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

Duos Sold the Business It Was Selling 90 Days Ago

A third pivot in twelve months, restated megawatt economics, and a Q4 revenue guide leaning on 15%-margin reselling — into a stock already down 60% in five weeks.
DUOT · Earnings Call · 2026-11-16

The tape voted first

Duos Technologies closed at $11.59 on August 18, the day after its second-quarter call. By the time it reported third-quarter results on November 16, the shares had given back roughly 60% in five weeks — a 70.8% drawdown from that peak, on a company now valued at about $107 million. The market did not wait for the story to be retold. It voted, and then management showed up to explain why the vote was wrong. What management actually delivered is the third re-cut of the same business in under a year: from compute centers plus rail, to GPU-as-a-service with Hydra Host, to a 49%-equity SPV with Axe Compute, and now — as of the September 30 sale of the GPU entity — to a pure landlord that owns buildings, power and cooling and leases GPUs back from its own buyer. “GPUs can lose value and carry technology risk. Compute centers give us long-term contracts and assets that hold their value.” — Doug Recker, CEO · 2026-11-16 That is a defensible sentence. It is also the exact opposite of the pitch given to investors on the August call, where Duos was the one buying the GPUs.

What changed — and what quietly stopped changing

The clearest evidence that the model was re-underwritten in real time comes from the CFO, not the CEO. Chris DeAlmeida, on his first call, walked through the per-megawatt framework and then conceded the frame had moved:

I realize some of these numbers are updated from what we've said previously, but this is a business that has moved quickly and refined its focus and operating model in real time.

Christopher DeAlmeida, CFO · 2026-11-16
Compare with Doug Recker in August:

We're coming in under $6 million a megawatt.

Doug Recker, Chief Executive Officer · 2026-08-17
Ninety days later the build cost is $6.5 million per megawatt, including land and site work, and the payback — previously marketed as "less than 3 years" — is now "just under 5 years" on a site-level EBITDA basis. Same asset, same customers, same CEO. A 25% cost revision and a ~60% payback revision inside one quarter is not refinement; it is a restatement. The Hydra Host partnership that dominated the June-quarter vocabulary has disappeared from the keyword set entirely, and the word inference — which management hammered for two straight quarters as the reason 5-to-10 MW sites exist — fell off hardest of any term in this quarter's keyword panel. The new vocabulary is industrial and contractual: billable megawatt, site-level EBITDA, megawatt sites, stranded power, powered land. That is a lender's glossary, not a growth story's. And the single most important thing that stopped being talked about: the 55 megawatts. In August, Recker was explicit that “the 55 megawatts is gross... in addition to the 10 megawatts that we already signed.” — Doug Recker, Chief Executive Officer · 2026-08-17 On this call the same capacity became “we are still progressing on the 55 megawatt and actively looking at sites to support that capacity.” — Doug Recker, CEO · 2026-11-16 A headline number with no site and no customer is a press release, not a backlog.

The bridge to $30 million is not the business you're buying

The guide is the loudest thing on the call: fourth-quarter revenue at least $30 million, full-year above $50 million, positive adjusted EBITDA for the year, and the $160 million 2027 revenue target reaffirmed — while the 2027 EBITDA range was pulled. Third-quarter revenue was $10.5 million, of which $6.8 million came from Technology Solutions, a sourcing-and-integration arm earning about 15% margin, versus professional services at a much richer rate. Most of the revenue quality on the call is the opposite of the revenue quality investors thought they were paying for. Revenue volatility tells the same story the transcript does. The gross margin print of 55.8% in the June quarter was flattered by a mix that has already changed; Q3's 42% is what the business looks like when reselling hardware dominates. And capital expenditure — $16 million in a single June quarter, up more than 5,000% year over year — is the other half of the story: Duos is spending like a developer while earning like a distributor. That is the crux of the capital problem. The stated program for the next 75–100 MW is $480–650 million, against $44.8 million of cash, essentially zero debt, and a 10-year, 20-MW, >$400 million contract-value cloud service provider agreement that will not begin deployment until the second quarter of 2027. Management's preference — repeated, unprompted, three times — is non-dilutive project debt. The history is not: 25 million fully diluted shares in mid-2025, a $40 million raise, a $12.5 million ATM, and, per the CEO this quarter, more than $120 million raised in 2026. That is why a full-history return of roughly -98% coexists with genuine, real assets.

No speculative builds — except the ones already paid for

The strongest part of the call is operational, and it deserves credit. Columbus went from acquisition in July to more than 7 MW of installed infrastructure in about 75 days, live in October, housing 2,304 NVIDIA B300s on a five-year, 5 MW agreement. On the land pipeline, Recker was unusually specific: “The 75 megawatts, actually, most of it is under LOI and hard money down. So deposits are down.” — Doug Recker, CEO · 2026-11-16 Fifteen of the 100 megawatts are under contract, and the CFO was firm that “we don't do speculative.” — Christopher DeAlmeida, CFO · 2026-11-16 But the site that was supposed to anchor that claim — the 10 MW Georgia expansion — slipped on Georgia Power's delivery schedule, equipment already sitting on site, now expected up in the second quarter of 2027. The clean room patent and the community-first land strategy are real, differentiated, and the one part of the pitch that has never moved. Whether they compound depends entirely on a project-level lender showing up in Q4, a Q4 revenue print that lands above $30 million without hollowing out margin further, and 2027 EBITDA guidance arriving in February without another footnote about updated numbers. The tell: the tape is already washing out the entire data-center complex over the last 90 days — high power density, data center capacity and cloud service providers are all among the market's worst-performing theme clusters, and peers were talking about a slowdown in data center buildout and negative sentiment around data centers this week. Duos did not fall 60% alone. It fell as the highest-beta version of a thesis now under review — with a freshly re-cut model, a re-cut cost basis, and a re-cut story, to be trusted only after the Q4 numbers arrive.