Anthropic and OpenAI Seek Smaller Data Center Deals to Accelerate AI Deployment
- Anthropic and OpenAI are actively seeking smaller data center deals ranging from 20 to 30 megawatts.
- The race for compute capacity has driven both AI labs to pursue a new tier of infrastructure projects.
- Two of those sources indicated that OpenAI had previously explored similar small-capacity deployment opportunities in the Nordics.
Anthropic and OpenAI are actively seeking smaller data center deals ranging from 20 to 30 megawatts. The move marks a sharp shift away from massive gigawatt-scale commitments as both companies race to deploy artificial intelligence infrastructure more rapidly, according to reporting by CNBC.
Anthropic and OpenAI Pivot to Smaller Compute Deals
The race for compute capacity has driven both AI labs to pursue a new tier of infrastructure projects. Over the past year, Anthropic and OpenAI signed substantial agreements for large-scale facilities requiring hundreds of megawatts or even gigawatts of power. But securing smaller allocations allows companies to deploy workloads significantly faster amid the ongoing AI boom.
Global Search for Compact Infrastructure
Two of those sources indicated that OpenAI had previously explored similar small-capacity deployment opportunities in the Nordics. Additionally, a source noted that both labs have consulted on capacity deployments of this scale within the United States.
Logistical Bottlenecks and Grid Congestion
This strategy directly targets the logistical bottlenecks of massive infrastructure buildouts. While massive campuses take years to construct and power, smaller compute allocations plug into existing facilities with far less friction.
Speed to Usable Capacity
Smaller capacity deals offer a critical operational advantage known as speed to usable capacity, an analyst told CNBC. Securing a few megawatts at an existing powered site proves far more practical than waiting years for a much larger block of power to become available in a single location.
For complex workloads that can operate across separate, distributed sites, a collection of smaller deployments can quickly add up to substantial operational capacity. This fragmented approach helps labs bypass severe grid congestion and long wait times for new transmission lines.
