Nvidia challenger FuriosaAI seeks up to $500M raise after declining Meta buyout
- ByStartupStory | January 23, 2026
As investment in AI hardware accelerates, the market for specialised chips is becoming increasingly crowded and competitive. Recently, we saw Cerebras Systems and Etched raise notable rounds to carve out positions for themselves in the market dominated by Nvidia.
Now, it is reported that Seoul-based AI chipmaker FuriosaAI is in discussions to raise between $300 million and $500 million in a Series D funding round. It is believed that this could set the stage for a public listing as early as 2027. The raise is being guided by Morgan Stanley and Mirae Asset Securities, signalling growing global investor confidence in an emerging challenger to Nvidia.
A funding push with a potential IPO
The upcoming Series D round is expected to be one of FuriosaAI’s last private raises before entering public markets. While the valuation has not been disclosed, the company was valued at $735 million in July after closing a $125 million Series C bridge round.
Earlier this year, the startup reportedly declined an $800 million takeover offer from Meta Platforms, reinforcing its intent to remain independent and scale on its own terms.
The new capital will be directed toward mass production of its second-generation RNGD chip, expansion into global markets, and development of a third-generation processor. Another notable development is that FuriosaAI is set to receive its first mass-produced shipment of RNGD chips from Taiwan Semiconductor Manufacturing Co., marking a move from development to large-scale deployment.
Rethinks how AI computations are done
Founded in 2017 by June Paik, a semiconductor veteran with two decades of experience at Samsung Electronics and AMD, FuriosaAI has built its identity around efficiency. Instead of optimising its chips for matrix multiplication, as GPUs do, the company redesigned its architecture around tensor contraction.
While both mathematical paths can reach the same result, tensor contraction allows computations to be executed more efficiently. FuriosaAI argues this shift unlocks better performance while consuming less power, a growing concern as AI workloads scale.
The RNGD chip (pronounced “renegade”) reflects this approach. According to the company, it delivers 2.25 times better inference performance per watt than traditional GPUs. The architecture enhances parallelism, allowing more calculations to run simultaneously, and increases data reuse, reducing memory access delays that often slow down AI systems.
Performance without excess power
From a hardware perspective, RNGD is designed for practical deployment. The chip can perform up to 512 trillion calculations per second using FP8 data and operates within a 180-watt thermal envelope, removing the need for complex water-cooling systems. It is offered as a PCIe card and also powers the NXT RNGD server, which combines eight accelerators to deliver up to four petaflops of performance.
This focus on efficiency aligns with FuriosaAI’s broader mission to run the world’s most advanced AI models without escalating energy costs. As investment surges across the AI hardware sector.
What’s next?
With two funding rounds in just six months, FuriosaAI is positioning itself as a sustainability-driven alternative to its competitors. If the Series D round closes as planned, the company will enter its next phase with capital, manufacturing momentum, and a clear path toward public markets.
Follow Startup Story
Related Posts
