Enfabrica, a Silicon Valley semiconductor firm developing networking chips for AI data centers, said on Tuesday that it had secured $125 million in venture funding. Nvidia also invested as a strategic partner.
Enfabrica, established by executives from Google (GOOGL.O) and Broadcom (AVGO.O), is part of a larger trend in which data centers are redesigned to produce generative AI technologies comparable to ChatGPT. The core of that change is a set of chips from Nvidia (NVDA.O), the most valuable semiconductor company in the world.
The issue with Nvidia’s graphics processing unit (GPU) processors is that they sometimes remain inactive because the networks that link them cannot provide them with data at a high enough rate.
To address this issue, Enfabrica has created a network chip that seeks to link the many components of a data center in novel ways. The Enfabrica chip builds a network that resembles a hub and spokes, which enables the Nvidia GPUs doing the data-crunching to pull data from several locations without encountering roadblocks.
Rochan Sankar, co-founder and CEO of Enfabrica, said that this results in considerably more effective use of GPUs, allowing the completion of the same amount of computational work with just roughly half as many chips. Such an accomplishment is seen favorably in the technology sector since busier chips are more economical.
According to Sankar, it’s no secret to Nvidia or anybody else out there that the cost curve has to decrease for AI computing to become genuinely commonplace. The crucial factor in this situation is making those GPUs more useful.
Gavin Baker, a former employee of Fidelity Investments and founder of Atreides Management, was on the board of Enfabrica before leading the Series B venture capital round. IAG Capital Partners, Liberty Global Ventures, Valor Equity Partners, Infinitum Partners, and Alumni Ventures were among the new investors recruited by Enfabrica. Sutter Hill Ventures, an earlier investor, also participated in the round.
