✦ Doubling FremontThin moat

Everything on this page depends on a building in Fremont, California. Bloom is spending roughly $100 million to double the plant's annual production run rate from about 1 gigawatt to 2, targeted for completion by the end of 2026, on a site the company says can eventually accommodate around 5 gigawatts of annual capacity¹. Against an order book measured in multiple gigawatts and a backlog reported around $20 billion, the arithmetic is uncomfortable: Bloom has sold considerably more than it can currently make.

Manufacturing scale is also where the economics improve. Fuel cells have never been cheap per kilowatt, and the path to durable profitability runs through volume — spreading fixed costs, buying materials better, and learning the stack down the cost curve. A company that reached a 34.3% non-GAAP gross margin in its first genuinely profitable stretch needs the second gigawatt to prove that number was structural rather than a moment of favourable mix.

The risk is symmetric and obvious. Capacity built for AI-cycle demand becomes a fixed cost if the cycle cools, and Bloom's history of losses is a reminder of what an underutilized factory does to this company's income statement. Watch product revenue against the ramp schedule through 2027, and watch gross margin as the new lines fill. Rising volume with rising margin is the whole bull case made visible; rising volume with falling margin would say Bloom is buying growth with price.