Crusoe Is Worth $30B: The AI Infrastructure Startups Raising Billions

The first half of September 2026 made it clear where venture capital is heading: AI infrastructure. Crusoe Energy raised more than $3 billion at a $30 billion valuation, Figure AI signed a $3.5 billion compute deal, and Dutch startup Euclyd raised €200 million with Samsung’s backing to challenge GPU economics. Three different bets, one bottleneck: power and chips.

Quick answer: who raised what?

Crusoe (AI data centers powered by its own energy) raised over $3 billion and tripled its valuation in about ten months, from $10 billion to $30 billion. Figure AI (humanoid robots) secured up to 100,000 Nvidia GPUs in a $3.5 billion agreement with Nscale. Euclyd (inference chips) closed a €200 million (~$230 million) Series A with Samsung participating, targeting commercial deployment in 2028.

Crusoe: from $10B to $30B in ten months

Crusoe’s round was co-led by Atreides Management and Valor Equity Partners. What supports the valuation is a five-year cloud agreement with Jane Street worth roughly $13 billion — contracted revenue before the data centers are even finished. The company’s model is to build compute capacity wherever energy is cheap and abundant, the input that has become AI’s physical limit.

Figure AI: 100,000 GPUs to train robots

Humanoid-robot maker Figure AI signed a $3.5 billion compute commitment with Nscale that can scale to 100,000 Nvidia GPUs and up to $6 billion in total. Deployment is targeted for the second half of 2027 in Barstow, Texas. It is a sign that training robots for the physical world demands the same scale of compute as large language models.

Euclyd: the challenge to GPU economics

The smallest of the three rounds may be the boldest. Rotterdam-based Euclyd received €200 million in Series A funding — with Samsung among the investors — to build a custom inference architecture that promises a much lower cost per token than conventional GPUs. Commercial deployment is not planned until 2028, but the mere fact that Samsung is backing a “GPU challenger” shows how uncomfortable the industry is with its dependence on Nvidia.

The rounds compared

Why this matters to you

For founders, the takeaway is that the application layer — the AI app — has become cheaper to build, while the infrastructure layer is where capital concentrates. That opens room for startups focused on niches (healthcare, agriculture, legal) that treat infrastructure as a commodity. For investors, the warning is the cycle: valuations that triple in ten months depend on giant contracts actually materializing. For consumers, the good news is that more compute supply tends to push AI service prices down over the coming years.

Frequently asked questions

What does Crusoe Energy do?

It builds and operates AI data centers in locations with abundant, cheap energy. In September 2026 it raised more than $3 billion at a $30 billion valuation.

How much is Figure AI worth today?

The company did not disclose a new valuation with this deal; the announcement was a $3.5 billion compute commitment with Nscale that could reach $6 billion.

Why did Samsung invest in Euclyd?

Euclyd is developing an inference architecture that promises lower cost than GPUs. For Samsung, it is a bet on reducing dependence on Nvidia.

At DigitalRadar, we follow where tech money is going. Stay on the radar so you do not miss the next round.