Velaura AI Raises $110M Series A Past a $1B Valuation — With a Bitcoin Mining Past

What limits AI now is not demand for compute but the electricity to run it. Hyperscalers are committing hundreds of billions to AI data centers while power availability, not silicon, sets the deployment calendar. For machines that operate in the physical world — robots, drones, autonomous systems — the constraint is tighter still: compute has to happen inside a battery and a thermal envelope. The competition has shifted from raw performance to performance per watt.

Velaura AI, an ultra-low-power AI compute company based in Santa Clara, announced on August 18 a $110 million Series A that takes its valuation past $1 billion.

The company used to make bitcoin mining chips

The “Series A” label is misleading. Velaura’s predecessor was Auradine, founded in 2022 to build Teraflux bitcoin miners — air-, hydro- and immersion-cooled systems running an in-house 3nm ASIC, aimed at a market dominated by Asian manufacturers like Bitmain and MicroBT. In April last year it raised a $153 million Series C led by StepStone Group, taking total funding past $300 million.

Then in March, the company renamed itself Velaura AI and launched Titan Core, redirecting the power-efficiency work from mining silicon into AI accelerators. Resetting the series letter to A tracks that reset. The investor list is the tell: Auradine backers Mayfield, Maverick Silicon, Premji Invest, Samsung Catalyst Fund and StepStone Group all returned, as did MARA Holdings — the publicly listed bitcoin miner that had been a customer. The business changed direction; the capital base carried over.

Seligman Ventures led this round, with Capricorn Investment Group joining as a new investor.

30 million ASICs, not 30,000

Titan Core is not a finished chip but digital chip IP and a design platform. It delivers a 2-4x improvement in performance per watt on the mathematical operations inside AI accelerators while holding performance steady, according to the company.

Its strongest claim is history rather than benchmarks. The underlying technology has been validated at commercial scale and shipped in more than 30 million ASICs on leading process nodes, with what the company describes as world-class yield and reliability. Selling mining rigs produced exactly the credential that matters here. In a field where most AI silicon startups are making performance claims from prototypes, having proven volume yield is an unusual position.

Velaura is extending that architecture in two directions: data center AI accelerators, and Physical AI — intelligent robots, drones and other embodied systems.

“Every advance in AI, from reasoning models to embodied intelligence, creates demand for more compute, and ultimately more power,” said co-founder and CEO Rajiv Khemani. “The next era of AI will be defined not only by better models, but also by fundamentally better compute economics.”

Two exits, and another go

The founders explain much of this round. Khemani co-founded Innovium in 2015, building the TERALYNX family of high-bandwidth Ethernet switching ASICs for hyperscale data centers, and sold it to Marvell for roughly $1.1 billion in 2021. Before that he was COO at Cavium. Auradine, now Velaura, came in 2022.

Co-founder and Chief Development Officer Manu Gulati is a silicon architect whose career spans Apple, Google and Qualcomm, and who co-founded NUVIA — acquired by Qualcomm for about $1.5 billion. The wider team draws from Apple, NVIDIA, Google, Qualcomm and Marvell.

“At Mayfield, we invest in people first. This is our fourth partnership with Rajiv Khemani and our second with Manu Gulati,” said Navin Chaddha, managing partner at Mayfield, noting involvement since inception. Umesh Padval, managing partner at Seligman Ventures, called Physical AI “one of the next major frontiers for AI” requiring “a fundamentally different approach to compute centered on extreme power efficiency,” and said Velaura is the firm’s first Physical AI investment.

Dipender Saluja, managing partner of Capricorn’s Technology Impact Funds, said AI’s energy footprint is “on track to become one of the most significant infrastructure challenges of this decade,” adding that Velaura “is attacking that problem at its root — the silicon itself — with technology that has shipped at scale.”

The capital funds development and commercialization of the compute portfolio, engineering and customer-facing hiring, and deeper work with strategic partners. The company says it has multiple hyperscaler engagements underway.

The competition sits at a different layer

AI chip startups have drawn $8.3 billion so far this year, but Velaura occupies a different layer from most of them.

The heat is in finished accelerators. Etched, which builds ASICs specialized for transformer inference, raised $700 million this month at a $21 billion valuation led by Jane Street — double its $10.3 billion mark from July — and says it has booked $1 billion in contract orders. d-Matrix raised $275 million at a $2 billion valuation on digital in-memory compute and entered volume production in June. EnCharge AI claims up to 20x less energy than leading AI chips using analog in-memory compute.

There are cautionary cases too. Groq peaked at $6.9 billion last September, then raised $350 million this month at $3.5 billion after Nvidia struck a licensing deal and hired away much of its talent — a reminder that how these startups entangle with big tech determines outcomes.

Velaura avoids that collision. Rather than selling finished chips against Nvidia, it sells the IP that goes inside accelerators — a position that can make the same companies competitors or customers. On the Physical AI side it overlaps with edge inference players like Hailo and Axelera AI.