Verda's $189M Raise: Europe Builds a Profitable AI Cloud

Finland's Verda raised $189 million this week, and the round is not the interesting part. The interesting part is the $165 million annualised revenue run-rate the company reported for July — reported alongside positive operating cash flow — while most American GPU clouds are still buying growth with borrowed money.

The company does not train frontier models. It rents the machines that do, the way a utility sells power.

Why a Finnish cloud can undercut the hyperscalers

Most companies called an "AI cloud" are resellers. They lease capacity from a hyperscaler or a colocation operator, mark it up, and call the bundle a platform. It owns two data centres, its GPU fleet, and the serving software in between.

Vertical integration is the whole margin story. Nordic electricity is cheap. Cooling is cheap. The power is contractually renewable, which matters to European enterprise buyers who now have a compliance line item for the carbon footprint of compute. None of that makes the company faster than a hyperscaler. It makes it cheaper to run.

The company crossed a $100M run-rate in June and says operations generate positive operating cash flow, a position almost unheard of for a capital-intensive infrastructure startup at this stage. Growth is being funded by customers as much as by investors.

A profitable cloud in a cash-burning market

The AI-infrastructure boom has been financed on the assumption that scale comes first and margins come later. Verda is a deliberate counter-example, and the comparison with a typical American neocloud is where the strategy becomes legible.

Company disclosures and press reports, 2026

As we wrote in July, the spread between bare-metal GPU pricing and hyperscaler list pricing has been the entire commercial argument for specialists like Verda. The new part is that the argument now survives contact with a balance sheet.

AI is becoming critical infrastructure, and our goal is to build the platform frontier AI teams can depend on as they scale.— Ruben Bryon, Founder and CEO, Verda

That claim is easy to make and hard to hold. Renting GPUs is a commodity business the moment capacity is abundant. The question investors are underwriting is whether owning the stack keeps its unit economics — the profit or loss on a single rented GPU-hour — intact when the next wave of capacity arrives.

The power bill is the new P&L line

Compute is, increasingly, a power trade wearing a software interface. For a cloud that rents GPUs by the hour, gross margin is set by three variables: the price of a megawatt-hour, the utilisation of the fleet, and the depreciation schedule on the accelerators. Two of those three are structurally favourable in the Nordics.

Cheap hydro and wind do not only lower the bill. They lower the volatility of the bill, which is what long-dated enterprise contracts actually price. A US operator signing a 2028 delivery slot is underwriting gas prices, transmission constraints and a queue of other data centres. It is underwriting weather it already knows.

Utilisation is the quieter risk. A GPU that sits idle still depreciates, and idle capacity is exactly what a fast build-out produces. A claim to positive operating cash flow implies the fleet runs hot today. The test is whether that holds through a construction cycle, when new capacity arrives before the demand that justifies it.

What the $189 million actually buys

The round was led by Emergence Capital, with MUFG Innovation Partners, Supermicro, Varma Mutual Pension Insurance and Lifeline Ventures participating. The capital is earmarked for compute capacity, product development, and expansion across Europe, the United States, the United Kingdom and Asia.

Verda already runs two Finnish data centres and has signalled a move into Sweden. It became an NVIDIA Preferred Partner in February 2026, which grants early access to new accelerator architectures, and it has a separate full-stack infrastructure partnership with Supermicro.

Bryon has told Bloomberg the company aims to raise as much as $1.5 billion in equity and debt this year, and up to $10 billion by 2027. Those are ambition figures, not commitments. But they sketch the intended shape of the business: a European hyperscaler, not a boutique GPU broker.

The risks: scale, valuation, and a crowded field

Verda declined to disclose an exact post-money valuation, confirming only that it exceeds $1 billion. In a market where fellow neoclouds have raised at far higher marks, the modesty is either discipline or a tell.

Can a profitable European challenger hold its pricing advantage through 2027?

Development scenarios