STOCKS | Nvidia Seeks Power and Memory Capacity to Ease Data Center Bottlenecks

Nvidia CEO Jensen Huang said the company is tracking every gigawatt-scale land, power, and shell resource around the world. According to Sina Finance, he made the remarks at Goldman Sachs' annual technology conference earlier this month. Nvidia views power supply as a major bottleneck because AI server chips cannot be brought online quickly without adequate electricity. The article said data center developers working with Google, Microsoft, and Oracle are worried that future power shortages could leave many chips idle. Recent land and power deals show Nvidia is using its balance sheet to steer limited electricity toward data centers deploying its own chips rather than projects using rival chips. The company and its large chip customers are also shifting away from betting everything on gigawatt-scale projects worth hundreds of billions of dollars. Smaller facilities and fragmented power resources are becoming more attractive to investors and data center builders. Two industry executives said uncertainty over community opposition and regulatory approvals makes it difficult to know when power will actually be connected, increasing the appeal of diversified smaller projects. Crusoe said this week it completed a new $3.9 billion funding round that included Nvidia, with the money to be used to expand containerized micro modular data centers with 1-megawatt capacity. Nvidia executive Dane Harris said the company is also looking for sites with existing idle power, with a focus on Nordic countries and Texas. Nvidia has also signed $20 billion in data center leases and plans to gradually sublease them to customers. The company introduced its MaxLPS power management system, which it said could allow customers to deploy 40% more GPUs under the same power allocation. The article also said Nvidia, Google, Emerald AI, Anthropic, and U.S. National Grid formed a coalition to promote grid-flexible data centers. Separately, the article said Nvidia has spent about $280 billion securing capacity from storage and compute chip makers, while memory shortages are expected to persist throughout 2026.