Bloom Energy's Power Connect isn't the headline — the deployment calendar is the stock

"We delivered 55 megawatts for Oracle in 55 days — down from 90 days." KR Sridhar, founder and CEO of Bloom Energy, said that in a July interview. Nineteen days after he said it, the company announced Power Connect, a new deployment system that, in Bloom's words, "cuts onsite power installation time by over 40%." The announcement was quiet by the standards of the stock it came from. Bloom is up roughly 197% this year, trades around $258 with a market cap near $76 billion, and sits at about 310 times trailing earnings — a number that makes most of the AI-power complex look boring by comparison. And the headline on a press release about faster wiring is exactly the kind of thing the market usually shrugs off.

The question for an investor who has watched this stock do what it has done all year: is the installation detail, or the $20 billion backlog everyone cites, the number that actually determines whether Bloom gets here? It turns out the backlog is the easy number. The real question is whether a company selling speed can deliver it, consistently, on a schedule, at a price. Power Connect is Bloom's answer to that. The 40% is a claim. The mechanism is the argument.

What Power Connect actually does

The claim is that the system cuts onsite installation time by over 40% by "moving critical electrical integration from the field to the factory." The unit of work — the thing you are actually paying a field electrician to do — is wiring, joining, and commissioning the fuel cell stacks into a live system. That work has historically been done on the customer's site, by electricians on the ground, against weather, site quirks, and the usual chaos of a construction crew. Power Connect pre-connects and pre-wires the systems, and tests them in Bloom's factories before they ship. On the site, electricians skip the repetitive wiring and integration work and focus on connecting to the building and commissioning.

Bloom itself flags the 40% figure as a forward-looking statement — the kind of number that gets a caution in the same press release. Treat it as such: it is a claim, not a measurement. Nobody has published a third-party audit of the reduction. The engineering is, on its face, plausible — moving assembly from a field environment to a factory is the same move that made solar panels and modular substations faster to deploy. The interesting question is not whether the reduction is real in the lab but whether it holds up across hundreds of sites simultaneously, with different site conditions and a workforce that is, per Bloom's own release, already strained by "skilled labor shortages."

Bloom is not selling a cheaper fuel cell. It is selling a faster one, and the installation bottleneck is the product.

The economics that make speed the whole pitch

Fuel cells are more expensive than the alternatives on a per-kilowatt basis. The installed capital cost for Bloom's solid oxide fuel cells is roughly $3,000 to $5,000 per kilowatt, versus $800 to $1,500 per kilowatt for a simple-cycle gas turbine and $300 to $700 per kilowatt for diesel generators. At the midpoints, a Bloom system costs roughly 3.5 times as much as a gas turbine to install. That is the number the sell-side glosses over, because the story they tell is about the 166% year-over-year revenue growth and the record margins, not the fact that customers are paying a serious premium for each kilowatt.

The reason anyone pays that premium is that grid interconnection queues in many U.S. markets now stretch three to seven years, and new gas turbine projects carry five-plus-year lead times. Bloom's modular systems deploy in under twelve months, and the 55-day Oracle delivery — vendor-reported, from a 90-day promise — is the extreme of that. The entire pitch is that the premium is worth it because you get power in months, not years.

That means the installation step is not an afterthought. It is the reason the price is what it is. If a fuel cell can be dropped in and wired in the field in weeks, the customer pays the premium because the alternative is waiting years. If that installation step slows down — if the wiring is inconsistent, if electricians are scarce, if commissioning drifts — the customer's schedule advantage shrinks, and the premium starts to look less like a bargain and more like a line item. The 40% number only matters in proportion to the premium it protects.

Backlog is not the constraint

Bloom's total backlog at the start of 2026 was $20 billion, split roughly into $6 billion of product backlog — up 140% year over year — and about $14 billion of service backlog, the annuity-like stream that grows with every fuel cell the company has ever sold. The product number is the one that matters here, and it maps cleanly onto what the company can actually build.

The anchor customer is Oracle. An initial 1.2 gigawatts of fuel-cell capacity has been contracted and is being deployed at Oracle's U.S. projects in 2026 and 2027, under a master agreement that can reach 2.8 gigawatts. AEP has committed to up to 1 gigawatt, with an initial order of 100 megawatts. Equinix has more than 100 megawatts across 19 data centers, of which roughly 75 megawatts were operational as of mid-year. Brookfield, which had a $5 billion commitment to deploy Bloom's cells, has expanded that relationship to $25 billion.

None of these are small, and the $20 billion total looks like the kind of number that should make the stock look cheap. It is not. The constraint is not demand. The constraint is production.

Bloom's Fremont, California factory is the main production line, and the company's stated target is to double its annual output from 1 gigawatt to 2 gigawatts by the end of 2026. That is roughly a third of the Oracle master agreement and a fraction of the total backlog. In other words, the backlog is several years of production at the target rate, not a single year's revenue. Which is precisely why a 40% cut in installation time is not a footnote: it is one of the few levers that lets the company convert the backlog into delivered megawatts faster without a corresponding step up in factory output. The backlog is a promise. Delivery is the product.

What the valuation already knows

The stock's multiple tells you what the market is already assuming. At about 24.5 times trailing revenue and roughly 310 times trailing earnings, Bloom trades at a premium to almost every peer in the electrical-equipment complex. GE Vernova is at about 5.9 times sales. Cummins is at about 2.2 times. Eaton is at about 5.3 times. Even Vertiv, a high-multiple AI-infrastructure name, is at about 8.3 times sales and 55 times trailing earnings. Bloom's multiple is not a mistake; it is a statement. The market has priced in a company that hits its guidance, ramps to 2 gigawatts on schedule, and converts a $20 billion backlog into delivered megawatts without a meaningful execution stumble.

The stock's own history this year supports that reading. A short-seller report in early July challenged whether the $20 billion backlog was as solid as it looked, pointing to gaps between the company's unaudited backlog and its audited remaining performance obligations, and raised supply questions around scandium, a rare-earth input to the fuel cell stacks. The stock dropped roughly 19% over the following month. A separate regulatory wrinkle — a second rejection of a natural gas pipeline permit for Oracle's Project Jupiter, which is tied to up to 2.5 gigawatts of Bloom fuel cells — added a different kind of risk to the mix, because the fuel cells need a continuous gas feed to run.

Bloom has denied the supply concerns, and its most recent quarter — $1.065 billion in revenue, up 166% year over year, with non-GAAP operating margin at 22.5% and EPS of $0.78 against a $0.40 consensus — is the kind of print that makes most of the criticism look premature. The company is profitable, cash-flow positive, and raising guidance every quarter. The problem is not that the numbers are bad. The problem is that the price assumes the numbers stay good.

Bloom's inclusion in the S&P 500 on September 21 is a mechanical bid from passive flows, not a fundamental signal. It will pull money into the name that was not there before, and that matters for a stock at this multiple. It does not change the underlying question, which is the same one Power Connect is trying to answer: can the company keep delivering at the pace the price assumes, and at the margin the numbers suggest?

The backlog is not the story. It is the setup. The number that decides whether Bloom keeps compounding or stumbles is the one on the deployment calendar — how many megawatts a quarter, and at what cost to install them. The 40% is the first independent data point that would tell you.