Multi-billion dollar infrastructure supports the development of AI

It takes a lot of computing power to be able to create AI – and as tech companies race to harness the power of AI models, there’s an equal race underway to build the infrastructure that can power them. At a recent callNvidia CEO Jensen Huang estimates that between $3 trillion and $4 trillion will be spent on AI infrastructure by the end of the decade — with most of that money coming from the AI ​​industry. Along the way, they’re putting major challenges on power grids and pushing business efficiency to the limits.

Below, we’ve covered everything we know about major AI projects, including major funding from Meta, Oracle, Microsoft, Google, and OpenAI. We will let you know as the volume continues and the numbers go up.

Microsoft’s 2019 investment in OpenAI

This is arguably the consensus that launched the entire AI community of today: In 2019Microsoft made a $1 billion investment in a buzzy nonprofit called OpenAI, which is best known for its partnership with Elon Musk. Most importantly, the deal made Microsoft the sole cloud provider for OpenAI – and as demand for the model grew, more Microsoft funding began to flow. in the form of Azure cloud credit not money.

It was good for both sides: Microsoft was able to command a lot of Azure sales, and OpenAI got a lot of money for its one big price. Over the next few years, Microsoft will make its investment around $14 billion – a move that should pay off big once OpenAI turns into a profitable company.

The partnership between the two companies has recently ended. Last year, OpenAI announced that it would don’t just use the Microsoft cloud anymoreinstead giving the company the right of first refusal on future needs but pursuing others if Azure can’t meet their needs. Microsoft has also begun looking at other types of foundations to develop its AI tools, establishing more freedom from the AI ​​giant.

OpenAI’s arrangement with Microsoft was so successful that it has become a common practice for AI services to sign with another cloud provider. Anthropic has received $8 billion in funding from Amazon, while making kernel updates on the company’s hardware to make it suitable for AI training. Google Cloud has also signed on small AI companies like Lovable and Windsurf as “computer partners,” although the agreements did not involve any revenue. And even OpenAI has returned to the well, receiving $ 100 billion in funding from Nvidia in Septembergiving them the power to buy more GPUs for the company.

Oracle update

On June 30, 2025, Oracle disclosed in the SEC that it had signed a $30 billion contract for cloud services with an unnamed partner; this is more than the company’s cloud revenue in all of the previous financial years. OpenAI was eventually revealed as a partner, protecting Oracle place next to Google as one of OpenAI’s string of co-hosts after Microsoft. Unsurprisingly, the company’s stock soared.

A few months later, it happened again. On September 10Oracle has unveiled a five-year, $300 billion contract for compute power, which will begin in 2027. Oracle’s stock he went up againbriefly making founder Larry Ellison the richest man in the world. The size of the agreement is surprising: OpenAI does not have $ 300 billion to spend, so the figure shows the great growth of both companies, and the limited faith.

But before a single dollar is spent, the deal has already established Oracle as one of the leading AI developers — and a financial force to be reckoned with.

Nvidia refunds

When AI labs struggle to create architecture, they buy GPUs from one company: Nvidia. The sale has left Nvidia cash-strapped — and it has been investing that money in the business in unconventional ways. In September 2025, Nvidia bought it 4% share in Intel competition for $ 5 billion – but the most surprising thing was to do with his customers. A week after the Intel deal was revealed, the company announced $100 billion investment in OpenAIpaid for by GPUs that can be used for OpenAI applications in progress. Nvidia also announced a similar partnership with Elon Musk’s xAI, and OpenAI was launched special GPU-for-stock configuration and AMD.

If it looks round, that’s because it is. Nvidia’s GPUs are valuable because they’re rare – and by selling them directly into the ever-expanding data system, Nvidia is making sure they stay that way. You can say the same about OpenAI’s privately held stock, which is valuable because it can’t be found through the public markets. At the moment, OpenAI and Nvidia are riding high and no one seems to be too worried – but if the threat starts to ring, this system will be under a lot of scrutiny.

Building tomorrow’s hyperscale data centers

For companies like Meta who already have it important heritage infrastructurethe matter is more difficult – although it is equally expensive. The CEO of Meta, Mark Zuckerberg, has said that the company plans to spend $ 600 billion on US infrastructure until the end of 2028.

In the first half of 2025, the company broke even Another $30 billion than last year, largely driven by the industry’s growing AI ambitions. Some of that damage goes to big cloud contractors, like recently $10 billion is associated with Google Cloudbut additional resources are being poured into two large data centers.

A new 2,250-acre property in Louisiana, called Hyperionwill cost $10 billion to build and provide about 5 gigawatts of computing power. In particular, the site includes arrangements with nuclear power plants to cope with the increase in energy consumption. A small facility in Ohio, called Prometheus, is expected to come online in 2026, powered by natural gas.

Such construction comes with real environmental costs. Elon Musk’s xAI built its own hybrid data center and power plant in South Memphis, Tennessee. The plant has become one of the biggest emitters in the region, thanks to the gas turbines it produces. experts say they are in violation of the Clean Air Act.

A picture of the Stargate moon

Just two days after his second inauguration last January, President Trump announced a partnership between SoftBank, OpenAI, and Oracle, meant to spend $500 billion to build AI infrastructure in the United States. Dubbed “Stargate” after the 1994 movie, the project was met with incredible fanfare, with Trump calling it “the biggest AI project in history.” OpenAI’s Sam Altman seems to agree, saying, “I think this is going to be a very important project in this era.”

Instead, the plan was for SoftBank to provide the funding, with Oracle working on the architecture with ideas from OpenAI. In charge of all was Trump, who promised to remove any obstacles that would delay the 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.

When the hype is over, the project has lost its momentum. In AugustBloomberg reported that the partners were unable to reach an agreement. However, the project has progressed with the construction of eight data centers in Abilene, Texasand the construction of the final building will be completed by the end of 2026.

An explosion of capex

“Capital expenses” are often very dry, depending on how much a company spends on physical assets. But as technology companies prepared to outline their capex plans for 2026, the rush for data center investments. it made the numbers more interesting – are very large.

Amazon was the capex leader, showing $200 billion in 2026 revenue (from $131 billion in 2025), while Google was second and estimated between $175 billion and $185 billion (from $91 billion in 2025). Meta estimates $ 115 billion to $ 135 billion (from $ 71 billion last year), although the figure is a little misleading because many data operations have been carried out. they left all their books. All told, hyperscalers plan to destroy approximately $700 billion in data center projects in 2026 alone.

It was enough money to confuse investors. The companies did not waver, however, explaining that AI infrastructure was critical to their companies. It has established a strange power. As you might expect, tech executives are more interested in AI than their Wall Street counterparts — and the more tech companies spend, the more nervous their bankers get. Add in more debt Many companies are taking those buildouts, and you’re starting to hear CFOs across the valley gritting their teeth.

This has not reduced the costs of AI use, but soon – if only, hyperscalers show that they can recoup those costs.

This article was originally published on September 22.