Delos Data, a startup specializing in AI inference hardware, has raised over $100 million in funding. The capital will accelerate its sales efforts and engineering hires, as the company introduces a new networking chip with 10x lower latency than rivals.
TL;DR
- Delos Data raises $100M+ to scale its AI inference hardware business.
- The funding will support hiring and sales expansion, following the launch of a new networking chip.
- The Delos Nonstop AI Data Interface offers 10x lower latency and improved efficiency for AI clusters.
What happened
Delos Data, a startup developing hardware for AI clusters, has secured over $100 million in funding from investors including Matrix, Playground, Socratic Partners, Capricorn, Matter Venture Partners, and IAG. The raise also included contributions from angel investors in the data center hardware industry.
The funding comes three months after Delos launched its first product, the Delos Nonstop AI Server, a computing appliance optimized for inference. Customers can equip it with up to four graphics cards, which connect via an open-source interconnect called OAM, optimized for AI use cases.
In conjunction with the funding announcement, Delos introduced the Delos Nonstop AI Data Interface, a chip designed to work with the Nonstop AI Server. This chip enables servers to connect to inference cluster networks, offering 10x lower latency, improved efficiency, and scalability compared to rival products. It also monitors clusters for malfunctions and automates some recovery processes.
Why it matters
For developers and startups, Delos's hardware offers a competitive alternative to Nvidia's Rubin appliances, with comparable network capacity and lower latency. The open-source interconnect and routing technology provide flexibility and scalability for AI clusters.
Investors see potential in Delos's focus on AI inference hardware, a critical component for scaling AI applications. The funding will enable Delos to accelerate its sales efforts and hire more engineers, potentially expanding its product lineup and market reach.
However, Delos faces competition from established players like Nvidia. The success of its hardware will depend on its ability to deliver on performance promises and gain traction in the market.
Key facts
- Funding amount: Over $100 million.
- Investors: Matrix, Playground, Socratic Partners, Capricorn, Matter Venture Partners, IAG, and angel investors from the data center hardware industry.
- Product: Delos Nonstop AI Server, optimized for inference with up to four graphics cards.
- Networking capacity: Each of the four chips in the Nonstop AI Server has nine OSFP ports with 1.6 terabits per second of bandwidth apiece.
- New product: Delos Nonstop AI Data Interface, a chip offering 10x lower latency and improved efficiency for AI clusters.
- Implementations: The Nonstop AI Data Interface is available as a PCIe card, a chiplet processing over 30 terabits of traffic per second, and a near-packaged optics (NPO) version with a capacity of over 10 terabits per second.
- Use of funds: Hiring more engineers and accelerating sales efforts.
Context
AI inference hardware is a critical component for scaling AI applications, enabling real-time processing and decision-making. As AI models grow larger and more complex, the demand for high-performance, scalable inference hardware is increasing.
Delos Data's focus on open-source interconnects and routing technologies aligns with industry trends towards open and interoperable AI hardware ecosystems. This approach could attract developers and startups looking for flexible and scalable solutions.
The competition in the AI inference hardware market is intense, with established players like Nvidia offering high-performance solutions. Delos's success will depend on its ability to differentiate its products and gain market traction.
