Velaura AI Raises $110M to Fix AI’s Power Problem

Velaura AI, a Santa Clara chip designer, closed a $110 million Series A on 10 July 2026 at a valuation of roughly $1.05 billion, with investors backing technology that cuts power consumption in AI data centres. The round was led by Seligman Ventures, with new investor Capricorn Investment Group and existing backers Samsung Catalyst Fund, StepStone Group and Maverick Silicon all participating.

Velaura isn’t building a faster chip. It’s selling the ability to make chips other companies already build run on less electricity — and it’s already inside more than 30 million of them.

Key takeaways

• Raised: $110 million Series A, closed 10 July 2026.
• Valuation: approximately $1.05 billion, post-money.
• Led by: Seligman Ventures, with new investor Capricorn Investment Group.
• Existing backers returning: Samsung Catalyst Fund, StepStone Group, Maverick Silicon.
• Scale already reached: Velaura’s technology has been deployed in more than 30 million chips.
• The business model: upfront licensing fee plus a royalty tied to the power savings customers achieve — a structure the CEO compares directly to Arm’s early per-chip licensing model.
• Named product: Titan Core, a proprietary chip-design platform claiming a 2–4x improvement in performance-per-watt for AI accelerator math.

What Velaura actually sells

Velaura AI develops low-power chip design technology and software for data centres and “physical AI” applications — robotics and autonomous systems. It does not manufacture finished chips. It licenses IP that other companies’ chip designs are built on, aimed specifically at cutting power draw without giving up performance.

Its flagship platform, Titan Core, is a proprietary chip-design system the company says delivers a 2–4x improvement in performance-per-watt specifically for the mathematical operations that dominate AI accelerator workloads. In an industry where power, not silicon, is becoming the binding constraint on how much AI compute can be built, that is the pitch investors are underwriting.

The Arm comparison, from the CEO’s own mouth

Rajiv Khemani, Velaura’s co-founder and CEO, described the pricing structure directly to Reuters: an upfront fee for the technology, plus a royalty tied to a share of the power savings the customer actually achieves. He confirmed that structure is similar to Arm’s per-chip licensing model — the same model Arm used before it started selling its own chips and became one of the most valuable companies in semiconductors.

“The next era of AI will be defined not only by better models, but also by fundamentally better compute economics,” Khemani said in a statement announcing the round.

That is a specific and checkable claim about the business, not marketing language. A royalty tied to measured power savings, rather than a flat per-unit fee, means Velaura’s revenue scales with how well the technology actually performs in the field — aligning its incentives with the customer’s, and giving outsiders a testable metric to watch as deployments grow.

Who is actually buying this

Khemani told Reuters that Velaura is engaged with three of the four largest cloud computing providers as potential customers, though he declined to name them. That detail matters more than the funding round itself: hyperscalers evaluate power-efficiency technology at enormous scale, and even a modest efficiency gain compounds into large savings across a data centre fleet running continuously.

The Samsung Catalyst Fund’s continued participation is a second signal worth reading. Samsung is both an investor and a potential manufacturing and design partner in the semiconductor supply chain, so a repeat check from its venture arm suggests more than a passive financial bet.

Why power, not speed, is the fight now

The AI industry has spent several years racing to build faster chips. Power has quietly become the harder constraint: data centre electricity demand is straining grids, and the physical limit on how much AI compute a company can deploy is increasingly set by how many megawatts it can secure, not how many GPUs it can buy.

That reframes the competitive question. A chip that is 20% slower but uses 60% less power can serve more total AI workload than a faster chip that a data centre cannot get enough electricity to run at scale. Velaura’s target market — companies already built around GPU or custom-silicon workloads that need more compute without more grid capacity — is betting on that reframing being real and durable, not a passing narrative.

Where this fits in the broader AI infrastructure build-out

Velaura’s raise lands alongside a wave of infrastructure investment chasing the same underlying constraint. Compute-hungry AI labs are themselves seeking speed gains through specialised hardware — see Cerebras’ new CS-4 inference chip, which competes on raw speed rather than efficiency. On the capital side, Nikhil Kamath’s Rs 200 crore backing of hyperscale data-centre operator CtrlS shows Indian capital chasing the same data-centre build-out from the real-estate and power side rather than the chip side.

What’s unverified

• Named hyperscaler customers. “Three of the four largest cloud providers” is Khemani’s characterization to Reuters; none has been named or confirmed independently.
• The 2–4x performance-per-watt figure is a company claim about Titan Core, not an independently benchmarked result.
• Revenue and margins have not been disclosed. The royalty structure is described qualitatively, not with actual rates.

Frequently asked questions

What does Velaura AI actually make?

Velaura licenses chip-design technology and software — principally its Titan Core platform — that reduces power consumption in AI data centre and robotics chips. It does not manufacture chips itself.

How much did Velaura AI raise and at what valuation?

$110 million in a Series A round closed 10 July 2026, at a valuation of roughly $1.05 billion, led by Seligman Ventures.

How does Velaura make money?

An upfront licensing fee plus an ongoing royalty tied to the share of power savings a customer achieves — a structure the company compares to Arm’s early per-chip licensing model.

Who are Velaura’s customers?

Not publicly named. CEO Rajiv Khemani said the company is engaged with three of the four largest cloud computing providers as potential customers.

The bottom line

Investors are not betting on Velaura building the fastest AI chip. They’re betting that as power becomes the binding constraint on AI infrastructure, a licensable efficiency layer — sold the way Arm sold instruction-set licenses before it made its own silicon — becomes as valuable as the chip itself. Thirty million chips already carrying the technology, and interest from three of the four largest cloud providers, are the two facts that make that bet look more than theoretical.

Reported from Reuters, via The Economic Times, and company statements.