Everything About AI
NextFin News - Verda, the Helsinki-based AI cloud infrastructure company formerly known as DataCrunch, has raised $189 million (€163 million) in an oversubscribed Series B round led by Emergence Capital, a move that lifts the company into unicorn territory and takes its total funding to more than $450 million. The round, announced September 22, 2026, arrives as the Finnish startup reports a $165 million annualized revenue run rate reached in July 2026 — a near-tripling from roughly $60 million at the start of the year — and positions Verda as Europe's latest challenger to the US-dominated specialized GPU cloud market.
Verda is a private company, so there is no stock-price reaction to measure; the round itself — oversubscribed, with a new lead investor replacing the previous round's lead — is the market signal, and it reprices the company above the $1 billion threshold that defines unicorn status.
A $189 Million Bet on a European Compute Challenger
The round's investor mix is itself a statement. Alongside lead investor Emergence Capital — the Silicon Valley firm behind enterprise-cloud names such as Salesforce, Veeva, Zoom and Box — the round included MUFG Innovation Partners, Supermicro, Varma Mutual Pension Insurance Company, Lifeline Ventures, 6 Degrees Capital (6DC), byFounders, Tesi (Finnish Industry Investment Ltd), and a group of angel investors including Ola Tørudbakken and Mark Saroufim. That is a coalition of US enterprise-software expertise, Japanese financial-institution capital, a US server hardware maker with direct exposure to NVIDIA supply chains, and the Nordic institutional investors that have backed Verda since its DataCrunch days.
Verda frames the raise as a closing window. In a statement accompanying the announcement, founder and CEO Ruben Bryon said:
There's a window right now to build one of the defining compute companies of this generation, and to do so from Europe. It won't be open for long.
The timing is not accidental. Verda has now closed three funding events in less than six months: $117 million in April 2026 led by Lifeline Ventures; an extension to $155 million in July when the Nordic Investment Bank joined a lender group that already included Nordea, OP, Danske and Norion; and now this $189 million equity round. The company says the capital will go toward multiplying compute capacity — more data-center capacity coming online — plus deeper investment in its platform, its in-house AI Lab, and the features customers already use.
The traction behind the check is the real story. Verda says it reached a $165 million annualized revenue run rate in July 2026, up from a $100 million run rate in the first half of the year and from just over $60 million in the first quarter of 2026. It says it is cash-flow positive, employs more than 250 people across Helsinki, London, Taipei and San Francisco, and powers AI workloads for organizations across more than 50 countries. Named customers include Aleph Alpha, which runs research infrastructure on Verda clusters with Verda engineers embedded on the software stack; Magnific, serving media generation at millions of requests per day; and Epsilon Health, training radiology AI models at native image resolution on a dedicated Verda cluster.
The Run-Rate Math, and What It Leaves Out
A $165 million run rate off more than $450 million of total capital raised is a capital intensity worth naming: Verda has raised roughly $2.70 for every dollar of annualized revenue. That is not an indictment — it is the price of admission in a category where capacity must be built before it can be sold. The relevant question is not whether the ratio is high, but whether deployed GPUs convert into sustained utilization fast enough that the next dollar of revenue requires less than the last. The Series B is the test of that curve.
The company's funding history shows the escalation clearly. It closed $13 million in seed funding in 2024, raised $64 million in Series A in 2025 — the round in which DataCrunch rebranded to Verda and earned NVIDIA Preferred Partner status — and has now added $344 million of equity and debt in 2026 alone. A startup that was cash-flow positive within weeks of its 2020 founding, launched by Bryon from a garage in Helsinki alongside co-founders Miłosz and Tamir, is now competing for the same institutional dollars as the US infrastructure specialists.
That escalation mirrors the broader European market. Finland's top-funded tech companies in the first half of 2026 included Verda's $117 million round, and the second half has brought a wave of large European infrastructure and deep-tech deals: cybersecurity startup Exein raised $270 million at a $1.7 billion valuation, satellite-intelligence company Open Cosmos secured €300 million, and rocket developer the Exploration Company closed a $450 million Series C. The $189 million round is Verda's entry into that cohort — and its claim on a slice of capital that, until recently, flowed almost exclusively to US founders.
Why Specialized AI Cloud Is a Structural Shift, Not a Cycle
The surface read is simple: AI demand is hot, so GPU cloud companies raise money. That is true and insufficient. The deeper question is whether the specialized AI cloud layer — companies like Verda, CoreWeave and Lambda Labs — is a durable structural tier of the computing stack, or a cyclical arbitrage that disappears once NVIDIA supply normalizes and the hyperscalers refocus on the segment.
Three pieces of evidence point to structure.
First, demand is fragmenting by workload. Training runs, large-scale inference, fine-tuning and rapid prototyping have different economics: burst versus sustained usage, latency sensitivity, interconnect requirements, data-sovereignty constraints. A general-purpose hyperscaler console is optimized for breadth, not for the cost-per-training-job or time-to-cluster economics that an AI-native team optimizes. Verda's pitch — vertical integration from data-center design through the platform and developer tooling — is a bet that workload-specific optimization beats one-size-fits-all capacity. That is a structural feature of how AI compute is consumed, not a temporary shortage.
The transmission channel matters. When a hyperscaler sells GPU time, it is selling a commodity input priced against a global hardware market. When a specialist sells a training cluster, it is selling an outcome — time-to-model, utilization efficiency, engineering support — priced against the customer's research velocity. Those are different products with different price elasticities. As long as AI teams optimize for research velocity rather than raw GPU-hour cost, the specialist layer has a defensible margin pocket. The moment procurement departments standardize on price-per-hour across all workloads, that pocket closes.
