It’s only Tuesday and AI chip startups have already soaked up $1.1B in funding
Summary
Tuesday saw roughly $1.1 billion of venture capital flow into AI chip startups as investors continue to back challengers to Nvidia. MatX led the wave with a $500m Series B to push its MatX One LLM-optimised accelerator, which combines SRAM for speed with HBM for KV cache storage and promises very high “FLOPS per mm2” and large-scale chip counts. Details remain thin but the firm claims high token throughput and an architecture designed to scale.
Dutch startup Axelera secured about $250m to advance low‑power RISC‑V AI accelerators aimed at edge workloads (Europa and Metis) while planning larger designs such as the Titania project with European partners. SambaNova picked up $350m and a multi‑year tie-up with Intel to ship its next‑gen SN50 dataflow accelerators (SoftBank will deploy them in Japan), showing diverse approaches — SRAM-heavy designs, RISC‑V, and dataflow — are all attracting capital.
Key Points
- Approximately $1.1bn in disclosed VC funding for AI chip startups on a single Tuesday.
- MatX raised $500m to build MatX One, an LLM-optimised accelerator blending SRAM speed with HBM for KV caches.
- Axelera secured $250m to scale low-power RISC‑V accelerators from edge use-cases towards datacentre workloads.
- SambaNova got $350m and an Intel collaboration to bring its SN50 dataflow accelerator to market, with SoftBank deployments planned.
- The funding underlines continuing investor appetite to fund alternatives to GPU dominance and a variety of technical strategies.
Why should I read this?
Quick and dirty: VCs are still chucking serious cash at AI silicon. If you care about who might displace GPUs, where compute innovation is going, or which startups to watch — this is worth a skim. We read the spin so you don’t have to.