Bloom Energy’s July 28 test is cash conversion

Oracle made permitting the Q2 setup

Bloom Energy enters its second-quarter 2026 earnings report on July 28, 2026, after Oracle expanded its Bloom relationship. CNBC reported on April 14 that Oracle expanded procurement to 2.8 GW, versus no previously quantified Oracle capacity in the supplied coverage. Oracle also received a $400 million Bloom warrant, versus no prior Oracle warrant disclosed in that coverage. Oracle tied capital to capacity.

Oracle’s Project Jupiter explains the urgency. On July 18, 2026, The Information reported that permitting hurdles pushed Oracle from gas turbines toward costlier fuel cells, adding billions to costs versus the original turbine plan. Oracle paid more because its preferred power design did not clear. Bloom’s place behind the Power Wall depends on moving faster than grids, turbines, and permits.

The broader TEXXR coverage record contains two Bloom articles from two sources, both first seen in 2026. That compares with one article from one source in the earlier entity snapshot and zero Bloom articles in the supplied 2014–2021 archive. The entity graph contains one confirmed Bloom edge, Oracle, versus zero other confirmed customer edges. The street gets one relationship, not a demand curve.

Bloom grew faster than its earnings

Bloom’s supplied FY2025 annual-report extract reports revenue of $2.0 billion, up 37% from $1.5 billion in FY2024. FY2024 revenue had risen 11% from $1.3 billion in FY2023. Bloom accelerated reported sales before Oracle’s 2026 expansion entered the coverage record. The allowed corpus contains no public annual-report URL, so these figures remain extract-level baselines rather than independently linked filing facts.

The same FY2025 extract reports EBITDA of $20.1 million, down 78% from $89.3 million in FY2024. FY2024 EBITDA had swung by $224.4 million from a $135.1 million loss in FY2023. The net loss widened 203%, to $88.4 million from $29.2 million. Bloom’s 37% revenue growth therefore failed to preserve FY2024’s earnings improvement.

The analytical question for July 28 is what changed between FY2024’s $224.4 million EBITDA swing and FY2025’s 78% decline despite 37% revenue growth.

Bloom cut capex while cash improved

Bloom’s supplied FY2025 extract reports capital expenditure of $56.8 million, down 4% from $58.9 million in FY2024. FY2024 capex had fallen 30% from $83.7 million in FY2023. Bloom raised revenue from $1.3 billion in FY2023 to $2.0 billion in FY2025 while cutting annual capex by $26.9 million over the same period. That pattern supports cash conversion, but only if the missing depreciation bridge does not hide a replacement burden.

Bloom’s extract reports FY2025 free cash flow of $57.2 million, up 73% from $33.1 million in FY2024. FY2024 free cash flow had improved by $489.4 million from negative $456.3 million in FY2023. Bloom produced positive free cash flow for two years, versus heavy cash burn in FY2023, even as FY2025 EBITDA fell.

The supplied extract states no depreciation figure for FY2025 or FY2024. Depreciation therefore has no numerical baseline, and its year-over-year change cannot be calculated. The July 28 record needs capex, depreciation, and free cash flow on one bridge. Without that bridge, the street cannot separate lower investment needs from deferred replacement spending.

Oracle proves demand but narrows the record

Oracle supplies Bloom’s only confirmed procurement edge: up to 2.8 GW, versus zero capacity attached to other named customers in the TEXXR graph. Both Bloom articles in the corpus also center on Oracle or Project Jupiter, versus zero articles showing another customer at similar scale. Oracle shows that one buyer accepted Bloom’s cost to escape a permitting block. Oracle does not establish customer breadth.

Bloom’s siting advantage also depends on competing power routes remaining slow. The Information reported on July 18 that permitting hurdles blocked Oracle’s preferred turbine route and pushed Project Jupiter toward fuel cells. Faster grid interconnections, turbine approvals, or competing generation would reduce that constraint. Bloom’s leverage rises when permission is scarce and falls when queues clear.

Bloom’s second-quarter disclosures can test both risks. Revenue, backlog, or deployed capacity tied to Oracle would measure conversion against the 2.8-GW ceiling. The same metrics for non-Oracle customers would show whether one confirmed edge is becoming a repeatable pattern.

The July 28 record has five checks

• Bloom’s second-quarter revenue: The comparison is Q2 2026 against Q2 2025, with FY2025’s 37% annual growth as the wider baseline.
• Bloom’s earnings conversion: EBITDA and net loss need comparison with the same 2025 quarter and with FY2025’s 78% EBITDA decline and 203% wider net loss.
• Bloom’s cash bridge: Capex and free cash flow start from FY2025 baselines of $56.8 million and $57.2 million. Depreciation needs a disclosed FY2025 or quarterly baseline because the supplied extract provides none.
• Oracle’s order conversion: Recognized revenue, backlog, deployment progress, or capacity would measure execution against Oracle’s up-to-2.8-GW commitment.
• Bloom’s customer breadth: Named non-Oracle revenue, backlog, or capacity would compare with the current graph baseline of zero other confirmed customer edges.

Bloom already has the permitting anecdote. The July 28 report must show whether constrained power markets produce cash without reopening FY2025’s earnings gap.

Sources

The record:

• Oracle expands its partnership with fuel cell maker Bloom Energy to procure up to 2.8 GW of capacity, after receiving a warrant to purchase $400M of Bloom stock — CNBC, 2026-04-14 — original · TEXXR record
• Permit hurdles push up costs for AI data centers; Oracle pivoted from gas turbines to costlier fuel cells for its Project Jupiter in NM, costing billions more — The Information, Sat, 18 Jul 2026 00:00:00 GMT — original · TEXXR record