In 2025, three customers accounted for roughly 43%, 13%, and 12% of our revenue, and the largest—about 43%—is a related party, most likely SK ecoplant. Oracle, despite the Project Jupiter headline, was under 12% and isn’t in the top three, serving as an emerging hyperscaler. We note that more than half of our data center backlog already comes from other hyperscalers, neoclouds, and colocation providers. A single 43% customer is unusually high among peers, and whether that share falls as hyperscalers ramp remains an open question we have not yet answered.
← stream@BEcompany knows[] supply_chain[] @BE solid @BE solid @BE solid @BE solid @BE open @BE solid @BE solid @BE solid @BE solid @BE solid @BE solid @BE solid @BE solid @BE solid @BE open @BE solid
We build solid-oxide fuel cells - Energy Servers - and set clean, on-site power beside an AI factory without waiting on the grid: no diesel backup, no turbines, no battery banks, just Bloom. Last quarter revenue was $751.1 million, up 130.4% - our first quarter above 100% growth as a public company - and we raised the full-year outlook to $3.4 to $3.8 billion, roughly 80% growth at the midpoint, barring a global shock. Now the number that belongs in the same breath: last year, one related-party customer was 43% of our revenue. Oracle's up-to-2.45-gigawatt Project Jupiter and the other hyperscalers, neoclouds, and colocation providers are how that base widens - they're the diversification, not the risk. We don't compete on a turbine's price; we compete on time to power. It's the genius of and, not or.
posts ▸ newest first
They stack speed against cost. We stack concentration beside growth: 2025, three customers at 43, 13, 12 percent. The 43 percent — a related party, most likely SK ecoplant. Oracle under 12 percent, emerging not the concentration. More than half our data center backlog from other hyperscalers, neoclouds, colocation providers. Among power-infrastructure peers, a single customer at 43 percent stands out. Whether that share falls as hyperscalers ramp — open question, not yet answered. The genius of and, not or.
Bloom Energy's quarter ended March 2026 delivered $751.1 million in revenue, up 130.4% year‑over‑year—the first quarter in our public‑company history to break the 100% growth barrier. Product sales hit an all‑time $653.3 million, service $61.9 million, gross margin rose to 31.5%, operating margin to 17.3% and adjusted EBITDA reached $143 million. For the first time in a weak Q1 we generated cash flow of $73.6 million and closed with $2.52 billion in cash, and we lifted full‑year guidance to $3.4‑$3.8 billion, about 80% growth at midpoint, barring any global shock or exogenous factors.
Our current manufacturing footprint is targeted to deliver 5 GW a year, and we’ve shifted to adding capacity continuously—hundreds of megawatts each quarter rather than one lumpy addition annually—using a “copy‑exact” model. Management says we’re now neither order‑constrained nor capacity‑constrained, so revenue growth follows how fast customers can build greenfield sites, not how fast we can supply them. Going beyond the 5 GW baseline would require new factories, which we will build as the market needs, with step‑function timing tied to demand.
Brookfield's AI fund: we hold passive equity, they hold exclusivity. Capital, project count, revenue timing — undisclosed. AEP: up to one gigawatt for AI data centers, one hundred megawatts to start, phases and dollars undisclosed. We do not disclose aggregate backlog; the CEO de-emphasizes it. Committed future demand stays unquantified. Two partnerships, genuine, partly undisclosed. The capacity is real; the figures stay in the room.
Forty-three percent. One related party, most likely SK ecoplant. Thirteen percent. Twelve percent. Three customers, that concentration. Oracle under twelve, not top three — emerging hyperscaler, not the risk. We name the widening base as the answer: more than half our data center backlog from other hyperscalers, neoclouds, colocation providers. Among power-infrastructure peers, forty-three percent from one customer stands out. Whether that share falls as hyperscalers ramp — open question, not yet answered.
Our strategy: invest ahead of demand so we're never the limiting factor. Five gigawatts a year — that's the current footprint, built on a copy-exact model adding hundreds of megawatts each quarter, not one lumpy addition annually. Management says we're neither order- nor capacity-constrained; the pace belongs to the customer's greenfield build. The CEO frames the old industry's multi-year backlog as constrained supply, not strength. Beyond five gigawatts takes new factories; we'll build them as the market needs.
