Ex-Google TPU Team’s MatX Raises $500M to Challenge Nvidia in the AI Chip Race

AI chip startup MatX has secured $500 million in Series B funding, stepping up its ambition to build processors it says could deliver up to 10× better performance than Nvidia GPUs for training and serving large language models.

The raise puts MatX squarely in the fast-intensifying battle to build the next generation of AI accelerators as demand for compute continues to surge.

Funding Round and Key Backers

The Series B round was led by Jane Street and Situational Awareness, an investment fund created by former OpenAI researcher Leopold Aschenbrenner.

Additional investors include:

• Marvell Technology
• NFDG
• Spark Capital
• Stripe co-founders Patrick Collison and John Collison

The new financing follows MatX’s roughly $100 million Series A in 2024, which was led by Spark Capital and valued the company at more than $300 million.

With the fresh capital, MatX now has significant runway to move from design to production.

Ambitious Positioning Against Nvidia

MatX is entering one of the most competitive segments in tech: AI compute hardware. The company claims its architecture could deliver up to 10× improvements over today’s Nvidia GPUs across both training and inference workloads.

If achieved, that level of performance gain would be highly disruptive in a market currently dominated by Nvidia’s CUDA ecosystem and data center GPUs.

However, the company still needs to prove its claims in real-world deployments, and commercial availability remains several years away.

Rising Competition From Etched and Others

The AI accelerator race is heating up quickly. Bloomberg recently reported that rival startup Etched also raised about $500 million, at a much higher $5 billion valuation, positioning it as one of MatX’s closest competitors in the emerging “post-GPU” category.

The broader trend is clear. As demand for large language model training explodes, investors are pouring capital into startups attempting to break Nvidia’s dominance.

The winners will likely be determined not just by raw chip performance, but by software ecosystem strength, manufacturing scale, and developer adoption.

Founders With Deep Google TPU Roots

MatX’s credibility rests heavily on its founding team’s background in Google’s custom AI hardware efforts.

CEO Reiner Pope previously led AI software development for Google’s TPU program

Co-founder Mike Gunter served as a lead hardware designer for TPUs

Google’s Tensor Processing Units are widely regarded as one of the few serious alternatives to Nvidia GPUs at hyperscale, giving the MatX team strong domain expertise.

The company was founded in 2023 with the goal of building purpose-built silicon optimized specifically for modern AI workloads.

Manufacturing Plan and Timeline

MatX plans to use TSMC as its manufacturing partner, a critical choice given the foundry’s leadership in advanced semiconductor fabrication.

According to the company’s roadmap:

• Chip manufacturing will ramp using the new funding
• First processors are expected to ship in 2027
• Initial focus is on large-scale LLM training and inference

The timeline reflects the long development cycles typical in advanced semiconductor design.

Why This Matters for the AI Infrastructure Race

The funding round highlights a broader structural shift in AI.

Three forces are converging:

• Explosive demand for AI compute
• Rising cost of Nvidia hardware
• Growing interest in specialized accelerators

Hyperscalers and AI labs are increasingly exploring alternatives to GPU-centric architectures, especially as training costs climb into the billions.

Startups like MatX are betting that purpose-built silicon can deliver better performance per watt and lower total cost of ownership.

The Real Challenge Ahead

Despite strong backing and technical pedigree, MatX faces significant hurdles.

Key risks include:

• Nvidia’s entrenched software ecosystem
• Long semiconductor development cycles
• Capital intensity of chip manufacturing
• Need for broad developer adoption
• Competition from well-funded rivals

Historically, breaking into the AI hardware stack has proven extremely difficult even for well-funded teams.

Bottom Line

MatX’s $500 million Series B marks one of the larger recent bets on next-generation AI accelerators. With founders from Google’s TPU program and backing from major investors, the startup is positioning itself as a serious long-term challenger to Nvidia.

But with chips not expected until 2027 and competition heating up, the real test will be whether MatX can translate ambitious performance claims into production-ready silicon and a developer ecosystem that can compete in the post-GPU era.