Jakarta, INTI - Running an AI product requires enormous computing capacity, and as the technology sector rushes to harness advanced AI models, a parallel competition is unfolding to construct the infrastructure needed to sustain them. During a recent earnings call, Nvidia CEO Jensen Huang projected that total spending on AI infrastructure could reach between $3 trillion and $4 trillion by the end of the decade, with AI companies supplying much of that capital. This surge in investment is placing significant pressure on power grids and stretching construction capabilities to their limits.

Below is an overview of the largest AI infrastructure initiatives, featuring substantial commitments from Meta, Oracle, Microsoft, Google, and OpenAI. The list will continue to evolve as investments expand and the boom intensifies.

Microsoft’s 2019 Investment in OpenAI 

Many consider Microsoft’s 2019 $1 billion investment in OpenAI, then a prominent nonprofit closely associated with Elon Musk, as the spark that ignited today’s AI boom. A crucial element of the agreement designated Microsoft as OpenAI’s exclusive cloud provider. As model training demands escalated, a larger share of Microsoft’s backing came in the form of Azure cloud credits rather than direct cash.

The arrangement proved mutually beneficial: Microsoft boosted Azure revenue, while OpenAI secured funding for its largest operational cost. Over time, Microsoft’s total investment climbed to nearly $14 billion, positioning it to benefit substantially once OpenAI transitions into a for-profit entity.

Recently, however, the partnership has evolved. OpenAI announced last year that it would no longer rely solely on Microsoft’s cloud, granting Microsoft a right of first refusal on future infrastructure needs while retaining the option to seek alternatives if Azure fell short. Microsoft has also begun testing additional foundation models to power its AI services, increasing its independence.

The success of this partnership set a precedent for similar arrangements. Anthropic secured $8 billion from Amazon while customizing hardware at the kernel level to optimize AI training. Google Cloud has also partnered with emerging AI firms such as Lovable and Windsurf as primary computing providers, though without direct equity investments. Meanwhile, OpenAI returned to Nvidia in September for a $100 billion investment, expanding its GPU capacity even further.

The Rise of Oracle 

On June 30, 2025, Oracle disclosed in an SEC filing that it had entered into a $30 billion cloud services agreement with an undisclosed partner, exceeding its total cloud revenue from the prior fiscal year. The partner was later identified as OpenAI, securing Oracle’s position alongside Google as one of OpenAI’s hosting providers beyond Microsoft. The announcement sent Oracle’s stock sharply higher.

Just months later, on September 10, Oracle unveiled a five-year, $300 billion compute power deal beginning in 2027. The news propelled its stock further, briefly making founder Larry Ellison the world’s richest individual. The scale of the agreement is extraordinary: OpenAI does not currently possess $300 billion in available capital, implying significant anticipated growth and substantial confidence from both sides.

Even before funds are deployed, the agreement has solidified Oracle’s status as a major AI infrastructure supplier and a formidable financial player.

Nvidia’s Investment Spree 

As AI laboratories rush to expand infrastructure, most are purchasing GPUs from Nvidia, a trend that has generated enormous cash flow for the chipmaker. Nvidia has reinvested those proceeds into the ecosystem in increasingly unconventional ways. In September 2025, the company acquired a 4% stake in rival Intel for $5 billion.

More surprising, however, were its arrangements with customers. A week after the Intel stake became public, Nvidia announced a $100 billion investment in OpenAI, paid in GPUs designated for ongoing data center projects. The company later revealed a similar agreement with Elon Musk’s xAI, while OpenAI entered a separate GPU-for-equity deal with AMD.

If that arrangement sounds circular, it’s because it is. Nvidia’s GPUs command high value largely due to their scarcity, and by channeling them directly into rapidly expanding data center projects, Nvidia helps preserve that scarcity. A similar logic applies to OpenAI’s privately held shares, whose exclusivity enhances their worth since they are not accessible through public markets. For now, both companies continue to benefit from strong momentum, with limited concern from observers. However, if growth slows, these reciprocal arrangements are likely to face closer examination.

