In a single week in February 2026, two AI chip startups collectively raised $750 million in funding—signaling a growing wave of venture capital backing for alternatives to Nvidia's overwhelming market dominance. MatX secured $500 million in Series B funding, while Axelera AI raised over $250 million, bringing total funding for these two companies to $750 million in just seven days.

This funding surge comes as Nvidia controls an astonishing 85-92% of the AI chip market, with a backlog exceeding $500 billion that analysts say could drive revenue above $300 billion in 2026. But VCs are betting billions that the market is ready for specialized silicon optimized for specific AI workloads rather than general-purpose GPUs.

What Happened: The Funding

MatX, founded by former Google TPU engineers Reiner Pope and Mike Gunter, raised $500 million in Series B funding led by Jane Street and Leopold Aschenbrenner's Situational Awareness fund, with participation from Marvell Technology and Stripe co-founders. The company has now raised over $600 million total since its 2022 founding.

The funds will accelerate development of MatX One, an LLM-optimized accelerator designed to handle pre-training, reinforcement learning, inference prefill, and decode—differentiating it from inference-only competitors like Groq. The chip combines ultra-fast SRAM for low latency with high-capacity HBM for long-context support and KV caching, aiming for higher throughput than current GPUs while maintaining SRAM-first latency.

Meanwhile, Dutch startup Axelera AI raised $250 million in a round led by Innovation Industries, with participation from BlackRock and Samsung Catalyst Fund—Europe's largest AI chip funding round. This brings Axelera's total funding to $450 million. The company focuses on power-efficient edge AI inference chips, targeting robotics, computer vision, and IoT devices.

The two funding rounds contributed to AI chip startups collectively attracting $1.1 billion in VC funding that week alone.

Why It Matters: Breaking Nvidia's Monopoly

Nvidia's dominance in AI chips is unprecedented. The company holds between 85-92% market share in AI accelerators, with a $500 billion backlog that ensures revenue of $300 billion or more in 2026. This monopoly-like position has created both opportunity and urgency for competitors.

VCs are responding. AI startup funding reached $202 billion in 2025, up 75% year-over-year, with 58% of that capital flowing to mega-rounds. The strategic shift is clear: investors are backing purpose-built alternatives that promise 10x efficiency gains for LLM training and edge inference.

Axelera's Europa chip exemplifies this approach. It matches Nvidia A100 compute while consuming just 1/6th the power (approximately 45W versus ~700W), processing over 13,000 FPS for vision models. The company uses digital in-memory computing (D-IMC) with SRAM crossbar arrays to minimize data movement—a fundamentally different architecture from general-purpose GPUs.

The edge AI market is projected to reach $255 billion by 2030, up from $106 billion in 2025—a 140% growth trajectory that makes the segment particularly attractive for specialized chips.

What's Next: The Race to Specialization

The $750 million in week-old funding represents more than capital—it's a bet that the AI hardware market is fragmenting. As LLM training moves from general pre-training to specialized fine-tuning and inference, purpose-built chips optimized for specific workloads could capture meaningful share from Nvidia's general-purpose architecture.

MatX plans to launch MatX One later in 2026, targeting datacenters with clusters capable of scaling to hundreds of thousands of chips. Axelera is expanding manufacturing and customer teams while building partnerships across robotics and edge computing ecosystems.

What to watch: Whether these specialized chips can actually deliver on efficiency promises in production environments, and whether Nvidia responds with its own specialized hardware lines. The $1.1 billion in weekly AI chip funding signals that VCs believe the monopoly is vulnerable—the question is just when and how.

Resources