The trillion-dollar race to build global AI infrastructure

Illustration photo.

Nvidia CEO Jensen Huang estimates that global AI infrastructure spending could reach $3 trillion to $4 trillion by the end of the decade, with much of it coming from tech companies themselves. The race is putting enormous pressure on the grid and pushing the industry’s ability to build to its limits.

* Microsoft and the inaugural deal with OpenAI

In 2019, Microsoft started the AI wave by investing $1 billion in OpenAI, becoming the startup’s exclusive cloud provider. As demand for AI model training grew, much of the investment was converted into Azure service credits. To date, Microsoft’s total investment in OpenAI is nearly $14 billion – an investment that is expected to pay off handsomely as OpenAI transitions to a for-profit model.

That relationship changed earlier this year when OpenAI stopped using Azure exclusively, only prioritizing Microsoft if it could meet its needs. At the same time, Microsoft also developed its own platform models to reduce its dependence. This type of partnership model quickly spread: Anthropic received $8 billion from Amazon, Google Cloud signed deals with small AI companies, and OpenAI itself continued to raise another $100 billion from Nvidia to expand its GPU infrastructure.

* Oracle rises to become the new "big guy"

Oracle has emerged with a series of large-scale deals. In June 2025, it signed a $30 billion cloud services contract with OpenAI – more than Oracle’s entire cloud revenue in the previous fiscal year. Three months later, the company announced a $300 billion five-year deal starting in 2027, sending its stock price soaring and briefly making founder Larry Ellison the richest person in the world .

Even without spending, these contracts solidify Oracle's position as one of the leading global AI infrastructure providers.

* Meta and the race to build data centers

For established giants like Meta, the infrastructure problem will be more complicated. Meta CEO Mark Zuckerberg said the company plans to spend $600 billion on infrastructure in the US by the end of 2028. In the first half of 2025 alone, Meta increased spending by $30 billion compared to the same period, mainly to serve the AI strategy.

In addition to major cloud deals like the $10 billion deal with Google Cloud, Meta is building two massive data centers. The Hyperion project in Louisiana is a 2,250-acre project worth about $10 billion and will provide 5 gigawatts of computing power powered by a local nuclear power plant. The smaller Prometheus center in Ohio is scheduled to open in 2026 and will run on natural gas.

However, the infrastructure boom also brings environmental risks. Billionaire Elon Musk's chatbot xAI built a power plant and data center in Tennessee, quickly becoming one of the region's largest sources of emissions.

* Stargate – the biggest AI project in history?

Two days after taking office for his second term, President Donald Trump announced the $500 billion “Stargate” project – a joint venture between SoftBank, OpenAI and Oracle to build AI infrastructure in the US. SoftBank will provide funding, Oracle will build with OpenAI as a consultant, and the Trump administration will pledge to remove regulatory barriers.

However, the project soon ran into financial doubts and difficulties in reaching consensus. Nevertheless, the project is still underway, with eight data centers in Abilene, Texas, expected to be completed by the end of 2026.

The AI infrastructure race is becoming the new strategic battlefield of the technology industry. With trillions of dollars committed to investment and a series of projects of unprecedented scale, large technology corporations are competing not only on AI models but also on computing platforms – the decisive factor in technological strength for many years to come.

Source: https://vtv.vn/cuoc-dua-nghin-ty-usd-xay-dung-ha-tang-ai-toan-cau-100250929155425364.htm