Meta, Nebius and Oracle in AI infra focus

Meta Platforms has agreed to purchase artificial intelligence computing capacity from Amsterdam-based Nebius Group under contracts worth up to $27 billion over five years, according to Reuters.

Under the agreement announced on March 16, Nebius will provide Meta with $12 billion of AI computing capacity across multiple locations by 2027. Meta also committed to buy up to an additional $15 billion of capacity planned by Nebius over the next five years if that capacity is not sold to other customers.

The combined commitments could bring the total value of the Meta–Nebius arrangement to as much as $27 billion, underscoring Meta’s drive to secure large-scale computing power for its AI initiatives.

Reuters reported that Nvidia (NVDA) plans to invest $2 billion to acquire an 8.3% stake in Nebius. Nebius uses Nvidia (NVDA) chips in its data centers, linking the graphics processor maker directly to the infrastructure Meta intends to use.

The Nvidia stake highlights the chipmaker’s participation in the AI data center ecosystem that supports large customers such as Meta, as Nebius expands its capacity footprint.

Alongside external capacity deals, Meta announced on March 11 that it is expanding its custom silicon program to power AI workloads. The company said it is developing and deploying four new generations of MTIA chips within the next two years.

Meta indicated that these new MTIA generations are aimed at supporting ranking and recommendations, as well as generative AI workloads. The chips expand its custom silicon portfolio, which Meta describes as central to its AI infrastructure strategy.

Meta’s broader technology products span personal computers, mobile devices, virtual reality, mixed reality headsets, and wearables, providing multiple platforms on which its AI capabilities can be deployed.

Meta has been integrating AI into WhatsApp, Instagram, and Facebook to increase user engagement and reinforce its advertising business. It also offers chatbots and foundational models that developers can use to build applications and AI tools.

The company is now focused on developing what it calls personal superintelligence, described as tailor-made for individuals to understand preferences, interact with their environments, and support users in achieving their goals.

To run these AI workloads, Meta has been investing heavily in data centers. In November 2025, the company said it planned to invest more than $600 billion in the United States over three years to build AI infrastructure.

Meta has also turned to external cloud providers to secure AI computing capacity. Oracle (ORCL) was reportedly in talks last year to sign a $20 billion AI cloud computing deal with Meta, reflecting Meta’s multi-partner approach to infrastructure sourcing.

Oracle (ORCL), traditionally known for database software, has seen its cloud computing business receive a major boost from AI demand. It operates in a market where demand for AI data center capacity significantly outpaces supply.

Oracle reported remaining performance obligations of $553 billion at the end of its fiscal third quarter of 2026, a 325% year-over-year increase, indicating a large backlog of contracts yet to be fulfilled.

In the same quarter, Oracle’s overall revenue rose 22% year over year to $17.2 billion, while cloud revenue grew 44% to $8.9 billion. Management has projected revenue growth to $67 billion in the current fiscal year and has indicated a target of $185 billion for fiscal 2029.

Meta is the eighth-largest publicly traded company globally with a $1.6 trillion market capitalization, driven by growth in its advertising business and AI-enabled tools for brands and advertisers.

Oracle’s expanding AI cloud business and substantial backlog have prompted comparisons with Meta’s growth profile. Some analysis suggests Oracle could experience faster revenue growth than Meta over the next several years, supported by demand for AI computing capacity.