Google and German energy company RWE invested €411 million ($469 million) in Proxima Fusion on July 7, valuing the nuclear fusion startup at €2.4 billion. The financing round is Google’s latest bet that fusion power can solve the energy equation for massive AI infrastructure.

Proxima Fusion is building stellarator-based nuclear fusion. The company wants a demonstrator running in the early 2030s and a commercial power plant in the latter half of the decade. That timeline is faster than most fusion companies are claiming.

Why Google Cares

AI training and inference require electricity. Lots of it. Data center power consumption is rising faster than grid capacity. Google signed an electricity purchase agreement with Commonwealth Fusion Systems in June 2025 and has also invested in that company. Now Proxima Fusion is a second bet on the same outcome: fusion power in the early 2030s.

If fusion works at commercial scale, Google secures long-term power. If it doesn’t, Google has lost investments but can still operate from conventional grids and renewables. It’s a hedge, but a serious one.

The Technology

Stellarators are a different approach to fusion than tokamaks. Instead of a doughnut-shaped plasma chamber, stellarators use a twisted, more complex geometry. They’re harder to build, but they have theoretical advantages in plasma stability. Proxima is betting on this path.

The company has demonstrated that its approach works in lab conditions. Now it’s raising money to scale up to a power plant. That jump—from lab to grid—has killed every fusion company in history. Proxima is aware of this.

Investor Confidence

XTX Ventures led the round. Google and RWE invested as strategic partners. Khosla Ventures returned. These are not passive investors. They’re betting capital on both technical success and commercial viability.

The valuation—€2.4 billion for a company three years old with no revenue—reflects belief that fusion is coming. Right now, belief is all the market has.

If Proxima gets power online in the 2030s, it changes AI economics forever. If it fails, it’s just another fusion casualty.