OVHcloud's Second Act: A Turnaround It Won't Let You Model

Growth re-accelerated to 10.5%, free cash flow turned positive for the first time since IPO — and management quietly deleted the one guide it couldn't control.

OVH.PA · Earnings Call · 2026-10-20

The Numbers Are Not Ambiguous

For two years OVHcloud was the European sovereign-cloud story that couldn't convert its industrial assets into a growth curve. This quarter that changed, and the change is measurable rather than rhetorical. Revenue reached €1,145 million, up 7.2% like-for-like and above the 5–7% guide; adjusted EBITDA margin hit 42%, up 160 basis points, with more than 70% of incremental revenue flowing through; net income was €9.4 million against €0.4 million. The shape of the year is the real headline.

Growth bottomed out in the second quarter at 5.1%. It then picked up to 6.9% in Q3 and reached 10.5% in Q4. This is our strongest quarter in almost 2 years and importantly, already at the pace of our FY '27 guidance.
Julien Rabette, Interim CFO · 2026-10-20

That Q4 run-rate is not a fluke of comparatives. Founder Octave Klaba returned as CEO, rebuilt the go-to-market into three segments — digital starters, scalers, corporates — and the customer engine restarted: outside the U.S. the base went from +0.7% month-over-month at the start of the year to +2% per month by August, with Klaba claiming more than 100,000 customers added in a single month. All three segments are now growing, corporates inflected to +7.8% in Q4 after three weak quarters, and Public Cloud is back above 20% (€255 million, +24.3% in Q4) helped by the 3AZ region in Paris and a marquee European Space Agency contract. The strategic asset base underneath it is enormous and underexploited: 250 megawatts of owned data center capacity, roughly 140 connected and ~110 still available, at a build cost of €15–20 million per megawatt. Most importantly for the equity story, levered free cash flow was +€24.7 million versus −€27 million a year earlier — “the first time since our IPO 5 years ago.” — Julien Rabette, Interim CFO · 2026-10-20 Leverage fell to 2.7x from 2.9x and return on capital employed rose to 7.7% from 7.1%, more than double the 3.3% of two years ago.

What Got Deleted From the Guidance

The tell is not in the results — it's in the guidance architecture. OVH went from four guided components to five, and the one it removed was CapEx. In its place: a commitment that ROCE grows to more than 9%, and that the leverage ratio holds at 2.7x. Klaba was disarmingly candid about the motive. “And why we removed CapEx because it was -- there was a few things that we didn't like about the CapEx... it means nothing. It's not -- it means a lot of things, but it's moved over the year.” — Octave Klaba, Founder, Chairman and CEO · 2026-10-20 He's right that the prior guide had become a moving target — 30–32%, then 33–35%, then landing at 32.6% ex-anticipation. But the fix reveals the underlying problem: CapEx is now the swing variable in an AI investment decision management refuses to pre-commit to, because demand is unproven. Rather than guide it, they've substituted two output ratios that implicitly cap it. That is CapEx discipline as a governance statement rather than a number — defensible, but it hands analysts a black box where they previously had a dial. There is a second, sharper disclosure retreat. Klaba said the company will deliberately share less: “We realized last year that a lot of this information were used by our competition... we gave too much and they use our secret sauce to go after us.” — Octave Klaba, Founder, Chairman and CEO · 2026-10-20 That is a legitimate competitive argument and also a convenient one. It is the reason given for not decomposing the "more than 10%" FY27 growth floor, and it left Bernstein's Derric Marcon asking three separate times what scenario was actually embedded in the guide. He never got an answer. The guide is a floor with an unknown mix of volume, repricing, and AI revenue.

AI: Five Bets Instead of One

OVH's AI answer is unusually honest about the commodity risk. Klaba explicitly framed GPU resale as a weak business — “reselling the GPUs, it's really competitive activity because you buy the same hardware, you deploy that in a little bit different data centers” — Octave Klaba, Founder, Chairman and CEO · 2026-10-20 — and then built a portfolio to de-risk it: GPUs, token-as-a-service (aided by the Gladia speech-to-text acquisition), Batch for AI, AI-for-SaaS, and the OVHai Workspace agentic platform going freemium at the November summit before GA in January/February. His argument for why now is refreshingly specific: each GPU generation cuts the cost of equivalent performance dramatically, so waiting was rational, and the B300/R200 class changes the economics. This is where the company's trajectory collides with the market's. The AI data center theme is one of the most powerful 360-day movers in the global tape — up over 50% across 40 advancing tickers — but over the last 90 days it has flipped to a decliner, with 47 names negative versus 6 positive, alongside data centers showing up in regional bank reports (FNB, KEY, AUB) as a credit and buildout theme rather than a growth one. OVH is arriving at the AI capex party with the tape already sceptical of the category. Tempering that: the underlying business is not AI-hyped — bare metal repositioning and a near-doubling of domain name volumes (AI coding tools making every project start with a domain) are prosaic, cash-generative volume drivers.

The Churn Question Nobody Answered

The most under-appreciated risk in this print sits in a single sentence: “the real increase of prices we had in the beginning of this month, one part, and there will be still in November, still in December...” — Octave Klaba, Founder, Chairman and CEO · 2026-10-20 The €25 million of free cash flow and the 160 basis points of margin were delivered partly on a price rise whose churn effect is, by management's own admission, unknowable for another quarter or two. Klaba hedged by noting OVH's public cloud is still roughly half of AWS pricing and that Azure raised European prices 11–17% weeks ago — a genuine competitive cover. But an unquantified price lever inside a floor-based guide is exactly the kind of thing that makes a "more than 10%" statement less reassuring than it sounds. So: a real turnaround, real cash, a genuinely differentiated North Star-style capital framework, and a management team choosing opacity at the precise moment investors wanted a bridge. The second act is credible; the reporting of it just got harder to audit.