Meta announced a $27 billion commitment to Nebius over five years for AI data center infrastructure, the second hyperscaler contract Nebius has secured in three months. The deal follows a $17 billion commitment from Microsoft announced in December, bringing Nebius's disclosed forward infrastructure pipeline to $44 billion across two anchor tenants.
Nebius, the Amsterdam-listed infrastructure provider spun out of Yandex's cloud business in 2023, operates GPU-dense data centers optimized for large language model training. The Meta contract covers both capacity expansion and long-term compute leasing, with Nebius responsible for site development, power procurement, and cooling systems across multiple European and North American facilities. Meta will take priority access to Nebius-operated clusters running Nvidia H100 and H200 chips, with options for next-generation Blackwell configurations as they deploy in late 2025. The contract includes performance guarantees on uptime and power efficiency, tying roughly 30 percent of the total value to delivery milestones over the first 24 months.
This is Meta's largest disclosed infrastructure commitment outside its own capital expenditure program. The company spent $38 billion on data centers and network equipment in 2024, but the Nebius contract suggests Meta is accelerating third-party capacity to supplement internal builds. The deal also confirms Nebius's position as the only non-hyperscaler infrastructure provider with two contracts above $15 billion, a tier previously reserved for relationships between cloud giants and sovereign wealth-backed data center developers. For context, CoreWeave—the venture-backed GPU cloud that went public in early 2025—has disclosed roughly $9 billion in forward commitments from Microsoft and other enterprise customers, but no single contract above $5 billion.
The timing matters for two reasons. First, Meta's AI spending is running ahead of revenue acceleration. The company guided to $64 billion to $68 billion in total capex for 2025, up from $38 billion in 2024, with roughly 75 percent earmarked for AI infrastructure. Gross margins on advertising revenue have compressed 180 basis points year-over-year as compute costs rise faster than yield optimization can recover. Outsourcing a portion of that burden to Nebius shifts some capital intensity off Meta's balance sheet while maintaining access to frontier compute. Second, Nebius now has the contracted revenue base to command better pricing on bulk GPU orders and long-term power agreements, creating a cost structure advantage over smaller competitors. The company has secured 2.4 gigawatts of power capacity under fixed-rate contracts extending through 2029, insulating it from energy price volatility that has destabilized other European data center operators.
Allocators should watch three follow-on events. Nebius is expected to file for a U.S. secondary listing by mid-2025, which would open the stock to institutional buyers currently restricted by Amsterdam exchange liquidity. The company has also hinted at a third hyperscaler contract in discussions, likely with a search or e-commerce platform scaling multimodal AI capabilities. Finally, Meta's next earnings call in late April will clarify whether the Nebius deal alters the company's internal capex guidance or if this represents incremental spending above the $68 billion ceiling.
Nebius shares closed up 41 percent in Amsterdam trading on volume eighteen times the 30-day average. The stock has now doubled since the Microsoft deal was announced in December.