South Korean inference chip startup FuriosaAI Inc. said today it’s ready to step up its challenge to Nvidia Corp. after raising a bumper $620 million round of funding to scale production of its energy-efficient artificial intelligence processors.

The round brings FuriosaAI’s valuation to a cool $2.9 billion, the startup said.

The investment was led by South Korea’s National Growth Fund, which injected $280 million of the total amount raised as part of a government-backed plan to boost the country’s domestic semiconductor industry. Global investors including Cisco Investments, Abu Dhabi-based NeoFort Holdings, QUDA Capital Partners, Jericho Capital and Worldline Capital added another $110 million, while existing backers including the Korea Development Bank, Kako Investment and NAVER contributed the rest.

The nine-year-old startup and its investors believe it has a critical advantage over Nvidia. According to them, the bottleneck in AI data centers now is not the availability of computational power, but the lack of sufficient energy to power all of those processors.

Nvidia’s power-hungry chips still dominate the industry, but many data centers are now struggling to scale up because they can’t get enough power to feed all of those processors. Energy grids and cooling systems are being stretched as far as they can go, which is why FuriosaAI’s inference chips, which are designed to maximize tokens-per-watt, are such a tempting alternative.

FuriosaAI introduced its energy-efficient Tensor Contraction Processor earlier this year. It’s the second generation of inference chips based on its RNGD architecture. Unlike Nvidia’s graphics processing units, which use a rigid two-dimensional matrix for processing computations, RNGD-based chips use a three-dimensional matrix.

By treating tensor contractions as the primary unit of compute and using deterministic execution on the silicon, the TCP chips keep energy-slurping data movement to a minimum. As a result, they use less than half the power of Nvidia’s GPUs, while offering compatibility with AI frameworks such as PyTorch and vLLM.

The startup’s chips have already won over a significant fanbase in South Korea. For instance, LG Group, the parent company of LG Electronics Co. Ltd., is using FuriosaAI’s chips to power its EXAONE models. Meanwhile, the Korean AI startup Upstage is using the chips to power live search summaries for millions of users, and Samsung Electronics Co. Ltd. offers access to the company’s silicon through its commercial neural processing units-as-a-service offering that launched in the summer.

The funds from today’s round will help FuriosaAI to scale up production volumes of the next-generation TCP chips, the startup said. It’s planning to produce 20,000 this year and double that to 40,000 in 2027, before scaling further in the coming years. At the same time, it’s also working on a third-generation processor that’s being co-developed with Broadcom Inc.

Co-founder and Chief Executive June Paik (pictured) said the size of today’s round reflects the strong confidence investors have in its technology. “We are entering a new phase of bringing our technology to the global AI infrastructure market at scale,” he insisted. “By accelerating the commercialization of RNGD and advancing the development of Gen 3, we aim to help shape the next generation of AI infrastructure.”

Yong Jung Park, a partner at NAVER Ventures, which has participated in every funding round dating back to FuriosaAI’s pre-seed in 2017, said the company has cleared the most difficult hurdle that semiconductor firms face – turning architecture into proven, commercial hardware. “We look for the moment when technology companies go from proving their technology to proving their business model,” Park said. “FuriosaAI is at that point now, with its chip in volume production and in deployment with customers.”

Image: FuriosaAI