Capital

Multiverse Raises $570M to Shrink AI Models

Multiverse Computing raised up to $570 million at a $1.7 billion valuation, co-led by Forgepoint Capital International, BNP Paribas SIVF and Bullhound Capital, to scale CompactifAI, its tensor-network technique for shrinking AI models by 80 to 95 percent. What a Basque-government-backed European compression play means for owners planning 2026-2027 AI infrastructure budgets.

CapitalBy Servola Tech Desk2026-08-135 min read

AI-assisted, edited by humans. | Editor-in-Chief: Leon Soliman | Editorial standards

Multiverse Raises $570M to Shrink AI Models

Key takeaways

Multiverse Computing, a quantum-inspired AI-compression startup based in San Sebastian, Spain, announced on July 27, 2026 a Series C round targeting up to $570 million (roughly 500 million euros) at a pre-money valuation of $1.7 billion (roughly 1.5 billion euros), co-led by Forgepoint Capital International, BNP Paribas SIVF and Bullhound Capital. The round brings Multiverse's total funding to roughly $800 million, an almost five-times step-up from the $215 million Series B it closed in June 2025, and the company says the round may remain open to additional select strategic investors.
The investor list stretches well beyond the three co-leads: Santander Alternative Investments, HP, Tikehau Capital, Orange Ventures, NAventures, Scania Invest, Qatar Development Bank, SETT, Zouk Capital, the Basque Government's Hazten Scale-Up Fund, the European Innovation Council's EIC Fund and Kutxa Fundazioa all took part, a mix of corporate strategics, sovereign and quasi-sovereign capital, and EU public funding rarely lined up together behind a single AI infrastructure round.
Multiverse's core product, CompactifAI, applies tensor networks, a mathematical toolset borrowed from quantum physics, to compress existing large AI models by 80 to 95 percent while preserving most of their accuracy and sharply cutting the energy consumption and inference latency of running them. The pitch is efficient AI from edge to cloud: a compressed model can run on far cheaper, smaller hardware, including edge devices, instead of requiring large-scale hyperscaler GPU clusters.
The round was independently reported by SiliconANGLE and The Quantum Insider alongside the company's own GlobeNewswire announcement, with all three confirming the same investor list and figures, and it lands in the middle of a 2026 news cycle otherwise dominated by US-based neocloud GPU-leasing deals and optical-interconnect scarcity stories, a contrast Multiverse's backers are explicitly betting on.
For any owner scoping a 2026-2027 AI deployment budget, Multiverse's near-$800 million funding stack is now a concrete reason to ask how much a given workload can be compressed before it actually needs perpetual, scarce and increasingly financialized hyperscaler GPU capacity at all, a question that used to be a research curiosity and is now a line item worth pricing before signing a multi-year cloud AI contract.

The round: $570 million at a $1.7 billion valuation, and who is behind it

Multiverse Computing announced on July 27, 2026 a Series C fundraising targeting up to $570 million, roughly 500 million euros, at a pre-money valuation of $1.7 billion, roughly 1.5 billion euros. The round, published via the company's own GlobeNewswire announcement and independently confirmed by Yahoo Finance, SiliconANGLE and The Quantum Insider, was co-led by Forgepoint Capital International, BNP Paribas SIVF and Bullhound Capital. Participating investors include Santander Alternative Investments, HP, Tikehau Capital, Orange Ventures, NAventures, Scania Invest, Qatar Development Bank, SETT, Zouk Capital, the Basque Government's Hazten Scale-Up Fund, the EIC Fund and Kutxa Fundazioa, and the company says the round may stay open to additional select strategic investors.

The raise brings Multiverse's total funding to roughly $800 million, an almost five-times step-up from the $215 million Series B it closed in June 2025. That pace, a pre-money valuation nearing $1.7 billion little more than a year after a $215 million round, is itself the story: it signals that institutional and sovereign-adjacent capital, not just specialist deep-tech venture funds, now treats AI model compression as a category worth underwriting at nine-figure scale.

CompactifAI: what tensor-network compression actually does

Multiverse's core product is CompactifAI, a technology that applies tensor networks, a mathematical toolset originally developed for quantum physics, to compress the size of existing large AI models by 80 to 95 percent while preserving most of their accuracy. The practical effect is a sharp reduction in the energy consumption and inference latency of running those models, since a compressed model needs far less compute to answer the same query.

The company frames this as efficient AI from edge to cloud: a model shrunk with CompactifAI can run on far cheaper, smaller hardware, including edge devices, instead of requiring the large-scale hyperscaler GPU clusters that most frontier and near-frontier models currently depend on. That is a claim about where AI can run, not just how well it runs, and it is the claim the entire round is underwriting.

