Evercore Says AI Demand for Bloom Energy Is Bigger Than You Think. The Real Problem Is the Price Tag.

Evercore ISI's Adam Sparkes recently reiterated an Outperform rating on Bloom Energy, arguing that AI's reach into the company's business is deeper than investors realize. The logic is sound: data centers need power, the grid can't always deliver it fast enough, and Bloom's solid oxide fuel cells offer an on-site alternative that sidesteps years of utility interconnection delays.

The problem isn't the premise. The problem is that the market has already run from roughly $35 a share a year ago to a $60 billion market cap, pricing in near-flawless execution across multi-gigawatt contracts with Oracle, Nebius, and Brookfield. At 267 times trailing earnings and 79 times forward earnings, Bloom Energy no longer trades like a company proving its business model. It trades like one that has already won.

I've been very surprised that investors aren't more troubled by what this valuation demands.

The revenue story is real — and that's the point

Bloom Energy's second quarter was a genuine breakout. Revenue hit $1.065 billion, a 165 percent year-over-year jump. Product revenue alone surged 215 percent to $935 million. Gross margins expanded 604 basis points to 34.3 percent on a non-GAAP basis. Operating income swung from a $3.5 million loss in Q2 2025 to $182 million in Q2 2026. Operating cash flow flipped from a $213 million burn to $226 million of generation. Free cash flow — operating cash flow minus capital expenditures — came in at $175 million.

Management then raised full-year 2026 guidance to $3.9 billion to $4.2 billion in revenue, with non-GAAP operating income of $800 million to $900 million. That implies roughly 100 percent year-over-year revenue growth at the midpoint.

These are not fabricated numbers. They're the kind of quarter that deserves attention. But attention and ownership are different things when the stock has already moved 169 percent year-to-date and more than 800 percent over the trailing twelve months.

What 267 times earnings means in practice

Forward P/E of 79 looks more grounded than the trailing multiple, but it only works if Bloom delivers at or above the top end of its own guidance — and then keeps accelerating from there. At that multiple, the market is paying for four to five years of current earnings power upfront, assuming no execution missteps, no supply-chain disruptions, and no slowdown in hyperscaler capital expenditure.

Compare that to GE Vernova at roughly 51 times earnings, Siemens Energy at 58 times, or Constellation Energy at 26 times. None of those companies have Bloom's growth trajectory. But none of them demand perfection either.

Bloom also carries $2.48 billion in recourse debt — long-term borrowings where the company's balance sheet is directly on the hook if projects underperform. That debt sits against $2.7 billion in cash, leaving net liquidity of roughly $220 million on a company whose quarterly capex is accelerating. The balance sheet is not precarious, but it isn't bulletproof either. Every quarter of heavy investment to meet deployment timelines tightens that cushion.

The Hunterbrook attack and the supply-chain question

The false narrative around Bloom Energy isn't just about valuation. It's also about whether the market has properly stress-tested the execution risks. In early July, short-seller Hunterbrook Capital published a report titled "Bloom's Big Lie," alleging that Bloom has understated its reliance on Chinese sources for scandium oxide — a rare-earth input critical to the conductivity and durability of its fuel cell stacks.

Bloom categorically rejected the claims, noting that scandium oxide is used only in small quantities and that it maintains a diversified global supply chain developed over two decades. The stock dropped 12 percent on the news and then recovered.

That recovery tells you what matters more than Hunterbrook's thesis or Bloom's denial: at these prices, the market doesn't want to hear about supply-chain risk. It's priced the stock as if Bloom has already solved every manufacturing constraint. The 12 percent drop was treated as a speed bump because the consensus narrative — AI power demand is infinite, Bloom has $20 billion in backlog, contracts are signed — proved more durable than a short report.

But consensus narratives are how stocks get expensive before they get challenged by something harder than a short seller. A material supply bottleneck, a delayed deployment, a hyperscaler that shifts its power strategy — any one of those would force a repricing that's currently impossible to absorb at 267 times trailing earnings.

The backlog illusion

Bloom's $20 billion total backlog sounds enormous. It's important to understand what that number means and what it doesn't. Product backlog is approximately $6 billion, representing 2.5 times year-over-year growth. The remaining $14 billion is service backlog — contracted maintenance agreements with terms ranging from 5 to 20 years on already-installed fuel cells. Service backlog is valuable recurring revenue, but it doesn't represent new orders. It represents past sales that will generate steady but modest-margin income over time.

So the growth engine is really the $6 billion product backlog, plus the pipeline of announced deals with Oracle (up to 2.8 gigawatts), Nebius (328 megawatts across three phases), and Brookfield (a commitment that expanded from $5 billion to $25 billion). Those are headline-grabbing contracts. But contracts aren't revenue until they're deployed, and deployment in the power infrastructure world takes time, site work, permitting, and physical manufacturing capacity.

The free cash flow test hasn't run long enough

Bloom generated $175 million of free cash flow in Q2 2026. That's impressive for a company that burned $213 million in operating cash flow in the prior-year quarter. But one quarter of positive free cash flow doesn't establish a pattern. Free cash flow in Q1 2026 was roughly $47 million — positive, but thin. Throughout 2025, the company barely turned positive on a full-year basis, posting $114 million in operating cash flow for the year. The business has been loss-making since at least 2019, and the first sustained profitability is happening right now.

That's the structural tension. The stock is valued like a mature growth company with a proven free cash flow engine. The reality is a company that has only just started printing free cash flow and faces a capital-intensive manufacturing ramp to meet its own raised guidance.

What I believe

The AI power demand story behind Bloom Energy is structurally real. Data centers need reliable on-site power, and fuel cells fill a genuine gap between grid capacity and deployment timelines. The revenue acceleration, margin expansion, and backlog growth are all confirmed by the numbers, not inferred from analyst optimism. Oracle going all-in on fuel cells for its New Mexico AI campus instead of gas turbines is a meaningful validation.

However, the market has already priced in that validation — and then some. At a $60 billion market cap, with 267 times trailing earnings, 79 times forward earnings, $2.48 billion in recourse debt, and a free cash flow track record of one strong quarter, the margin for error has evaporated. This is no longer a stock where you get paid for taking execution risk. You're paying a premium to hope the execution is flawless.

In my opinion, the false narrative here isn't that investors underestimate Bloom's AI demand. It's that the stock's move has been so large, so fast, and so narrative-driven that investors have forgotten what multiple they're paying. The Evercore note is directionally right about the opportunity but silent on the valuation trap that comes with pricing a story before the cash flow has had time to prove it.

I rate Bloom Energy as a Hold. The underlying business trajectory is compelling, but at current levels, the risk-reward is unattractive. A pullback toward the $180-$200 range — where forward earnings would still stretch but not demand perfection — would present a more defensible entry. For investors who already own the position, I don't recommend selling a company with this kind of growth momentum and backlog visibility. But for those on the sidelines waiting for the AI infrastructure trade to catch up, Bloom Energy has already caught up. The question now is whether its earnings can justify a stock price that has moved years ahead of itself.