Oracle sidesteps grid bottlenecks with massive 2.8GW bloom energy fuel cell expansion

In a bid to accelerate its "supercluster" AI infrastructure without relying on the strained U.S. electrical grid, Oracle has nearly tripled its fuel cell supply agreement with Bloom Energy to a staggering 2.8 gigawatts (GW).

The expanded deal, announced on April 13, 2026, follows a successful pilot where Bloom delivered a fully operational power system to an Oracle site in just 55 days, a timeline traditional utility companies often measure in years.

Powering the "AI supercycle"

The partnership reflects the desperate race for power as hyperscalers build out high-density data centers for clients like OpenAI and xAI. Oracle has already contracted an initial 1.2 GW of capacity, with deployment across U.S. projects continuing through 2027.

On-site independence

Bloom’s solid oxide fuel cells generate electricity through an electrochemical reaction without combustion, allowing Oracle to bring data centers online "behind the meter" and independent of the public grid.

Analysts estimate the incremental supply agreement represents between $4 billion and $4.5 billion in potential revenue for Bloom Energy.

Deepening financial ties

The agreement is more than a simple vendor contract; it signals a deep strategic and financial integration between the two firms.

Equity warrants

On April 9, 2026, Bloom issued a warrant allowing Oracle to purchase approximately 3.5 million shares at a fixed price of $113.28. If fully exercised, these warrants would inject nearly $400 million into Bloom Energy, helping the manufacturer scale its production to meet Oracle's unprecedented demand.

Bloom Energy shares surged nearly 30% following the news, briefly touching $210 per share, as investors rewarded the company for its "grid-free" deployment speed.

A contrasting strategy

The announcement comes as Oracle navigates a complex period of aggressive spending and workforce restructuring. While the company is committing $50 billion in capital expenditure this year to build its AI "power foundation," it has recently undergone significant layoffs, with internal estimates suggesting as many as 30,000 roles were impacted in March 2026.