**Position: long shares. Not financial advice — do your own DD.**
**TL;DR:** AI data centers can’t get power fast enough, the grid can’t keep up, and Bloom sells the one fix you can deploy in months instead of years — modular on-site fuel cells. Backlog, customers, and the financial turn are all real. The catch: it’s priced like everyone already knows it. This is a hold-through-volatility thesis, not a back-up-the-truck-at-any-price one.
**The Setup**
Bloom makes solid-oxide fuel cells that turn natural gas, biogas, or hydrogen into on-site electricity, no combustion. Boring industrial story for 20 years — then AI happened. Hyperscalers need gigawatts *now*, and grid interconnection takes years. Bloom deploys behind-the-meter power in months. The whole thesis in one line: **they sell speed, and speed is the scarcest thing in the AI buildout.**
**Why it’s a hold, not a trade**
**1. The backlog is huge and named (\~$20–24B):**
**Oracle** — expanded \~2.8 GW master agreement (scrapped the gas turbines/diesel for a full AI campus)
**Brookfield** — $5B partnership to build “AI factories”
**Nebius** — up to $2.6B for data center power
Plus existing C&I names like Walmart and FedEx
**2. The financials actually turned (Q1 2026):**
Revenue **$751M, +130% YoY**; product revenue +208%
Non-GAAP EPS **$0.44** vs $0.03 a year ago
Adjusted EBITDA **$143M** vs $25M
Free cash flow positive, second straight year of positive operating cash flow
FY2026 guidance raised to **$3.4–3.8B revenue**, \~34% non-GAAP gross margin, **$1.85–2.25** adj. EPS
**3. They’re not diluting you.** CEO said outright they don’t need to issue stock to fund expansion. Rare in this sector.
**4. Policy tailwind.** FERC is letting large energy users connect faster — positive read-through for behind-the-meter providers. Also a floated future S&P 500 inclusion candidate.
**The risks (read this part twice)**
**Valuation is stretched.** Trades well above the average analyst target (\~$267), forward P/E north of 140x. Plenty of analysts sit at Hold/Market Perform — the business bull case is accepted, the *stock-at-this-price* bull case is not.
**Insiders are net sellers** (\~$83M trailing year, much of it scheduled after a \~15x run, but still).
**Execution is the whole game now.** Partner Crusoe paused a \~1.8 GW project — backlog ≠ revenue until it’s delivered. Permitting and supply chain can slip.
**Volatility is brutal.** Beta \~3.7, \~16% swings in a typical week. If a 30–40% drawdown makes you panic-sell, this isn’t your hold.
**“Clean-ish,” not clean.** Most servers run on natural gas — some policy/ESG exposure.
**Bottom line**
The structural story — AI power demand is a multi-year bottleneck and Bloom sells the fastest fix — is strong enough to hold through cycles. The risk isn’t the business breaking; it’s the **stock getting ahead of the business** and having to grow into the multiple. If you believe in the 5–10 year AI buildout and can ignore red days, BE is a defensible long-term hold. Buying at highs? Scale in instead of going all at once.
**Not financial advice. Size for the volatility.**