FuriosaAI targets $500M funding round as RNGD chip production begins, eyes 2027 IPO

FuriosaAI Launches Major Funding Round to Scale AI Chip Production

FuriosaAI, a South Korea-based AI chip designer, is seeking to raise between $300 million and $500 million in a Series D funding round as it positions itself to compete directly with industry leader Nvidia1

1

2

2

. The Seoul-based startup has appointed Morgan Stanley and Mirae Asset Securities Co., Korea's largest investment bank, as co-advisers for the fundraising effort1

1

. This AI chip developer reached a valuation of $735 million as of July 2025, when it closed a $125 million investment from institutional investors2

2

. Notably, the company reportedly turned down an $800 million acquisition offer from Meta Platforms Inc. just months earlier, signaling its ambitions for independence and growth.

Mass Production Milestone and Global Expansion Plans

The funding round comes at a critical juncture for FuriosaAI, as the company is set to receive its first mass-produced shipment of RNGD AI chips from Taiwan Semiconductor Manufacturing Co. (TSMC) later this month1

1

. Proceeds from the capital raise will fund the mass production of the second-generation RNGD chip, support global expansion efforts, and accelerate development of a third-generation chip1

1

. The RNGD chip, pronounced "renegade," delivers 2.25 times the inference performance-per-watt of traditional graphics processing units, according to the company1

1

. This efficiency advantage positions FuriosaAI's AI inference chips as compelling alternatives in a market dominated by power-hungry accelerators.

Innovative Tensor Contraction Architecture Sets RNGD Apart

What distinguishes the RNGD chip from conventional AI accelerators is its underlying mathematical approach. Instead of optimizing for matrix multiplications like server-grade graphics processing units, FuriosaAI designed its processor around tensor contraction operations2

2

. The company asserts that tensor contraction can produce identical results to matrix multiplication but with greater efficiency, translating into faster AI performance2

2

. This architectural choice improves neural network parallelism, allowing more calculations to execute simultaneously rather than sequentially, while also increasing data reuse to minimize memory retrieval delays2

2

.

The RNGD can perform up to 512 trillion calculations per second on data stored in the FP8 format and ships in a PCIe card with a thermal footprint of 180 watts, eliminating the need for water cooling2

2

. FuriosaAI also offers the NXT RNGD server, which includes eight RNGD accelerators and provides up to 4 petaflops of performance2

2

.

Path to IPO Amid Surging AI Hardware Investment

The funding round could reportedly be FuriosaAI's last before an initial public offering, with the company eyeing a potential listing as early as 20271

1

2

2

. Founded in 2017 by June Paik, a veteran of Samsung Electronics Co. and Advanced Micro Devices Inc., FuriosaAI specializes in high-efficiency AI inference chips1

1

. The milestone arrives amid a boom in AI hardware investment, with competitors securing substantial capital. Groq Inc. raised $750 million in a round last year ahead of a strategic partnership with Nvidia, while Cerebras Systems Inc. pursued $1 billion in new capital at a $22 billion valuation, and Etched secured a $500 million funding round1

1

. For investors and industry watchers, FuriosaAI's trajectory signals growing confidence in alternatives to established chip architectures, particularly as enterprises seek more energy-efficient solutions for AI inference workloads at scale.