Bloom Energy is supplying fuel-cell power to data centers as artificial intelligence workloads create a massive surge in electricity demand.

This shift is critical because modern AI models require energy at a scale that traditional power grids struggle to support. As companies deploy next-generation chips, the need for reliable, onsite power generation has become a primary bottleneck for AI scaling.

KR Sridhar, the founder, chairman, and CEO of Bloom Energy, discussed the trend during an interview on CNBC TV18’s “Voices From The Valley.” Sridhar said that a single GPU-based rack now consumes power equivalent to 300 homes [1]. This level of consumption is driving data center operators to seek clean and reliable power through Bloom Energy's fuel-cell systems.

The financial impact of this demand has been significant. Bloom Energy’s share price rose 1,511% over the past year [2], and the company's market capitalization has exceeded $80 billion [2]. These gains reflect a broader market bet on the infrastructure required to sustain the AI boom.

Operational growth has mirrored the stock performance. Revenue grew 130% in the first quarter, a spike attributed to AI-driven fuel-cell orders [2]. The company also secured a 2.8 GW fuel-cell order from Oracle specifically for AI data centers [2].

Sridhar said that the energy requirements for these facilities are no longer incremental. The transition to high-density GPU racks means that power is now the limiting factor for the physical expansion of AI capabilities. By providing fuel cells, Bloom Energy allows operators to bypass grid limitations, enabling faster deployment of computing clusters.

“A single GPU-based rack consumes power equivalent to 300 homes.”

The rapid growth of Bloom Energy highlights a critical transition in the AI race, where the primary constraint has shifted from software and chip availability to physical power infrastructure. The reliance on fuel-cell technology suggests that the existing electrical grid cannot keep pace with the energy density required by modern GPUs, potentially leading to a decentralized power model for the tech industry.