FuriosaAI raises $620m at $2.9b valuation for RNGD AI chip
South Korean chip designer FuriosaAI has raised $620 million at a $2.9 billion valuation and says its RNGD chip beat Nvidia's RTX Pro 6000 in its own efficiency test.

South Korean AI chip company FuriosaAI has raised $620 million in a funding round that values it at $2.9 billion, putting it among the better-funded challengers to Nvidia in the market for AI inference hardware.
Alongside the funding, the company made a performance claim about its RNGD processor. According to FuriosaAI, RNGD outperformed Nvidia's RTX Pro 6000 in an efficiency benchmark that the company ran itself.
What the benchmark does and does not show
The comparison is worth noting, but it comes with a caveat: the test was conducted by FuriosaAI, not by an independent third party. Efficiency results depend heavily on the workloads, models and settings chosen, so the figure is best read as the company's own positioning rather than a settled industry verdict. The material available does not detail the test methodology or the size of the gap.
Still, the choice of comparison is telling. The RTX Pro 6000 is a professional-grade Nvidia product, and measuring against it signals that FuriosaAI is targeting customers who care about performance per unit of energy and cost when running AI models, rather than only raw speed.
Why it matters
Nvidia dominates AI computing, and buyers have been looking for alternatives that can lower costs and ease supply constraints. A $620 million round gives FuriosaAI substantial capital to pursue that opportunity, from chip development to manufacturing and customer support, which are expensive for any chip designer.
The valuation of $2.9 billion also underlines continuing investor appetite for AI hardware startups outside the United States, with South Korea emerging as a notable base for such companies.
The key test now is whether FuriosaAI can turn its efficiency claim into customer deployments and whether independent benchmarks confirm the result. Those outcomes will determine if RNGD becomes a credible option for data centers and enterprises weighing alternatives to Nvidia's hardware.
More from this day
See all →
DRAM Contract Prices Hit Record in October as NAND Stalls

Sibilliter's Lypet AI Pet Care App Named Korea's A-Ventures No. 90

Edutem's SpeakingRise AI Korean Platform Adopted by UW



