SUNMMIO

Positron AI, founded in 2023 in Reno, Nevada, USA, develops multimodal AI acceleration hardware.

Series C · $1.2B · 2026-09 · Valuation $5.0B

D-Matrix, founded in 2019 in Santa Clara, California, USA, develops AI inference chips and systems for data centers.

Series C · $450M · 2025-11 · Valuation $2.0B

DensityAI, founded by former leaders of Tesla's Dojo supercomputer program and based in Mountain View, CA, builds full-stack AI compute infrastructure — custom chips using 3D DRAM stacking and near-memory compute, plus hardware and software — for high-density AI data centers, now positioned as a fast inference solution for frontier models. It reportedly has a conditional chip-procurement agreement with AWS contingent on performance benchmarks.

Xingyun Integration, founded in 2023 in Beijing, China, is a GPU chip developer.

Undisclosed · 2026-09

Moxian Technology, founded in 2018 in Shenzhen, China, designs AI inference chips based on sparsification algorithms. By increasing the ratio of effective computation, it significantly reduces inference costs and power consumption, targeting cloud AI compute cards.

$139M

Founded in May 2026 in Shenzhen by Chen Peng, former GM of HiSilicon's Kunpeng processor business and former chip product-line president at Horizon Robotics, the company designs edge and on-device high-compute inference chips aimed at running very large models on a single chip.

Angel · $100M · 2026-09

Olix, founded in 2024 in London, develops photonic AI chips targeting inference workloads to reduce high-bandwidth memory constraints.

Series B · $532M · 2026-08 · Valuation $3.3B

Fractile, founded in 2022 in London, UK, develops advanced AI chips that run AI models 100x faster than traditional methods at 10x lower cost.

Series B · $236M · 2026-05

Founded in 2015 in Paris, designs AI inference chips for automotive, mobile, and data center applications, including its Jotunn8 (J8) processor.

Undisclosed · $46M · 2025-05

Founded in 2020 in Seongnam, South Korea, Rebellions designs energy-efficient AI inference chips.

Pre-IPO · $898M · 2026-03 · Valuation $2.3B

Spun out of SenseTime's large chip division and independently established in Hangzhou in 2025, Sunrise focuses on developing and commercializing AI inference GPUs and multimodal inference chips, with its flagship S1/S2 products already in mass production.

Undisclosed · $582M · 2026-04 · Valuation $1.5B

EUCLYD, founded in 2024 by Bernardo Kastrup and Atul Sinha at the High Tech Campus in Eindhoven, the Netherlands, builds ultra-efficient AI infrastructure for foundation models. It combines programmable ASIC compute (flagship product Craftwerk), the rack-scale CWS 32 system, processor-memory co-design and system-level optimization to reduce the power consumption, memory bandwidth requirements and infrastructure complexity of AI inference. Target customers are the enterprise, sovereign and hyperscale AI markets, addressing what the company calls the industry's 'AI efficiency wall'.

Series A · 2026-09