Snapshot
- Ticker: BE (NYSE)
- Bucket: Power & Energy Infrastructure
- Q4 2025 fund position: **875.5 M common (10.08 M sh) + $35.5 M call notional (0.41 M sh) — 16.51 % of 13F, #1 holding
- HQ: San Jose, CA
- Fiscal year: calendar (Dec 31)
Business Overview
Bloom Energy makes solid-oxide fuel cells (SOFC) — the Bloom Energy Server — that generate electricity on-site from natural gas, hydrogen, or biogas at ~52 % electrical efficiency, with combined heat & power efficiency above 90 %. A single “Energy Server” module produces ~250 kW; sites are commonly built in MW-scale arrays.
Key product properties:
- Operates behind the meter — requires no utility interconnect
- Fast deployment — 6–12 months versus 5+ years for utility power
- Tier-IV-compatible — 99.999 %+ uptime
- Near-zero air pollutant emissions even running on natural gas
- Hydrogen-ready as the H2 economy develops
The 2024 AEP (American Electric Power) channel partnership (up to 1 GW master agreement structure) and the 2025 expansion of the AI data-center pipeline reframed Bloom from a niche distributed-power vendor into the marquee on-site generator for the AI era.
Why On-Site SOFC for AI Data Centers?
Three properties combine to make Bloom uniquely positioned:
- Speed. A 100 MW Bloom installation can be operational in well under a year. A new 100 MW utility interconnect is typically a 4–7 year project. AI data-center developers cannot wait.
- Reliability. Tier-IV-compatible 99.999 %+ uptime, with redundancy that grid power often cannot match.
- Carbon profile. Even running on natural gas, Bloom is materially cleaner than open-cycle gas turbines and is hydrogen-ready as the H2 economy develops.
The combination makes Bloom the de facto solution when an AI hyperscaler needs reliable hundreds-of-MW of capacity in 12 months and the grid can’t deliver. The 2025 backlog reportedly grew to multiple GW.
Financial Trajectory
| Metric (USD M) | FY22 | FY23 | FY24 | FY25 (reported) |
|---|---|---|---|---|
| Revenue | ~1,200 | 1,335 | 1,475 | >2,000 |
| YoY growth | — | +11 % | +10 % | +35 %+ |
| Gross margin (non-GAAP, directional) | sub-20 % | low-20s | mid-20s | improving |
| Operating income | loss | loss | near breakeven (adj.) | turning positive |
| Net income | loss | loss | narrowing | improving |
The directional story (matching the Q4 2025 fund weight): margins climbing as the AI-DC mix grows; the company is approaching sustained operating profitability; revenue inflected sharply in FY25 with the AEP-channel deals beginning to convert backlog to acceptance.
Bloom’s revenue is heavily Q4-weighted because acceptance milestones (the GAAP revenue trigger) cluster at year-end project completions.
Segment / Product Lines
Bloom reports primarily two product lines plus smaller installation/electricity revenue:
| Segment | % of revenue (historical) | Notes |
|---|---|---|
| Product (Energy Server sales) | ~70–75 % | The headline lever |
| Service (multi-year operations) | ~20–25 % | Annuity-like, growing with installed base |
| Installation / Electricity | balance | Ancillary |
The Service line has structurally higher margins and grows mechanically with the installed base — a hidden compounding lever that the AI capex story tends to overlook.
Operational KPIs
- MW accepted/shipped (the headline volume metric):
- FY24: ~300–400 MW range
- FY25: meaningful step-up tied to AEP deal conversion
- Backlog: multi-GW by 2025, with the AEP master agreement structured around up to 1 GW of master-agreement capacity
- Hyperscaler customers:
- Equinix (long-standing, multiple expansions)
- AEP (channel partner 2024 — end customers under NDA but understood to include leading AI hyperscalers)
- Multiple direct AI-data-center accounts (under NDA)
Balance Sheet
Bloom historically carried meaningful convertible debt (~500–700 M cash. Net-debt/EBITDA is not a clean metric while EBITDA is near zero, but the trajectory is improving as EBITDA turns positive in FY24-FY25.
No buyback or dividend — Bloom remains in growth-investment mode.
Why It Fits the Thesis
Bloom is the single highest-conviction position in the entire fund — it is #1 in the 13F at 15.9 % common + 0.6 % call overlay = 16.5 % combined. The bucket logic is clear: if AI scaling continues and power is the binding constraint, then on-site generation that can be deployed in <1 year is the only short-term solution, and Bloom is the only US-public pure-play on that solution.
The position grew from “new, large size” in Q3 2025 to #1 long with calls layered on in Q4 2025 — one of the most decisive single-quarter conviction moves in fund history.
Position History in the Fund
| Quarter | Position |
|---|---|
| Q3 2025 | New, large size (entered as top-3 immediately) |
| Q4 2025 | Increased to 10.08 M sh + 408 K-share call notional — #1 long |
Risks
- Natural gas price exposure — feedstock cost risk if gas spikes; dilutes the fuel-cell efficiency advantage.
- Backlog conversion — pipeline must translate to acceptance revenue on the implied schedule; project delays compress IRR.
- Competition from gas turbines (GE Vernova, Siemens Energy) and reciprocating gen-sets, which can also be deployed quickly at lower upfront cost (worse opex).
- Single-product company — concentration in SOFC technology; if a competing on-site solution wins share, Bloom has limited fallback.
- Customer concentration — AEP channel and a small number of hyperscaler end-customers represent most of the AI-DC pipeline.
- Long sales cycles — multi-quarter project timelines mean misses can be lumpy and visible.
Sources
- Bloom Energy 10-K and 10-Q filings (annual + quarterly)
- AEP partnership press release (2024)
- Investor presentations
- Earnings call transcripts