High-performance networking chip startup Enfabrica Corp. today announced a significant new cash injection, raising $115 million as part of an oversubscribed equity financing round.

Today’s Series C round was led by Spark Capital and saw participation from new investors including Maverick Silicon and VentureTech Alliance. Existing investors Atreides Management, Sutter Hill Ventures, Alumni Ventures, IAG Capital and Liberty Global Ventures, which all took part in its Series B round last year, also participated.

Alongside the capital raise, Enfabrica lifted the lid on its upcoming new product, a “groundbreaking” 3.2 terabyte-per-second accelerated compute fabric of ACF SuperNIC chip, which it says will help to cement its status as a key player in the fast-growing artificial intelligence infrastructure industry.

Enfabrica believes its technology will sit at the heart of future graphics processing unit compute networks, which are the engine of modern AI applications. As AI technologies become more powerful and large language models require increasing amounts of data, its efficient server fabric chips will be able to solve critical challenges around network traffic congestion, input/output and memory scaling.

The startup was founded back in 2020 by its Chief Executive Officer Rochan Sankar and Chief Development Officer Shrijeet Mukherjee, plus other engineers, but it spent a lot of time in the shadows, secretly building out its technology. It only came to attention when it emerged from stealth mode in March 2023, before hitting the headlines when it raised $125 million in its Series B funding round the following September.

With the rise of chatbots such as OpenAI’s ChatGPT fueling enormous demand for generative AI applications and, now, AI agents, the startup launched its AI network interconnect chips at just the right time. It claims that its ACF SuperNIC chips can provide unmatched scalability and performance, with a lower total cost of ownership, for distributed AI workloads, plus extended-reality, high-performance computing and in-memory database applications.

As the company points out, the growth of AI is set to put enormous pressure on enterprises to sustain workload growth while keeping the number of GPUs and other processors they use to a minimum. That’s because GPUs are in high demand and very costly and hard to come by. Using its high-performance, converged memory and network fabric chips, companies will be able to squeeze the most performance out of their AI chips, Enfabrica says.

The new Enfabrica ACF SuperNIC is said to provide high-radix, high-bandwidth and concurrent PCIe/Ethernet multipathing and data mover capabilities, enabling it to connect between four and eight of the most advanced GPUs per server system, bringing what the company says is “unprecedented performance, scale and resiliency to AI clusters.”

It delivers four times the bandwidth and multipath resiliency of any other network interface controller currently available, the company said. With an 800-gigabit-per-second Ethernet interface, a high radix of 32 network ports and 160 PCIe lanes, it enables customers to build an enormous AI cluster containing more than 500,000 GPUs using an efficient, two-tier network design.

The startup said the ACF SuperNIC chip will become available in initial quantities in the first quarter of next year, and can be pre-ordered from Enfabrica and select partners from today.

Sankar said today’s round sets the stage for his company to become one of the industry’s leading AI networking chip and network software providers.

“We were the first to draw up the concept of a high-bandwidth network interface controller chip optimized for accelerated computing clusters,” he said. “The incredible syndicate of investors who are supporting our journey speaks to the commercial viability and value of our ACF SuperNIC silicon.”

Image: SiliconANGLE/Freepik AI