Velaura AI raises $110M to push ultra-low-power AI compute with Titan Core

Velaura AI, a startup that pivoted from bitcoin mining to AI silicon, just raised $110 million in Series A funding at a valuation above $1 billion. The company's Titan Core platform offers modular processor building blocks and a toolflow designed to cut the energy cost of AI compute by 2-4x on matrix-multiplication workloads. For builders shipping inference at scale or deploying AI at the edge, this signals a growing push to make power efficiency a first-class design constraint.

From crypto mining to AI silicon

Velaura was originally Auradine, a bitcoin-mining chip company. Its most advanced crypto processor, the liquid-cooled Teraflux AH3880, could perform 600 trillion computations per second and included an EnergyTune feature to reduce power draw during grid constraints. In March 2026, the company rebranded and shifted its focus to AI compute with the launch of Titan Core.

Titan Core is not a single chip. It is a suite of pre-packaged cell designs, a proprietary toolflow, and professional services that let customers design their own low-power AI processors. Chip teams provide an RTL file describing the processor, and Velaura's tooling handles the low-voltage circuit implementation. The company claims Titan Core supports 3nm and 2nm manufacturing nodes and is already working with multiple hyperscalers on chip projects.

Why energy efficiency matters for AI builders

Matrix multiplications can account for up to 70% of an AI chip's power usage, according to Velaura. Reducing that by a factor of two to four translates directly into lower operating costs and thermal budgets. For a builder running large-scale inference, that could mean more throughput per watt or the ability to deploy AI in power-constrained environments like edge devices or robotics.

Velaura claims the savings can reach $1,300 per chip over three years. That number is a vendor projection, but even a fraction of that would shift the economics for high-volume deployments.

What Titan Core changes in practice

The modular approach means a team building a custom AI accelerator does not need to design every transistor-level cell from scratch. Titan Core's cell library provides pre-optimized low-voltage building blocks, and the proprietary toolflow aims to improve manufacturing yield and reliability. For hyperscalers designing their own inference chips, this could reduce R&D time and risk.

Beyond data centers, Velaura targets edge AI and Physical AI applications including robotics and autonomous systems. The EnergyTune feature, inherited from the crypto chip, allows chips to dynamically lower power consumption when grid capacity is tight, which matters for edge deployments with limited or variable power.

Caveats to watch

The energy savings and node support claims are based on company statements and press coverage, not independent benchmarks. Real-world performance will depend on customer designs, manufacturing process maturity, and actual deployment conditions. Velaura has not disclosed which hyperscalers it is working with or when first customer chips will tape out. The company's pivot from crypto to AI is recent, and its track record in AI silicon is unproven.

The funding will go toward engineering and customer-facing teams, so expect more details on specific partnerships and tape-out timelines in the coming quarters.