Velaura AI Raises $110M As Data-Centre Power Becomes The Next Chip Battle

Velaura AI has raised $110 million at a valuation above $1 billion, and the funding round points to where the AI infrastructure race is moving next; power efficiency.

The company said in a Business Wire announcement that the Series A round will help advance ultra-low-power AI compute infrastructure. Reuters-linked reports say the round was led by Seligman Ventures, with participation from Capricorn Investment Group, Samsung Catalyst Fund, StepStone Group and Maverick Silicon.

The headline number is useful, but the more important point is the problem Velaura is trying to solve. AI data centres are no longer constrained only by access to GPUs. They are increasingly constrained by power, cooling, operating cost and the ability to keep large compute clusters running economically.

That is why investors are backing chip-design companies that promise lower power consumption. If a startup can reduce the energy required for AI workloads without sacrificing performance, it can affect both cost and deployment speed. In a market where power connections can delay data-centre projects, efficiency becomes strategic.

The timing fits the broader AI infrastructure story. Microsoft, OpenAI, Google, Meta, Amazon and xAI are all trying to secure compute, but the physical world is pushing back. Chips need buildings, cooling, grid capacity and financing. A better chip architecture can only solve part of that problem, but power efficiency is one of the parts that matters most.

We recently wrote about Microsoft’s AI buildout running into the hard math of chips and power. Velaura belongs in that same conversation. The next infrastructure winners may not only be the firms buying the most Nvidia GPUs. They may also be the companies that make each watt of AI compute do more work.

This is also why AI chip startups are still attracting money despite Nvidia’s dominance. Nvidia owns the centre of the market, but not every workload or data-centre problem will be solved by one kind of accelerator. Startups can compete around efficiency, inference, specialized workloads, networking, memory or physical AI systems.

Velaura’s pitch appears especially relevant for data centres and physical AI systems, where operating cost matters over long periods. Training large models gets the headlines, but inference at scale is the bill many companies will live with every day. Lower-power hardware can make that bill easier to manage.

There is execution risk. Chip startups are difficult businesses. They require deep engineering, manufacturing partnerships, customer validation and enough software support to make hardware useful. A valuation above $1 billion does not guarantee commercial adoption.

Still, the funding round captures a real market shift. AI infrastructure is no longer only about who can buy the most expensive chips. It is about who can deploy compute sustainably, affordably and at scale. If power is becoming the bottleneck, then chip efficiency is becoming one of the most important battles in AI.