hyperscale -datacenters
For companies like Meta that already have significant older infrastructure, the story is more complicated – though it is just as expensive. Mark Zuckerberg has said Meta is planning to spend $ 600 billion on US infrastructure through the end of 2028.
In just the first half of 2025, the company spent $ 30 billion more than the previous year, largely driven by the company’s growing AI ambitions. Some of these expenses go against big ticket cloud contracts, such as a recent $ 10 billion deal with Google Cloud, but even more resources are poured into two massive new data centers.
A new 2,250 hectare place in Louisiana, called Hyperion, costs an estimated $ 10 billion to build out and give an estimated 5 gigawatt calculation power. In particular, the site includes an event with a local nuclear power plant to handle the increased energy load. A smaller place in Ohio, called Prometheus, is expected to come online in 2026, powered by natural gas.
This kind of structure comes with real environmental costs. Elon Musk’s Xai built his own hybrid data center and power generation system in South Memphis, Tennessee. The plant has quickly become one of the county’s largest broadcasts of smog-producing chemicals thanks to a number of natural gast turbines that experts say violate the law of pure air.
Stargate Moonshot
Only two days after his second inauguration, President Trump announced a joint venture between Softbank, Openai and Oracle intended to spend $ 500 billion on building AI infrastructure in the United States. The project, named “Stargate” after the 1994 movie, arrived with incredible amounts of hype, with Trump calling it “the biggest AI infrastructure project in history.” Sam Altman seemed to agree and said, “I think this will be the most important project in this era.”
In broad strokes, the plan was for Softbank to deliver the financing, with Oracle handling the construction with input from Openai. To oversee it all was Trump, who promised to clear away all regulatory obstacles that had to slow down the building. But there was doubt from the beginning, including from Elon Musk, Altman’s business rival, who claimed that the project did not have the available funds.
Since hype is dead, the project has lost some speed. In August, Bloomberg reported that the partners did not reach consensus. Nevertheless, the project has progressed with the construction of eight data centers in the Abilene, Texas, with construction on the final building to be completed by the end of 2026.
This article was first published on September 22.