Bloom Energy (BE) was up +23.98%.

Not because fuel cells suddenly became fun. Because investors are finally treating power as the limiter for AI data centers.

And once you see that, today’s move starts to make more sense than it probably did on the first read.

Oracle says it plans to procure up to 2.8 gigawatts (GW) of Bloom’s fuel-cell systems for AI and cloud data centers.

“Up to” is doing real work there. It’s not a single guaranteed purchase order for 2.8 GW. It’s more like a ceiling for what could be rolled out over time.

But still. 2.8 GW is not small. You don’t put a number like that on paper unless you expect to need a lot of electricity.

What matters is the problem Oracle is trying to solve. Data centers are being built quickly. The grid often isn’t. Permits take time. Upgrades take time. Getting a big new connection can take time. Meanwhile, the AI arms race is very comfortable spending money and very uncomfortable waiting.

So companies look for “good enough, on time” power. Bloom’s systems generate electricity on-site. That’s the whole pitch.

Here’s where I’m at: the market isn’t just cheering an Oracle logo. It’s betting that Oracle is showing everyone the next line item in the AI budget. Not just chips, not just buildings – power.

Slight tangent, but it matters: when a big-name buyer makes a public move, investors immediately start playing the “who’s next” game. That game alone can re-rate a stock, at least in the short term.

Bloom isn’t starting from zero either. In its Q4 2025 report (released February 5, 2026), the company put up a much larger quarter versus the prior year:

• Q4 2025 revenue: $777.7M (vs. $572.4M in Q4 2024)
• FY 2025 revenue: $2.024B
• Non-GAAP gross margin (FY 2025): 30.3%
• Non-GAAP EPS (Q4 2025): $0.45 diluted
• Adjusted EBITDA (FY 2025): $271.6M

That doesn’t “prove” the Oracle ramp. But it helps explain why the market was willing to pay up today. Investors prefer big growth stories when the recent numbers aren’t falling apart.

One more thing people are watching: Jefferies upgraded Bloom on April 14, pointing to better visibility tied to Oracle. Upgrades don’t make a business. But they can pull more attention – and attention matters when a stock starts moving.

Now for the boring part that decides whether this move sticks. How much of “up to 2.8 GW” turns into real orders. How fast it ships. And whether Bloom can scale without giving back margins to win the work.

If we get clearer timelines and the first chunks land cleanly, investors will keep widening the circle around this story. If it drags, or the economics disappoint, the stock can give back a lot in a hurry.

Either way, it’s hard to unsee the takeaway: AI is forcing a new question – where does the power come from?