Velaura AI Raises $110 Million to Develop Lower-Power AI Chips

Key Takeaways

• Velaura AI secured a $110 million Series A led by Seligman Ventures, lifting its valuation above $1 billion.
• The chip designer is working with three of the four largest cloud computing providers, although their identities were not disclosed.
• Velaura AI is targeting the power consumption and operating costs associated with running increasingly large AI workloads in data centers.

Velaura AI has entered the ranks of billion-dollar semiconductor startups after raising $110 million to develop chips intended to reduce the energy demands of artificial intelligence infrastructure.

The Series A was led by Seligman Ventures, with Samsung Catalyst Fund and StepStone Group among the participating investors, according to Reuters. The financing valued Velaura AI at more than $1 billion, giving the chip designer substantial backing as cloud operators search for ways to expand AI capacity without allowing power and infrastructure expenses to rise at the same rate.

Velaura AI is already engaged with three of the four largest cloud computing providers, a company executive told Reuters. The executive declined to identify those providers. That distinction matters because an engagement can cover several stages, from technical evaluation and software testing to a potential production deployment. It does not, on its own, establish that Velaura AI’s chips have been selected for large-scale commercial use.

Getting into discussions with major cloud operators is an important early test, as these companies operate some of the world’s largest AI clusters and evaluate processors against demanding requirements for throughput, reliability, energy use and integration with existing infrastructure.

Raw processing speed is only part of the buying decision now. Data center operators also have to account for electricity availability, cooling systems, networking capacity and the cost of keeping accelerators busy. A processor that performs well but spends too much time waiting for data, or requires extensive changes to software, may offer limited operational savings.

Gartner forecast worldwide AI chip revenue of about $71 billion in 2024, representing growth of 33% from 2023. AI accelerators used in servers were expected to account for roughly $21 billion of that total, with the server accelerator segment projected to reach $33 billion by 2028.

Those figures help explain why investors continue to fund specialized silicon despite the cost and technical risk of chip development. AI accelerators are becoming a routine part of server configurations as cloud and enterprise customers deploy generative AI, machine learning and inference services. Industry reporting from TechRepublic has likewise highlighted the growing enterprise focus on the infrastructure required to support AI systems, not merely the applications placed on top of them.

Velaura AI is entering a market led by NVIDIA’s GPU-based accelerators and contested by AMD’s Instinct portfolio, along with companies designing application-specific integrated circuits. The startup’s pitch centers on lowering power consumption and operating costs, two areas where cloud customers have strong incentives to consider alternatives.

Whether efficiency gains can overcome the advantages held by established platforms will depend on more than chip architecture. Software compatibility is often a decisive factor. Developers typically expect support for widely used frameworks such as PyTorch and TensorFlow, while data center operators need processors that can work with established networking and interconnect technologies. Compatibility with PCI Express and IEEE Ethernet standards can help reduce integration friction, although production readiness also depends on drivers, compilers, management tools and dependable access to chips.

Large cloud providers increasingly develop their own AI silicon, particularly for inference and selected internal workloads. Velaura AI therefore may compete with internal chip programs while also seeking those same operators as customers or development partners. That makes flexibility valuable. A design that addresses a clear performance-per-watt or cost-per-query requirement could still find room within a mixed computing environment.

The $110 million round gives Velaura AI resources to continue development and pursue demanding cloud evaluations. It also raises expectations. The next indicators will be concrete deployments, repeatable performance results and evidence that customers can adopt the technology without rebuilding their software environments. In AI infrastructure, an impressive chip is a start; turning it into an economical, usable system is the harder part.