The counter-bet to the AI-infrastructure chokepoint economy

Most of the capital pouring into AI infrastructure in 2026 is chasing the opposite strategy: securing access to a chokepoint. Optical interconnect components are scarce, neocloud operators are leasing GPU capacity on leveraged balance sheets, and hyperscalers are locked into a capex arms race to build ever-larger clusters, all of it premised on the idea that running capable AI requires ever more of that scarce, expensive infrastructure. Multiverse's bet is structurally different, and it is the genuine analytical hook of this round: rather than trying to out-build the chokepoint, CompactifAI tries to make the chokepoint irrelevant to a large share of AI workloads.

If a technique like CompactifAI can reliably shrink a capable model by 80 to 95 percent with minimal accuracy loss, an owner no longer needs perpetual access to scarce, increasingly financialized hyperscaler GPU capacity to run useful AI in production, they can deploy a compressed model on owned or far cheaper commodity and edge hardware instead. That is a direct, structural threat to the vendor-lock-in and recurring-cloud-spend model that most current AI infrastructure investment assumes will persist indefinitely.

A Basque-backed European bet against hyperscaler dependence

There is also a genuine European strategic-autonomy angle worth stating plainly. Multiverse is headquartered in San Sebastian, in Spain's Basque Country, and its Series C carries direct backing from the Basque Government's Hazten Scale-Up Fund, the European Innovation Council's EIC Fund and Kutxa Fundazioa alongside private and corporate capital. This is a European company built with explicit public-institution support, not a spinout from a US hyperscaler or a Silicon Valley deep-tech fund.

That matters because it credibly challenges an assumption baked into most 2026 AI infrastructure investment, including the US-based neocloud and optical-interconnect deals dominating the same news cycle: that frontier-capable AI has to run on US hyperscaler-scale infrastructure at all. A near-$1.7 billion valuation for a European compression company backed by regional and EU public capital is a direct counterweight to that assumption, not a footnote to it.

What this means if you are scoping a 2026-2027 AI deployment budget

For an owner planning AI infrastructure spend over the next 18 months, Multiverse's round turns a research curiosity into a budget-line question: how much can this specific workload be compressed before it actually needs cloud GPUs at all. That question now has a nearly $800 million answer behind it in the form of a funded, independently verified vendor, which changes the calculus before signing a multi-year cloud AI contract that assumes perpetual hyperscaler dependence.

For hyperscalers, neocloud operators and the infrastructure vendors selling into the current chokepoint economy, Multiverse is a live signal that at least part of the market, and a well-capitalized part at that, is betting the chokepoint itself will not hold for every workload. Whether CompactifAI's compression claims prove out at production scale across a wide range of models is still to be tested, but the money behind the bet is no longer trivial.

Frequently asked questions

How much did Multiverse Computing raise, and at what valuation?

Multiverse Computing announced on July 27, 2026 a Series C round targeting up to $570 million, roughly 500 million euros, at a pre-money valuation of $1.7 billion, roughly 1.5 billion euros, co-led by Forgepoint Capital International, BNP Paribas SIVF and Bullhound Capital. Total funding to date is roughly $800 million.

What does CompactifAI actually do?

CompactifAI applies tensor networks, a mathematical toolset from quantum physics, to compress existing large AI models by 80 to 95 percent while preserving most of their accuracy, cutting the energy consumption and inference latency needed to run them on far cheaper hardware, including edge devices.

Who are the other investors in Multiverse's Series C, besides the three co-leads?

Participants include Santander Alternative Investments, HP, Tikehau Capital, Orange Ventures, NAventures, Scania Invest, Qatar Development Bank, SETT, Zouk Capital, the Basque Government's Hazten Scale-Up Fund, the European Innovation Council's EIC Fund and Kutxa Fundazioa, with the round possibly remaining open to additional select strategic investors.

Multiverse did not raise $570 million because AI compression is fashionable, it raised it because a European, publicly-backed challenger just put nearly $800 million behind the bet that the hyperscaler chokepoint is optional, not structural, and every owner still signing multi-year cloud GPU contracts now has a compression benchmark to explain why they did not check first.

Multiverse ComputingCompactifAISeries CSpainQuantum-Inspired AIModel Compression

Only 6% of Finance Leaders Want Full AI Autonomy

Moss raised EUR30m at a EUR1bn valuation, but the real news is a survey where 48% of finance leaders want AI control and just 6% want autonomy.

5 min read

Capital

$1.2 Billion for AI Agents That Finish the Call

HappyRobot raised a $150 million Series C at a $1.2 billion valuation, co-led by Prysm Capital and Eurazeo, with DHL and Uber among its 150-plus customers.

4 min read

Decisions

The Third Pivot Was the Warning Magic Leap Missed

Magic Leap filed a WARN notice for 193 layoffs effective 1 October 2026, its fourth strategic pivot after $4 billion raised. The lesson is the pattern, not the pivot.