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

DigitalOcean Grew 40% and Got 50% Cheaper: The Megawatt Trade Is Being Repriced, Not the Business

A fifth straight quarter of acceleration, a third raise to 2027 guidance, and a stock 40% lower in 90 days — the gap between the P&L and the tape is the story.
DOCN · Earnings Call · 2026-11-03

The Disconnect

DigitalOcean printed the best quarter in its public history on 2026-11-03 and its stock has spent the last three months being sold. Revenue of $321 million grew 40% year over year — “Revenue grew 40% year-over-year, our highest in a decade and our fifth consecutive quarter of acceleration.” — Padmanabhan Srinivasan, Chief Executive Officer · 2026-11-03 A record $160 million of net new ARR landed in the quarter, adjusted EBITDA margin was 45%, and management raised full-year 2027 growth from 50%-plus to 60%-or-more, off a higher 2026 base. The tape disagrees violently. DOCN is down 39.6% over the last 90 trading days and sits 50.1% below its 2026-06-15 peak of $181.29 — an unwind that followed a +171% ten-week melt-up. That is the whole dossier in one line: an accelerating operating business inside a de-rating theme.

The Language Rotated Before the Numbers Did

This is the most telling thing in the transcript. In the prior quarter, DigitalOcean's two highest-ranked company keywords were Open weight and Inference Engine — the open-weight model adoption story and the inference router. Both have fallen clean out of this quarter's top set, and "Open weight" has migrated into the decliners, alongside "OpenRouter." Nine months ago Paddy was hammering the same point from the opposite direction: “you, I think, quoted that it's risen from roughly 15% to now nearly 75% of token volume since launch.” — Padmanabhan Srinivasan, Chief Executive Officer · 2026-08-04 What replaced it is a different — and more defensible — narrative frame. The new top themes are yield per megawatt, agentic flow, complete platform, financial profile and Token volume. Management is no longer selling a router or an open-weight catalog; it is selling the arithmetic of a fleet.

it is not just about the megawatts we energize. It is about the yield we are getting from the fleet that we have, both on GPUs, but also CPUs.

Padmanabhan Srinivasan, Chief Executive Officer · 2026-11-03
The number attached is $18 million of ARR per average active megawatt across the entire fleet. That is a real progression — in May 2026, Paddy was defending a lower figure: “We definitely expect that we can increase that $13 million per ARR per megawatt over time.” — Padmanabhan Srinivasan, Chief Executive Officer · 2026-05-05 The claim now is that 90% of AI customer revenue comes from accounts that also consume core cloud primitives, and that customers on four or more products drive ~60% of total revenue. Whether or not you buy the framing, the metric moved in the right direction and management gave it a denominator nobody can fudge: total ARR over average active megawatts.

The Third Raise

The guidance cadence is the fundamental signal. Twenty-twenty-seven went from 30% to 50%-plus in February, to "60% or more" now. Matt put the compounding in plain terms: “Since our initial guidance in February, we have raised the midpoint of our 2026 guide by $120 million. On top of this, we've increased our 2027 growth rate outlook from 30% to 60% plus, an increase of over $0.5 billion at the midpoint.” — Matt Steinfort, Chief Financial Officer · 2026-11-03 And it is a direct answer to a question the company deferred in August. Then, Matt was explicit that he would not put a number on 2027: “it's premature to give a specific number. But clearly, the message is we're exiting the year a lot faster growth than what we had said we were... we expect there to be additional upside.” — Matt Steinfort, Chief Financial Officer · 2026-08-04 Three months later the upside is quantified. That is a management team converting a promissory note into a number.

Funding Is the Subtext — and the Stock's Problem

The single largest change in capital structure this quarter was not disclosed in a segment table: a $1 billion at-the-market equity program, layered on top of a $785 million equipment lease facility signed in September, roughly $845 million of cash and securities, an undrawn $400 million revolver, and the planned retirement of $312 million of 2026 converts. Net leverage is 1.2x. Paddy's framing is openly defensive:“As concern grows about funding AI growth, we sit with minimal leverage and a flexible balance sheet.” — Padmanabhan Srinivasan, Chief Executive Officer · 2026-11-03

The very proactive and opportunistic putting in place of an ATM today just gives us option value... We believe that we have the ability to fund all of our growth in 2027.

Matt Steinfort, Chief Financial Officer · 2026-11-03
Note the new keyword option value — it did not exist in this company's vocabulary a quarter ago. Neither did CapEx per megawatt, which Matt now volunteers at "mid-30s" for a GB300-class build and "closer to 40" for Vera Rubin, alongside a commitment that underwriting thresholds stay constant. That is the right answer to the right question. It is also a benchmark the market can now hold them to quarterly.

Why the Tape Is Short the Megawatts

The drawdown is not idiosyncratic. The entire capacity complex has been repriced: the global tape shows data center capacity down 32% over 90 days with DOCN sitting in the same eight-name basket as Equinix, Applied Digital and IREN; "AI data center" is off more than 20% over 30 days across a 55-ticker swath with negative breadth. High bandwidth memory, co-packaged optics, critical IT load — every keyword that denominates this build-out is falling together. Against that, the reported fundamentals still describe a premium asset. Price to revenue of 16.3x is up 422% year over year; price to free cash flow is 165.8x. Yet the operating leverage story is already bending: gross margin of 55.0% is down 4.9pp year over year as leased capacity and depreciation land ahead of revenue, and effective net cash sits at -$170 million. That is precisely the mechanism investors fear — Memphis came online weeks early, but the lease expense arrives before the ramp does.

The One Thing Worth Watching

The bull case now rests on a lever no competitor can copy quickly: agents as customers, not just workloads. Roughly half of unique visitors to digitalocean.com are agents, and more than 37,000 unique customers deployed through one this year. That connects to a genuinely broad wave — personal agents showed up across Booking, Shopify, PayPal and DoorDash this reporting season — but DigitalOcean's angle, agent-initiated provisioning, is company-unique. Capacity tells you the ceiling: ~75 MW live, another ~75 MW in 2027, ~30 MW in 2028, roughly 180 MW committed, all with Tier 1 providers. The business is accelerating, the guidance is compounding, and the stock trades as if the megawatts are a liability rather than a yield. On this evidence, the disagreement is about timing and terminal multiples — not about whether the tokens are flowing.