MatX’s Series B Fuels a New Front in the AI Chip Wars

MOUNTAIN VIEW, Calif.: MatX has raised $500 million in Series B funding to scale its AI accelerator chips for data centers, taking direct aim at Nvidia in the AI silicon market.

The startup focuses on custom hardware optimized for transformer-based LLM inference. Co-founders Mike Gunter and Reiner Pope bring deep credibility to the venture, having previously led TPU development at Google before leaving to build accelerators purpose-designed for modern AI workloads.

The chip startup competes directly against Nvidia in the high-stakes AI silicon market. MatX targets large language model workloads with custom hardware optimized for inference. The round fuels aggressive chip development as demand for AI compute keeps surging.

MatX founders Mike Gunter and Reiner Pope previously worked at Google on TPU teams. The pair built the chips powering Google’s largest AI training efforts. They left to design accelerators specifically optimized for transformer architectures.

The startup’s chips reportedly deliver superior performance per dollar on language model workloads. Customers running massive AI inference pipelines see significant cost reductions versus GPUs. The advantage matters as inference costs eclipse training spend across the industry.

The round arrives amid a brutal compute shortage hitting every major AI lab. Anthropic recently tapped SpaceX’s Colossus data center for emergency compute capacity. Even well-funded labs struggle to secure sufficient infrastructure at necessary scale.

Nvidia continues dominating AI chip sales with over 90% market share globally. Challengers like Cerebras, Groq, and Tenstorrent target specific workload niches. MatX bets that transformer-optimized silicon can win meaningful share in inference.

“The economics of AI inference will determine which companies survive the next decade,” says a venture investor familiar with the round.

The funding follows Cerebras’ recent IPO filing targeting a $26.6 billion valuation. Lambda Labs also pursues a public listing in the second half of 2026. AI infrastructure plays attract sustained capital despite broader market uncertainty.

MatX plans to use proceeds for engineering hires and fabrication partnerships. The company has not disclosed customer names or revenue figures publicly.