Velaura AI, formerly Bitcoin-mining chip firm Auradine, raised $110 million at a valuation above $1 billion to sell chip-design technology that cuts AI data-center power use. The raise reflects a real shift: as AI's electricity demand surges, the binding constraint is becoming power, not compute, and efficiency is becoming the product.

[For more news, click here] 

For years, the story around artificial intelligence infrastructure has been circling around chips, who can get the most graphics processors, and how fast. The constraint that decides the outcome? Electricity. A Santa Clara startup has just raised a large round on the premise that the next phase of AI build-out will be won on watts.

Velaura AI said on Tuesday it raised $110 million in a Series A funding round that valued the chip-designing startup at more than $1 billion, with investors backing technology that can lower power consumption and operating costs at AI data centers. The round was led by Seligman Ventures, with participation from new investor Capricorn Investment Group, alongside existing backers Samsung Catalyst Fund, StepStone Group, and Maverick Silicon.

Velaura, which was known as Auradine until it rebranded in March 2026, develops low-power chips and design software for data centers and for so-called physical AI applications such as robotics and autonomous systems. Its flagship product, a chip-design and intellectual-property platform called Titan Core, is pitched as a way to cut the power a customer's AI accelerator draws without a full chip redesign, and the company says the technology has already been deployed in more than 30 million chips. Velaura says Titan Core can roughly halve overall chip power for AI accelerators, which would translate to hundreds of watts saved on a typical high-end processor, though that figure is the company's own and has not been independently benchmarked.

"The next era of AI will be defined not only by better models, but also by fundamentally better compute economics," said Rajiv Khemani, co-founder and CEO of Velaura AI, in a statement.

Why Power Became the Constraint

In its landmark Energy and AI analysis, the International Energy Agency projected that electricity demand from data centers worldwide will more than double by 2030 to around 945 terawatt-hours, slightly more than the entire electricity consumption of Japan today, and that demand from AI-optimized data centers will more than quadruple over the same period. The agency also warned that grids are already straining under the load, estimating that around a fifth of planned data-center projects could face delays tied to power access.

The pressure is especially acute in the United States, where Velaura is based. The IEA has projected that American data centers will consume more electricity by 2030 than all of the country's energy-intensive manufacturing combined, including aluminum, steel, cement, and chemicals. When power becomes short in supply, the value of squeezing more computation out of every watt rises sharply. That is the market Velaura is aiming at, and its own framing is blunt: the company argues that power is becoming the limiting constraint for compute everywhere.

Getting Paid by the Watt

How Velaura plans to make money is through upfront fees. The startup charges an upfront fee for its technology, plus a royalty tied to a share of the power savings customers achieve, a structure Khemani confirmed is similar to Arm's per-chip licensing model before Arm began selling its own chips. Taking a cut of the electricity a customer no longer has to buy is an unusual arrangement, and a capital-light one, since Velaura licenses design tools and intellectual property rather than manufacturing finished chips. In principle that gives it a wider addressable market than a company betting on a single processor.

The founders have the credentials to warrant attention as well. Khemani previously ran Cavium, acquired by Marvell for roughly $6 billion, and co-founded Innovium, the Ethernet-switch designer Marvell bought for about $1.1 billion in 2021, and Velaura's ranks include veterans of Nvidia, Qualcomm, Apple, Google, and Marvell. There is also an unusual origin story. As Auradine, the company built energy-efficient silicon for Bitcoin mining, a business that is fundamentally a power-optimization contest, where every watt saved improves margin. Having proven low-power design in one of the most electricity-sensitive industries on the planet, the team is now applying the same techniques to AI. Notably, one of the largest publicly traded Bitcoin miners, MARA Holdings, remains an investor.

The Claims Worth Checking

While The power-savings figures are Velaura's own and await independent validation, the company says it is engaged with three of the four largest cloud providers as potential customers but declined to name them, so that traction cannot yet be verified from the outside. The more than 30 million chips it cites are largely a legacy of the Bitcoin-mining business rather than AI deployments, which are newer and smaller. And the round's "Series A" label is something of a reset, since as Auradine the company had already raised across earlier Series A, B and C rounds totaling more than $300 million, a detail that reflects the pivot more than a company at the true start of its funding life.

A long list of well-funded startups, including Groq, Cerebras, SambaNova, Tenstorrent, and Etched, is chasing different angles on efficient AI silicon, while Nvidia, Arm, and the hyperscalers designing their own chips all have power efficiency in their sights. Velaura's licensing model and its focus on cutting power rather than winning raw benchmarks are its attempt to carve out a distinct lane.

Why it Matters Beyond One Round

Strip away the specifics and Velaura's raise is a marker of where the AI infrastructure contest is heading. If the growth of AI is increasingly capped by the availability of electricity and the strain on grids, then performance per watt, not peak performance, starts to decide which systems get built and where. Seligman Ventures, which led the round, framed the opportunity around both the rising power demands of AI data centers and the growing need for energy-efficient computing in robotics, where battery and thermal limits make every watt count even more.

The first stage of the AI boom rewarded whoever could amass the most compute. The next one may reward whoever can run it on the least power, which turns efficiency from an engineering footnote into the product itself.

Related Articles

How 1001 Raised $30 Million to Bring Sovereign AI to Ports, Airports and Energy Grids

Amber Lands €7M to Turn Enterprise Knowledge into Autonomous AI

Cequence Security Launches Agent Personas to Automate AI Agent Governance Across the Enterprise