FuriosaAI Secures $620M at $2.9B Valuation to Scale Energy-Efficient AI Inference Chips

FuriosaAI Funding Round Quadruples Valuation

FuriosaAI, a South Korean AI startup specializing in energy-efficient AI processors, has closed a $620 million Series D funding round that values the company at $2.9 billion—a fourfold increase from its $735 million valuation just last year1

1

. The substantial investment signals growing confidence in the company's ability to address one of the AI infrastructure market's most pressing challenges: the data center power bottleneck that's limiting expansion across the industry.

South Korea's National Growth Fund led the round with a $280 million investment, representing the government's strategic push to strengthen the country's domestic semiconductor capabilities2

2

. International investors contributed $110 million, with US-based Infinitum serving as the largest foreign participant. The investor roster includes Jericho Capital, Worldline Capital, Cisco Investments, Abu Dhabi's NeoFort Holding, and Singapore's QUDA Capital Partners, whose founder Charlie Foo—a former Nvidia executive—now advises FuriosaAI1

1

.

RNGD Inference Chips Target Power Efficiency

The company's second-generation AI inference chip, RNGD (pronounced "renegade"), entered mass production in January through TSMC manufacturing. FuriosaAI positions these chips as a direct response to the energy constraints plaguing modern data centers, where computational power availability has become secondary to securing sufficient electricity to run processors. "We see strong demand and opportunities as we are scaling our business globally," chief executive June Paik stated1

1

.

The RNGD architecture represents a fundamental departure from traditional GPU design. While Nvidia's graphics processing units rely on rigid two-dimensional matrices for computations, RNGD-based chips employ three-dimensional matrices that treat tensor contractions as the primary compute unit2

2

. This Tensor Contraction Processor design minimizes energy-intensive data movement through deterministic execution on silicon, using less than half the power of Nvidia's GPUs while maintaining competitive performance.

Each RNGD card draws just 180 watts and requires only air cooling—a stark contrast to the liquid cooling systems many competing chips demand. FuriosaAI claims the chip matches Nvidia's RTX PRO 6000 throughput card-for-card while serving twice as many users per kilowatt, maximizing tokens-per-watt efficiency1

1

.

Production Scale and Customer Adoption

TSMC is manufacturing 20,000 RNGD chips for FuriosaAI this year, with plans to more than double production to over 40,000 units in 20271

1

. The company has already secured the memory capacity and factory allocation necessary for this expansion—a significant achievement given that Micron expects industry-wide memory shortages to intensify through 2028.

The South Korean AI startup has built meaningful traction with major customers. Samsung SDS launched a commercial cloud service powered by RNGD chips in July, while LG Group runs its EXAONE models on FuriosaAI's processors1

1

. Korean AI company Upstage leverages the chips to deliver live search summaries to millions of users2

2

. Paik indicated the company is in active discussions with potential customers across Asia-Pacific and the Middle East.

Next-Generation Development with Broadcom

The funding round will accelerate development of FuriosaAI's third-generation processor, which the company is co-developing with Broadcom and other strategic partners. This next iteration will utilize a cutting-edge 2nm multi-die design paired with HBM4 memory—representing the latest advances in semiconductor manufacturing and high-bandwidth memory technology1

1

.

"We are entering a new phase of bringing our technology to the global AI infrastructure market at scale," Paik emphasized. "By accelerating the commercialization of RNGD and advancing the development of Gen 3, we aim to help shape the next generation of AI infrastructure"2

2

.

John Yetimoglu, chief investment officer at Infinitum, highlighted the rarity of FuriosaAI's achievement: "Taking an advanced AI chip from design to volume production is an exceptionally difficult undertaking, and few companies have demonstrated the ability to do so successfully"1

1

. The company now operates offices across the United States, Singapore, and Lisbon, where it partnered with Equinix in July to bring RNGD technology to European markets.