CoreWeave (CRWV.US), the dominant cloud computing provider powering AI giants including OpenAI, Microsoft, and Google, unveiled ambitious plans to construct a massive AI data center in Lancaster, Pennsylvania, with investments reaching $6 billion. This strategic expansion comes amid feverish infrastructure development across the tech sector: Nvidia (NVDA.US) has resumed H20 AI GPU sales in China; Meta (META.US) is funneling hundreds of billions into hyperscale AI facilities targeting artificial general intelligence (AGI), with its first site launching next year; and Amazon AWS sealed an expanded deal with SK Group to build South Korea's largest 100-megawatt data center.
The unrelenting global surge in AI computing demand, turbocharged by colossal investments from tech titans, signals sustained momentum for computing power leaders. Wall Street analysts now project Nvidia's market capitalization could sail toward $5 trillion—with some bullish forecasts eyeing $6 trillion—as the AI revolution continues supercharging valuations.
CEO Michael Intrator disclosed CoreWeave will initially deploy 100 megawatts of computing capacity at the Pennsylvania site, scalable to 300 megawatts. Announced alongside former President Donald Trump at a Pittsburgh CEO roundtable, the project will generate 600 high-wage construction jobs and 175 permanent operations positions. The facility will house next-generation Nvidia AI GPUs and NVL server racks to train what Intrator termed "the planet's largest AI models" requiring unprecedented computational resources.
As Nvidia's privileged early-adopter partner—often called "Nvidia's favored son" for its preferential GPU access—CoreWeave became the inaugural provider to deploy Blackwell Ultra GB300 NVL72 servers. This advantage fueled its recent $9 billion acquisition of data center operator Core Scientific, expanding its U.S. high-performance network to meet insatiable demand from cloud providers and AI developers.
"Demand patterns show relentless escalation," Intrator observed. "Three years ago, we discussed 10-megawatt deployments. Then 100 megawatts. Now conversations pivot decisively toward gigawatt-scale infrastructure." A single gigawatt facility could power approximately 750,000 U.S. homes. With the first 100-megawatt phase slated for 2026 completion, Intrator anticipates "intense market competition" for the site's capacity.
CoreWeave will lease the Lancaster campus from developers Chirisa Technology Parks and Machine Investment Group. Investor enthusiasm remains white-hot—shares have tripled since March's market debut, closing at $132.37 before surging 9% premarket Tuesday on the expansion news. The project forms part of Trump's $70 billion energy and AI investment package, complementing SoftBank, OpenAI, and Oracle's $100 billion "Stargate" hyperscale initiative.
CoreWeave's meteoric rise stems from prescient bets on Nvidia's ecosystem. As the first cloud provider to deploy H200 Tensor Core GPUs in 2023, it leveraged exclusive partnerships across CUDA software and Blackwell hardware to force even Microsoft into leasing its resources. Its core offering—bulk access to premium Nvidia GPU clusters—enables elastic scaling for massive AI training workloads like large language models, alongside traditional HPC applications spanning scientific simulation and financial analytics.
Comments