Building Tomorrow’s Hyperscale Data Centers 

For companies such as Meta, which already operate extensive legacy infrastructure, expansion presents both complexity and high costs. CEO Mark Zuckerberg has stated that Meta intends to invest $600 billion in U.S. infrastructure through 2028.

During the first half of 2025 alone, Meta’s spending exceeded the previous year’s level by $30 billion, largely fueled by its expanding AI initiatives. While some of this capital is allocated to major cloud agreements, including a recent $10 billion deal with Google Cloud, a significant portion is directed toward constructing two massive data centers.

One of these projects is a 2,250-acre campus in Louisiana named Hyperion, projected to cost around $10 billion and deliver approximately 5 gigawatts of computing capacity. The facility includes a partnership with a nearby nuclear power plant to manage its energy requirements. A smaller Ohio-based site, Prometheus, is scheduled to launch in 2026 and will rely on natural gas for power.

Such large-scale developments carry environmental implications. Elon Musk’s xAI established a hybrid data center and power generation facility in South Memphis, Tennessee. The operation has rapidly become one of the region’s largest sources of smog-forming emissions, attributed to multiple natural gas turbines that experts argue breach the Clean Air Act.

The Stargate Moonshot 

Shortly after his second inauguration last January, Donald Trump announced a joint venture involving SoftBank, OpenAI, and Oracle to allocate $500 billion toward AI infrastructure development in the United States. Branded “Stargate,” in reference to the 1994 film, the initiative was introduced with considerable fanfare, with Trump describing it as “the largest AI infrastructure project in history.” OpenAI’s Sam Altman echoed that sentiment, stating, ”I think this will be the most important project of this era.”

Broadly, the arrangement envisioned SoftBank supplying capital, Oracle managing construction with OpenAI’s input, and Trump facilitating regulatory approvals. Skepticism emerged early, including criticism from Elon Musk, who questioned whether sufficient funding was truly secured.

As initial enthusiasm subsided, the project encountered slower progress. In August, Bloomberg reported difficulties among the partners in reaching agreement. Nevertheless, development has continued, with eight data centers under construction in Abilene, Texas, and the final structure expected to be completed by the end of 2026.

The Capex Crunch 

Capital expenditures, typically a routine financial metric, have taken on new significance as technology giants unveiled ambitious spending plans for 2026. The surge in data center investment has dramatically expanded projected figures.

Amazon leads the pack with an anticipated $200 billion in 2026 capex, up from $131 billion in 2025. Google follows with projections ranging between $175 billion and $185 billion, compared with $91 billion the prior year. Meta forecasts spending between $115 billion and $135 billion, rising from $71 billion, though some projects remain off its balance sheet. Altogether, hyperscalers are preparing to invest nearly $700 billion in data centers in 2026 alone.

These vast sums have unsettled certain investors. Still, corporate leaders remain steadfast, arguing that AI infrastructure is essential to long-term competitiveness. This divergence has created tension: while executives express strong confidence in AI’s prospects, financial markets show greater caution. Combined with rising debt levels used to finance expansion, concerns are growing among corporate finance teams.

For now, AI-related spending continues unabated. Whether it remains sustainable will depend on hyperscalers’ ability to translate these extraordinary investments into tangible returns.

Conclusion 

The AI boom is no longer just about smarter models, it is increasingly defined by trillion-dollar infrastructure bets. From GPU-for-equity swaps to hyperscale data center buildouts and massive capex projections, tech giants are committing unprecedented capital to secure computing power. While the momentum remains strong, the sustainability of this expansion will ultimately depend on whether AI-driven revenues can justify the scale of investment. If returns fall short, today’s bold infrastructure race could quickly face tougher financial and regulatory scrutiny.

Read more: Indonesia–US Data Exchange: What Are the Impacts and Risks for the National Data Center Sector?