South Korean AI chipmaker FuriosaAI has raised $620 million in a Series D financing round that values the company at $2.9 billion, the company announced on October 29, 2026. The round, reported by Bloomberg to quadruple the company’s previous valuation, gives the Seoul-based startup a much larger funding base as it scales production of its RNGD inference processor and takes on Nvidia’s dominance in specialised AI hardware.
South Korea’s National Growth Fund contributed $280 million of the round, the largest single commitment, while international investors contributed about $110 million. Those international backers include Infinitum, Jericho Capital, QUDA Capital, Cisco Investments, NeoFort Holding and Worldline Capital. Returning investors include Naver, DSC Investment, Korea Development Bank, Nextrans and Kakao Investment.
Key Takeaways
- FuriosaAI raised $620 million in a Series D round at a $2.9 billion post-money valuation, announced on October 29, 2026.
- South Korea’s National Growth Fund put in $280 million, with about $110 million from international investors including Cisco Investments and Infinitum.
- The funding will scale production of the RNGD inference chip, which entered mass production in January 2026, and support a third-generation accelerator being developed with Broadcom.
What FuriosaAI actually builds
FuriosaAI, founded in 2017, designs processors optimised for AI inference, the process through which trained models generate responses. That is a different job from training a model in the first place, and it is where a growing share of AI computing cost now sits: every chatbot answer, search summary and generated image runs through inference hardware.
Its second-generation processor, RNGD, is available as a 180-watt accelerator card and in server configurations containing multiple accelerators. The company says RNGD entered mass production in January 2026 and expects to produce approximately 20,000 cards during 2026, with manufacturing and memory capacity secured to more than double that output in 2027. Those production figures are company targets, not completed sales, but they give a measurable test of whether its chips can move from pilot deployments to volume supply.
FuriosaAI says its design treats tensor contractions as the primary unit of compute, rather than the two-dimensional matrix processing used in graphics processors, and claims the design consumes less than half the electricity of Nvidia hardware. That efficiency claim matters because inference runs constantly, and electricity is now one of the largest operating costs for AI services.
Customers already in deployment
The round is notable because FuriosaAI is not only selling a chip design. The company says RNGD is already in production deployments with leading enterprises, including Samsung SDS and LG Group. LG Group deploys the silicon for its EXAONE models, Upstage runs live search summaries on it, and Samsung SDS offers the processors through a cloud service, according to reports of the announcement.
That commercial traction was central to the investor case. John Yetimoglu, chief investment officer with Infinitum, said taking an advanced AI chip from design to volume production is exceptionally difficult, and that few companies have demonstrated the ability to do so successfully. He added that FuriosaAI has gone further, securing production deployments with leading enterprises such as Samsung SDS and LG Group.
The funding will scale production of a chip that is already in deployment, rather than funding a design that has yet to reach customers.
Why investors are hunting for Nvidia alternatives
Nvidia’s chips dominate AI data centres, but that dominance is itself the opportunity for challengers. Investors are hunting for alternatives that are more energy-efficient and more specialised, particularly for inference, where a chip tuned to one workload can beat a general-purpose accelerator on cost per answer.
The same logic is driving competition across the AI stack, from new AI models to cheaper AI coding tools. Hardware is the most capital-intensive part of that race, which is why a $620 million round, backed in large part by a state-backed growth fund, is a signal of how seriously South Korea takes building a domestic alternative.
- $620 million Series D at a $2.9 billion post-money valuation
- $280 million from South Korea’s National Growth Fund, about $110 million from international investors
- RNGD in mass production since January 2026, with about 20,000 cards targeted for 2026
- Deployments reported with Samsung SDS, LG Group and Upstage
- Third-generation accelerator in development with Broadcom
The next test: volume and the third generation
FuriosaAI is also developing a third-generation accelerator with partners including Broadcom, aimed at large-scale AI infrastructure. The Series D funds both tracks at once: scaling RNGD deployments globally and advancing that next chip.
The risk is execution. Chip startups face long manufacturing cycles, tight memory supply and customers that are slow to qualify new hardware. A valuation that has quadrupled in one round raises the bar for the next one: FuriosaAI must now show that its 2027 production ramp lands, that its deployments expand beyond its early customers, and that the Broadcom partnership produces a competitive third-generation product.
What to Watch Next
Watch for FuriosaAI’s 2027 production numbers against its target of more than doubling 2026 output, for new customer announcements beyond Samsung SDS, LG Group and Upstage, and for further detail on the third-generation accelerator being developed with Broadcom. Each of those milestones will show whether this round bought scale, or only time.