Open in interactive viewer → charts, metric popovers & call review

QumulusAI Doubled Revenue and Got Halved

A first positive EBITDA quarter and a $515M backlog, into a tape that spent October repricing the entire compute buildout on credit risk
QMLS · Earnings Call · 2026-11-10

The quarter the tape didn't want

QumulusAI's third quarter as a public company was, on its own arithmetic, its best yet. Revenue of $14.8 million, up 376% year-over-year and more than double the second quarter's $6.7 million. Adjusted EBITDA positive at $3.7 million, a 25% margin, the first positive print in the company's short public life. Remaining performance obligations of $515 million, roughly three times the $173.1 million at June 30. And the stock is down 50.3% over the 31 trading days since it listed, peaking at $6.53 on October 2 and halving from there — a $108 million market capitalization. That gap is not a company story, it is a tape story, and the tape is legible in the global theme data. AI cloud service names are down roughly 46% over ninety days; data center capacity down 37%; critical IT load down 37%; high bandwidth memory down 32%; co-packaged optics down 32%. Management even narrated it on the call: “The market spent weeks repricing AI infrastructure companies, their credit first and then their equity.” — Michael Maniscalco, CEO · 2026-11-10 Qumulus is the highest-beta expression of that fear — $248 million of debt and finance lease obligations against $25.3 million of shareholders' equity, sitting on top of a legacy Bitcoin mining business in Oklahoma that belongs to another group the market de-rated in the same window.

Reframing the question

The most telling thing on this call is which argument the CEO chose to pick. Three months ago the pitch was contracts, customers, and demand. This quarter he opened by naming the bear case and claiming to have answered it: “The question has shifted from how large the customer contract is to whether we can deliver, finance and run it at a profit. This quarter, we answered all 3.” — Michael Maniscalco, CEO · 2026-11-10 Mostly, he did. Eight megawatts are energized and billing across five colocation sites; 5.45 more are contracted. Third-quarter signings were $352 million, bringing 2026 signings to $529 million. And the funding mechanics are real: customer prepayments of $71 million in the quarter, deferred revenue up to $79 million from $30.5 million, operating cash flow positive $32 million, and roughly $194 million of equipment placed in service — the reason D&A jumped $8 million sequentially.

A customer signs a term take-or-pay contract and puts down a deposit. That deposit covers the portion of the GPU cost that isn't financed and equipment financing tied to that contract covers the rest... We have no unsecured bonds. The customer's commitment funds the hardware and the hardware earns the customer's commitment back. That's the model.

Michael Maniscalco, CEO · 2026-11-10
The contrast with August is stark. Then, the financing environment was an unalloyed positive: “there are more options now than ever that we're seeing out there, and that's encouraging.” — Michael Maniscalco, Chief Executive Officer · 2026-08-25 Now the same executive is describing a credit crunch in the small-colocation market. Every equipment financing tied to a contract is the direct answer to a question investors didn't ask in August but asked relentlessly in October.

Where the constraint moved — and the fine print

In August the bottleneck was internal: “the largest constraint to our growth right now is the land powered shell in the near term.” — Michael Maniscalco, Chief Executive Officer · 2026-08-25 The phrase Land Power shell was a top-three theme of the Q2 call. It has vanished from Q3. In its place: “Demand is not our constraint. Capacity is.” — Michael Maniscalco, CEO · 2026-11-10 And the capacity that is scarce increasingly belongs to someone else — “what we're seeing right now as far as scarcity is the data center capacity itself.” — Michael Maniscalco, CEO · 2026-11-10 The 3.75-megawatt Atlanta tranche slipped from early 2027 to Q2 2027, blamed on legacy tenants vacating on time, retrofit windows, and the owner-operator's financing. The squeeze is second-order and worth understanding: small colocation operators, spooked by the credit repricing, “made a pretty fast adjustment to say that they do not want to sell to non-IG in those instances.” — Michael Maniscalco, CEO · 2026-11-10 The CEO calls that a broad stroke overreaction and says he is pursuing creative ways around it, including a credit wrapper instrument — an admission that the counterparty-credit problem has reached down into the supply side of a business built on small pockets of power. Then there is the crack in the narrative. Michael Donovan of Compass Point asked whether the new 1.2 megawatts in Denver were tied to existing contracts. “The new capacity is that we just signed that's currently being marketed. It is not yet sold.” — Scott Krosnowski, CFO · 2026-11-10 The CEO had already told the room that we sold every megawatt we've energized before it was turned on — true of the past, no longer true of the present. Guidance of 18 megawatts of year-end contracted capacity was reaffirmed with 13.45 megawatts in hand and the balance in late stage negotiation; the $300 million forward ARR target was reaffirmed with $209 million actually signed. Investors can do that subtraction in their heads: roughly 4.5 megawatts and $91 million of annualized revenue to close in six weeks, against a supply chain the company just told them is slipping. One genuinely new note is the mix shift toward anchored by creditworthy enterprise customers, with DRW cited as the template. That is the correct strategic response to a credit-led de-rating — enterprise counterparties make both the lender and the data center owner more comfortable — but it also means the growth rate bends toward enterprise procurement timelines.

What stopped being said

Two concepts quietly fell off the call. The first is Land Power shell, above. The second is convertible notes, a top-five theme in Q2 on the back of the issuance. In Q3 they appear only as a $32.3 million noncash fair-value gain that moves inversely with our share price — which is to say the instruments that flattered the quarter are the ones that hurt the stock. Meanwhile the guidance framework itself got a hedge dressed as housekeeping: “we expect to update the way we provide forward guidance when we report the fourth quarter as we look ahead to 2027.” — Scott Krosnowski, CFO · 2026-11-10 The honest read: Qumulus just crossed from story to arithmetic — $0.5 billion of signed contract value, positive EBITDA, and contract revenue that amortizes its own hardware — at precisely the moment the market decided that only arithmetic on credit counts. The model is coherent. The question the half-off share price is asking is narrower and nastier: at what spread does a $25 million-equity company finance the next block of GPUs, and what happens to the revenue-per-megawatt math if the answer is worse than last quarter's?