Good call on Bloom Energy $BE with a director on Crusoe's energy team TLDR: Constructive on $BE through the speed-to-power window: everything they can build gets sold, the regulatory-constraint use case is expanding, and the edge market is a genuine second leg. The bear case is not execution risk, it is terminal value—an unsubsidized, unconstrained power market where Bloom's cost curve has not moved enough. Key insights: Behind-the-meter adoption is not a technology preference, it is a hedge against binary regulatory risk. Moratoriums, temporary pauses and Governor Abbott's ERCOT announcements do not degrade a project's economics—they kill it outright, and approved power becomes worthless if a data center moratorium lands on top of it. He has seen two separate tenants elect to proceed with behind-the-meter projects alongside their grid portfolio, explicitly as diversification against that risk. Crusoe deliberately sites behind-the-meter gas away from rural communities to strip out both grid-connected regulatory exposure and community sentiment risk. Texas is the tell. It has historically been the easiest place in the US to get grid power because ERCOT is deregulated—no capacity market booking, no requirement to point specific generation at a specific load, with scarcity pricing left to incentivize the build-out. The fact that delays and regulatory uncertainty are showing up there is what has shifted tenant behavior. Working from a 100 GW / five-year data center demand frame, which he treats as aggressive but attainable: Six months ago: roughly 70 GW served by grid, assuming turbine production picks up, no fuel constraints, and several other unlocks. Some observers penciled 10-20 GW of SMR by end of period. Today: near-term grid share compresses to perhaps 50-60 GW, with the forward split moving closer to 50/50. Then it reverts. Once ratepayer protection is formalized, the grid reasserts as the cheapest and most reliable long-term supply. The mechanism for reversion is the Arizona construct APS has pushed—growth pays for growth, where all incremental upgrade costs are borne by the data center operator. He expects that to be cemented and formalized across every market, and expects it to take six to twelve months before politicians stop feeling their seats are threatened. The Bloom Bull Case He does not dispute near-term demand at all. "as much reliable Bloom capacity or solid oxide fuel cell capacity that can come online, will be deployed." If US deployable manufacturing produces 2-8 GW over the next two years, it gets absorbed. The premium is not a problem for buyers whose binding constraint is capacity. He also believes the product works. Hundreds of megawatts is deliverable, the systems are reliable, downtime is low because units are swappable, and the architecture is fully modular. The Bear Case Is About Price - "I don't see how they compete on price." Ten years out, in an unconstrained power market, he does not believe Bloom is competitive. He specifically does not believe the roughly 10% annual cost-down, on the grounds that the technology is structurally hard to make cheaper. He also flags a cost item he thinks the market underweights: the full system swap over a ten-year cycle, which makes ongoing O&M more expensive than a gas gen equivalent. Initial capex is the wrong lens; actual LCOE is critical to assess. He extends this into a coherent explanation of Bloom's own behavior. On the question of why they have not simply built the next facility if demand is as strong as claimed — his answer is not that they are sandbagging. He thinks they are building as fast as they can, and that the constraint is the supply chain, not the building. Solid oxide is an extremely small US market. You can put up a structure; scaling the full assembly chain behind it on the same timeline is the harder problem, unless more of it moves offshore. Crusoe Has Zero Bloom Projects Today - "Today, actually, we don't have any projects that are relying on Bloom fuel cells." The strategic rationale for staying at arm's length: "We've been a follower in this instance... we will accept it once the utility does." Buying 500 MW of Bloom directly means absorbing the regulatory risk that utilities may not accept solid oxide as high-rated ELCC capacity. Let the utility carry that. Turbines, recips, aeros and engines are all accepted technologies with known extreme-weather behavior and established effective load carrying capacity ratings. Solar, wind and battery now have them too. Solid oxide does not, because utilities have not yet observed large-scale fuel cell fleets through heat events and cold snaps. AEP has gotten comfortable off the back of the roughly 80 MW deployment plus smaller installations and Bloom's published test results. That is the template, and it is why the next few hundred megawatts of live operating hours matter far more than any order announcement. Where Bloom Actually Wins His model is not that Bloom wins on merit in a fair fight. It is that Bloom is the path of least resistance when a specific constraint blocks a project that already has hundreds of millions of development dollars sunk into it. If the binding constraint is price, Bloom loses. If it is speed to power, Bloom sometimes wins. If it is emissions, air permitting, noise or a regulatory restriction someone failed to plan for, Bloom wins. He reads the Nebius Vineland switch from gas gensets to solid oxide exactly this way — anti-genset pushback threatening a contract worth billions, with an obvious substitution available. He expects more of that, in lumpy project-specific chunks rather than as a smooth share gain. He also identifies the next leg of NIMBY-ism, which he thinks is underpriced: if communities dislike data centers, they dislike new gas generators considerably more. A fuel cell is lower emissions, quieter and a different class of asset. He calls it artful. That is a real, non-obvious tailwind. The Edge and Inference Market Is the Bigger Prize Crusoe is planning heavily for 10-50 MW modular builds, which he sizes at 20-40 GW over five years and would anchor at 20 GW in isolation. Amazon, NVIDIA, Tesla and xAI are all chasing the same edge market. "Bloom will leapfrog any gas combustion." In metro locations you cannot air-permit gas gen at all — this is a hard prohibition, not a noise preference. But he immediately caps the enthusiasm: much of that edge capacity is low-hanging fruit, converted Bitcoin sites at 5-18 MW with existing grid interconnects. "Grid power will always beat it." Bloom is the answer for incremental capacity and backup, not the base case. This is why he holds solid oxide at roughly 10 GW of the 100 GW mix, against 60-70% backstopped by combustion gas. He validates the native DC output argument, but for a better reason than efficiency. Fewer conversion losses lower the price, yes. The larger point is that it removes dependence on transformers and switchgear — equipment that is not only expensive but carries lead times that are themselves the binding constraint. A 345 kV breaker is two years out. Engineering around that bottleneck is precisely the kind of scenario where a project with sunk capital selects Bloom. What Changes His Mind He names two variables explicitly. Power price curves — if Bloom's costs do not fall, or if turbine pricing keeps rising, the relative position shifts and Bloom is in the money reasonably soon. And political sentiment, which he calls a big unknown and which he thinks is the more likely driver of fuel cell share than any technical milestone. Notably, a successful 200 MW deployment alone does not change his view. He already believes they can do it. source: Tegus