Deep|BE: The Debate Has Shifted from AIDC Demand to Delivery Execution
Executive Summary
AIDC power scarcity remains the structural driver. We estimate North America faces a ~15GW AI power shortfall in 2027, as grid interconnection, turbine lead times, and downstream power infrastructure constrain how much announced capacity can actually be energized. This keeps time-to-power at a premium and supports Bloom’s behind-the-meter SOFC value proposition.
The debate has shifted from demand to execution. Production slots appear substantially booked through 2028, but the market is increasingly focused on whether Bloom can convert headline capacity into shippable MW and site deliveries. We model ~2.7GW YE27 manufacturing capacity and ~2.2GW of 2027 deliveries vs. ~2.0GW consensus, with supplier ramp, testing, yield and service capacity becoming more relevant constraints.
Scandium looks manageable near term, but service capacity risk is overlooked. Our channel checks suggest Bloom has diversified scandium sourcing across Malaysia, Japan and Canada, while current pricing does not indicate acute scarcity. The more overlooked risk is whether Bloom’s service organization can scale quickly enough to support a rapidly expanding multi-GW installed base.
Bloom is more than a temporary gas-turbine shortage trade. We do not expect turbine supply to meaningfully normalize by 2030, extending Bloom’s time-to-power advantage. Longer term, modularity and native-DC compatibility with emerging 800V AIDC architectures could give SOFC a structural role even as conventional generation capacity expands.
Backlog conversion is now the key rerating driver. Recent delays at Jupiter and Vineland look more like normal project-level timing risk than backlog impairment, but increasing customer concentration makes annual revenue more sensitive to permitting, construction and acceptance cadence. Following the July short reports, BE’s forward EV/EBITDA has reset from ~85x to ~42x; we see 2H26 delivery execution, 2027 guidance and progress on major AIDC projects as the key catalysts for Street estimates and the multiple to recover.
Why now: Following a series of short reports in July, BE’s forward EV/EBITDA has reset from ~85x to ~42x. Despite a strong 26Q2 beat-and-raise, the stock has yet to regain momentum, suggesting the debate has shifted beyond the well-understood AIDC power demand. We see this as a good time to dig into the more important questions: how quickly Bloom can ramp capacity, deliver, and convert its AIDC backlog into revenue.
Key Debates & Our View
#1 Can Bloom add capacity and, more importantly, deliver fast enough to meet demand?
Our View: We think Bloom can support the near-term delivery ramp. Our channel checks suggest production slots are substantially booked through 2028, with Bloom internally targeting roughly 4GW of annualized manufacturing capacity by YE27 while we model ~2.7GW, reflecting a more realistic ramp in systems manufacturing than the company’s internal stretch target. For actual deliveries, we model ~2.2GW of 2027 deliveries versus ~2.0GW consensus. Scandium does not appear to be an immediate bottleneck, but service capacity could become a more meaningful constraint as the installed base scales.
Debate: Bloom’s demand strength is already well understood, and incremental project announcements alone are unlikely to drive another meaningful rerate. The debate has shifted to whether Bloom can physically supply what it has already sold. However, we think the market is still over-focused on headline capacity expansion and under-focused on actual delivery capability. Most capacity forecasts remain largely top-down, anchored to the size of the AIDC power shortfall and assumed BTM share, rather than a bottom-up assessment of Bloom-specific constraints such as effective manufacturing throughput, yield, supplier readiness, testing capacity, service infrastructure and site-level conversion.
A recent short report has highlighted potential supply-chain bottlenecks, including scandium availability and broader manufacturing constraints. The key question has therefore shifted from headline demand to whether new capacity, suppliers, and yields can ramp fast enough for booked demand to convert into shippable MW on schedule.
We believe the risk is manageable. Management says its diversified supply chain can support up to 25GW/year of production and is not dependent on China, while current scandium pricing shows little evidence of tight supply or an imminent shortage. Our channel checks suggest Bloom has diversified its scandium sourcing since the early 2010s, with supply relationships across Malaysia, Japan, and Canada that together could provide 15 tons of annual ex-China supply, or about half of its current needs. The same checks suggest Bloom’s scandium intensity is closer to ~15 tons per GW, materially below some bearish estimates based on older cell designs and thicker electrolyte assumptions. This suggests scandium is unlikely to be a near-term bottleneck, but the margin of safety narrows as Bloom scales further into multi-GW annual production. We therefore continue to monitor qualified ex-China supply growth and actual scandium intensity per GW as the key variables.
A more underappreciated risk is service capacity. We believe that Bloom’s current service organization is not yet sized for a much larger multi-GW installed base, meaning the company will likely need to expand internal field-service capacity and develop more third-party service partners. As deployments scale, technician availability, spare parts, stack replacement, and field-response capability could become increasingly important constraints on uptime and customer satisfaction, even if manufacturing itself keeps pace.
点赞
收藏
点踩
分享