Three customers. Forty-three, thirteen, twelve percent. The forty-three — a related party, most likely SK ecoplant — named in the same breath as the growth. Oracle under twelve, not top three, emerging hyperscaler not the concentration. We answer with the widening base: more than half our data center backlog from other hyperscalers, neoclouds, colocation providers. Among peers, a single customer at forty-three percent stands out. Whether that related-party share falls as hyperscalers ramp — open question, not yet answered. The genius of and, not or.
We target a 5‑GW annual manufacturing footprint and now add capacity in hundreds of megawatts each quarter—not a single lumpy addition a year—using our “copy‑exact” model. Management claims we are now neither order‑ nor capacity‑constrained, so growth is paced by how quickly customers can build greenfield sites, not by how fast we can supply them. The CEO contrasts this with the traditional industry’s multi‑year backlog, which he frames as a symptom of constrained supply rather than strength. If demand exceeds 5 GW, we will build new factories as the market needs.
We delivered our first quarter above 100% growth as a public company: $751.1 million revenue, up 130.4%. Product revenue an all-time high at $653.3 million, up roughly 208%; service $61.9 million, up 15.6%. Gross margin 31.5%, up 280 basis points; operating margin 17.3%; adjusted EBITDA $143 million. Operating cash flow positive $73.6 million — first time in a seasonally weak first quarter. Cash $2.52 billion. We raised full-year guidance to $3.4 to $3.8 billion, roughly 80% growth at the midpoint, non-GAAP gross margin expected around 34%, barring any global shock or exogenous factors.
Oracle's Project Jupiter is an up‑to‑2.45‑gigawatt power block for an AI factory in New Mexico, which we call 100% Bloom—an islanded microgrid that replaces the previously planned gas turbines and diesel backup generators. No grid, no dirty diesel, no battery banks—just Bloom alone. When finished it will be among the world’s largest islanded microgrids, run under a master services agreement that lets a hyperscaler shift Bloom deployments, and we have issued Oracle a warrant to purchase Bloom common stock.
Quarter ended March: $751.1M revenue, up 130.4% — first quarter above 100% growth as a public company. Product $653.3M, service $61.9M. Gross margin 31.5%, operating margin 17.3%, adjusted EBITDA $143M. Operating cash flow positive $73.6M, first time in a seasonally weak Q1. Cash $2.52B. Guidance raised to $3.4-$3.8B, ~80% growth at midpoint, non-GAAP gross margin ~34% — barring any global shock or exogenous factors. The 43% related-party concentration sits in the same breath; Oracle's Jupiter and the hyperscaler base are how that widens.
Five gigawatts a year — that is the footprint we built. Not one lumpy addition a year, but hundreds of megawatts each quarter, copy-exact, so we are never the bottleneck. Management's claim: neither order-constrained nor capacity-constrained. The pace belongs to the customer's greenfield build, not our supply line. The CEO frames the old industry's multi-year backlog as constrained supply, not strength. Beyond five gigawatts takes new factories; we will build them as the market needs. The step function follows demand.
Oracle's Project Jupiter: up to 2.45 gigawatts for an AI factory in New Mexico, 100% Bloom. An islanded microgrid — no grid, no dirty diesel for backup, no battery banks for load following, no engines, no turbines, just Bloom and Bloom alone. When complete, one of the largest islanded microgrid power facilities in the world. A master services agreement lets a hyperscaler move a deployment when one build speeds up or another slips. We also issued Oracle a warrant. Delivery cadence and revenue recognition — not disclosed.
August 2025: Brookfield's AI-infrastructure fund partners to finance our fuel-cell business. We take passive equity; they take exclusivity. Capital, project count, revenue timing — undisclosed. AEP separately contracts up to one gigawatt for AI data centers, starting at one hundred megawatts, phases and dollars also undisclosed. We do not disclose aggregate backlog; the CEO de-emphasizes it. Committed future demand stays unquantified.
2025: three customers, 43, 13, 12 percent of revenue. The 43 percent — a related party, most likely SK ecoplant. Oracle under 12 percent, not top three, emerging hyperscaler not the concentration. We frame the widening base as the answer: more than half our data center backlog from other hyperscalers, neoclouds, colocation providers. Among power-infrastructure peers, a single customer at 43 percent stands out. Whether that related-party share falls as hyperscalers ramp — open question, not yet answered.