BE

Bloom Energy makes fuel cells — refrigerator-sized boxes that generate electricity onsite from natural gas or hydrogen, without connecting to the grid. The company is over 20 years old and spent most of that time as a niche clean-energy story. Then AI data centers ran into a problem nobody solved for them: you can build a data center in 18 months, but getting new grid power connected can take five years or more. Bloom sells the answer. Last quarter it crossed $1 billion in revenue for the first time, up 165% from a year ago, with adjusted earnings of $0.78 per share against the $0.42 analysts expected. The stock rose about 10%. The part that impressed us most isn't the growth — it's that the cash showed up. We've flagged repeatedly this year that a lot of AI-adjacent companies report strong profits while burning cash to build capacity. Bloom went the other direction: operating cash flow swung to a positive $226 million from a negative $213 million a year ago, and the company ended the quarter with $2.7 billion in cash. Gross margins expanded to 34% and operating income went from roughly breakeven to $182 million. Management raised full-year guidance to $3.9–4.2 billion in revenue. And Brookfield, the infrastructure investment giant, expanded its partnership with Bloom from $5 billion to $25 billion — an outside party with real money agreeing this is worth financing at scale. Two cautions. First, insiders have sold Bloom stock 32 times in the past six months and bought exactly zero times, including one executive selling about $41 million worth. That's now the fifth or sixth company where we've flagged this, and while it's rarely a smoking gun, it's a consistent pattern across the whole AI complex. Second, the stock has been violent: up about 70% for the year, but down nearly half from its June high before this report. Fuel cells also face real competition — gas turbines, other fuel cell makers, and eventually the grid itself catching up. That last one is the actual risk worth watching. Bloom's advantage exists because utilities can't build fast enough. The thing to track is the backlog and whether new customers keep signing, because the day grid interconnection times start shrinking is the day Bloom stops being essential and starts being one option among several. (written with the help of Claude)