Nvidia controls 80%+ of the AI accelerator market — but 43 US startups have raised $17.7 billion to change that. This dashboard maps the entire private AI chip landscape: from transformer-specific ASICs (Etched, Groq) to photonic interconnects (Lightmatter, Ayar Labs) to brain-inspired analog chips (Unconventional AI, Rain AI).
Cerebras Systems became the first AI chip IPO of 2026, opening at $385/share ($66B market cap). Groq's LPU technology was licensed by Nvidia for ~$20B. Etched emerged from stealth with $800M raised and $1B+ in signed contracts. The race to dethrone Nvidia is accelerating across every compute layer — datacenter, edge, networking, and exotic physics.
Cerebras Systems leads with $4B+ raised (now public at $66B market cap), followed by Groq (~$3B, LPU inference), SambaNova (~$1.5B, RDU accelerator), Astera Labs (~$860M, connectivity silicon, now public), and Lightmatter (~$850M, photonic compute). The top 12 companies account for over 80% of total funding.
The 43 startups span 7 segments: Datacenter ASICs/accelerators (13 companies, $11.3B — the most funded), Photonics (6 companies using light instead of electrons), Edge AI (9 companies for on-device inference), Analog/In-Memory compute (4 companies), Exotic/Neuromorphic (4 companies including brain-inspired and thermodynamic approaches), Interconnect/Networking (6 companies connecting GPU clusters), and Infrastructure (1 company).
Several are finding wedges. Etched's Sohu ASIC claims to replace ~160 H100s with one 8-chip server for transformer inference. Groq's LPU delivers deterministic ultra-low latency. Cerebras's wafer-scale chip eliminates chip-to-chip bottlenecks. Lightmatter and Ayar Labs attack the interconnect layer where GPUs waste time waiting for data. The most likely path to disruption is specialization — beating Nvidia on specific workloads rather than general-purpose compute.
Intel Capital leads with 8 portfolio companies, followed by Fidelity (4), Atreides (4), and BlackRock (3). Trading firms like Jane Street, Jump Trading, and Two Sigma have also entered the space, backing Etched and Groq — a sign that low-latency inference has financial applications beyond AI.
The AI Chip Wars dashboard tracks the public incumbents — Nvidia, AMD, Intel, Google, Amazon, and other major players. This dashboard focuses exclusively on the 43 private US startups trying to disrupt them, with detailed funding, valuation, investor, and technology data for each company.