An AI product requires a lot of processing power, and while the IT sector competes to harness the potential of AI models, another race is taking place to construct the infrastructure that will support them. Nvidia CEO Jensen Huang said on a recent earnings call that between $3 and $4 trillion will be spent on AI infrastructure by the end of the decade, with a large portion of the funding coming from AI businesses. They are pushing the building capacity of the industry to its limit and putting tremendous burden on electricity networks in the process.
We have outlined all we know about the largest AI infrastructure initiatives below, including significant investments from Meta, Oracle, Microsoft, Google, and OpenAI. As the boom continues and the numbers continue to rise, we will keep it updated.
Microsoft’s $1 Billion Investment In OpenAI
This agreement may have been the catalyst for the current AI boom: Microsoft invested $1 billion in OpenAI, a popular non-profit organization best known for its ties to Elon Musk, in 2019. Importantly, the agreement established Microsoft as OpenAI’s sole cloud supplier. As model-training demands increased, a larger portion of Microsoft’s contribution began to take the form of Azure cloud credit rather than cash. Both parties benefited greatly from the agreement: OpenAI received more revenue for its largest single expense, while Microsoft was able to claim more Azure sales. Microsoft would increase its investment to around $14 billion in the ensuing years, a decision that will be highly beneficial when OpenAI becomes a for-profit business.
More recently, the two businesses’ collaboration has come to an end. OpenAI declared in January that it would no longer be solely relying on Microsoft’s cloud, instead granting the firm the first say in any future infrastructure requests while looking at other options in the event that Azure was unable to satisfy its requirements. In order to become even more independent from the AI behemoth, Microsoft recently started looking into other foundation models to power its AI products.
Because of the success of OpenAI’s partnership with Microsoft, it is now standard procedure for AI services to sign up with a specific cloud provider. Amazon has invested $8 billion in Anthropic, which is also modifying the company’s hardware at the kernel level to improve its suitability for AI training. Smaller AI firms like Loveable and Windsurf have also joined Google Cloud as “principal computing partners,” though no money was invested in those agreements. Additionally, OpenAI has returned to its roots after Nvidia invested $100 billion in September, enabling it to purchase additional GPUs from the firm.
The Rise Of Oracle
In an SEC filing on June 30, 2025, Oracle disclosed that it had inked a $30 billion cloud services agreement with an undisclosed partner, surpassing the company’s cloud revenues for the entire preceding fiscal year. on the end, OpenAI was identified as the partner, guaranteeing Oracle a place alongside Google on OpenAI’s list of post-Microsoft hosting partners. It should come as no surprise that the company’s stock soared.
A few months later, it occurred once more. Oracle announced a $300 billion, five-year contract for computing power on September 10th, with the transaction starting in 2027. Larry Ellison, the founder of Oracle, briefly became the richest man in the world as the company’s stock surged even higher. The magnitude of the transaction is astounding: OpenAI does not have $300 billion to invest, therefore the amount assumes tremendous expansion for both businesses and a great deal of faith. However, even before any money is exchanged, the agreement has solidified Oracle’s position as a major supplier of AI infrastructure and a formidable financial power.
Building Tomorrow’s Hyperscale Data Centers
For organizations such as Meta that already have substantial legacy infrastructure, the situation is more complex, yet no less costly. According to Mark Zuckerberg, Meta intends to invest $600 billion in US infrastructure by the end of 2028. Due in large part to the firm’s expanding AI goals, the company spent $30 billion more in the first half of 2025 than it did the year before. While some of that money is allocated to high-profile cloud contracts, such as the recent $10 billion agreement with Google Cloud, additional funds are being invested in two enormous new data centers. The construction of a new 2,250-acre Hyperion site in Louisiana is expected to cost $10 billion and produce an estimated 5 gigawatts of computing power. Interestingly, the complex has a deal with a nearby nuclear power plant to manage the extra energy demand. Prometheus, a smaller natural gas-powered plant in Ohio, is anticipated to go online in 2026.
There are actual environmental costs associated with that type of buildout. In South Memphis, Tennessee, Elon Musk’s xAI constructed its own hybrid data center and power plant. Thanks to a series of natural gas turbines that experts claim violate the Clean Air Act, the plant has rapidly grown to become one of the county’s biggest emitters of chemicals that cause smog.
The Stargate Moonshot
President Trump launched a $500 billion joint venture between SoftBank, OpenAI, and Oracle to construct AI infrastructure in the US just two days after his second inauguration. Trump referred to the project, which was named “Stargate” after the 1994 movie, as “the largest AI infrastructure project in history.” It arrived with an enormous amount of excitement. “I think this will be the most important endeavor of our era,” said Sam Altman, who appeared to concur.
The general idea was that Oracle would handle the buildout with assistance from OpenAI, while SoftBank would supply the cash. Trump was in charge of it all and pledged to remove any regulatory obstacles that might cause the construction to lag. However, there were skepticism from the start, particularly from Altman’s business adversary Elon Musk, who asserted that the initiative lacked the necessary funding.
The project has slowed significantly as the initial excitement has subsided. According to a Bloomberg story from August, the partners were unable to agree. The project has nevertheless proceeded with the construction of eight data centers in Abilene, Texas, with the completion of the last structure anticipated by the end of 2026.
Leave feedback about this