Second, the capital intensity of the category creates a moat that favors players who can reach scale. Competing in AI infrastructure means spending billions upfront on data centers and GPUs before revenue catches up. IDC projects global AI infrastructure spending will reach $487 billion in 2026 and exceed $1 trillion by 2029, and the leading US pure-play providers are already operating at that scale: CoreWeave reported first-quarter 2026 revenue of $2.078 billion, more than double the $982 million it posted a year earlier, alongside a record revenue backlog with hyperscalers and large enterprises. The gap between what it costs to enter and what it takes to win is widening, not narrowing. Verda's ability to raise $189 million after already being cash-flow positive is a signal that it has crossed the credibility threshold with institutional capital.
Third, sovereignty is a permanent constraint, not a mood. European enterprises in health, finance and the public sector face data-residency and regulatory requirements that make a Helsinki-based, EU-jurisdiction provider a structurally different choice from a US hyperscaler, regardless of price. Verda's Finnish data centers run on 100% renewable energy, and the company markets itself on digital sovereignty as much as on performance. As the EU's AI regulatory framework hardens, that positioning shifts from a differentiator to a procurement prerequisite for a whole class of customers.
The Counter-Thesis: Commoditization Is Real, and It Is the Real Risk
The strongest case against Verda's thesis is not that AI demand cools. It is that GPU cloud becomes a commodity, and the specialists get squeezed between hyperscaler price wars on one side and NVIDIA's own increasing willingness to allocate capacity directly on the other.
This is not a strawman. The hyperscalers have deeper balance sheets, global footprints and existing enterprise relationships. If NVIDIA H200 and B200 supply normalizes — and lead times compress from quarters to weeks — the scarcity premium that lets specialists charge above list price evaporates. At that point the market competes on price per GPU-hour, and the player with the cheapest cost of capital wins. Verda's Nordic renewable-power advantage lowers operating cost, but it does not erase the scale gap with a Microsoft or Amazon.
There is also an execution risk specific to Verda's geography. A Helsinki-based team serving US enterprise customers must match the proximity and response speed of US-based rivals on the ground. The company has opened a San Francisco office and hired a CFO with Morgan Stanley and Nokia corporate-development experience — Mikko Einiö, appointed in August 2026 — precisely because it knows the next phase is a capital-markets and enterprise-sales game as much as an engineering one.
The counter-thesis, stated fairly: Verda is a well-run company in a hot market, but the specialized GPU cloud layer may be a transitional arbitrage rather than a permanent tier, and the winners will be determined by cost of capital and proximity to customers — two advantages that still lean toward the US incumbents.
The Falsifying Signal
The judgment that specialized AI cloud is structural rests on one observable condition: that workload-specific economics continue to beat hyperscaler general-purpose pricing for a meaningful share of AI workloads. The specific falsifying signal is this — if, over the next four quarters, hyperscaler pricing for comparable GPU instances falls by more than 30% while NVIDIA lead times normalize to under four weeks, and specialized providers' gross margins compress below 25%, the "durable tier" thesis is wrong and this is a cyclical arbitrage after all. Watch pricing disclosures and margin commentary from the public pure-plays over the next two earnings cycles; that is where the verdict will appear first.
What the $189 Million Changes — and What It Does Not
The immediate effect of this round is capacity, plain and simple. More data centers coming online means Verda can accept workloads it would otherwise have to turn away in a supply-constrained market. The second-order effect is credibility: unicorn status and a $165 million run rate move Verda from "promising challenger" to a vendor that appears on enterprise shortlists alongside the US specialists. The third-order effect is the signal it sends to European founders and investors — that a compute company can be built from Helsinki, not just consumed there.
Who benefits: European enterprises with data-sovereignty requirements; AI teams that need burst capacity without hyperscaler procurement friction; and the Nordic deep-tech ecosystem, which now has a flagship infrastructure name. Who is exposed: smaller GPU resellers with no vertical integration and no sovereign angle — the middle layer that neither matches hyperscaler scale nor offers what Verda offers.
Split by horizon:
- Short term (next 6–12 months): the story is capacity deployment and whether Verda can convert the oversubscribed round into deployed GPUs fast enough to keep pace with its run-rate growth. Execution here is the metric.
- Medium term (12–24 months): the story is margin defense. Can Verda hold pricing power as supply normalizes, or does it get pulled into the commodity lane? This is where the falsifying signal above becomes live.
- Long term (structural): the story is whether Europe produces a lasting independent compute tier. Verda's bet is that sovereignty, renewable cost advantage and workload-specific engineering compound into a defensible position.
Base case: Verda deploys the capital, holds its niche as Europe's sovereign AI cloud provider, and grows into its valuation on the back of enterprise contracts. Upside case: GPU scarcity persists longer than expected, pricing power holds, and Verda becomes the default European alternative to US specialists. Downside case: supply normalizes quickly, hyperscalers cut prices, and Verda's margins compress; the round then buys time rather than a moat.
Verda's $189 million is less a bet on AI demand — that trade is already crowded — than a bet that compute has a political geography, and that Europe will pay a premium to keep its AI infrastructure on European soil. That premium is real today. Whether it survives the next hardware cycle is the question this round is meant to answer.
Explore more exclusive insights at nextfin.ai.
