Connect your agent to akta.pro and build company lists like this across sectors and geos.
Private AI chip companies, Series A or later, ranked by Q3 2026 primary-subject news volume: funding, products and services, revenue model, news signals and headcount growth for each.
| Company | Funding | Products & Services | Revenue Model | Signals (Q3 2026) | Headcount growth (Q3) |
|---|---|---|---|---|---|
| Etched.ai ↗ AI Inference Semiconductors 🇺🇸 San Jose, USA | Funding $2.13B Last raised $700.0M Aug 2026 · Series D | Sohu, a transformer-specific AI ASIC chip accelerating inference for LLMs, sparse MoEs and diffusion models, packaged into a rack-scale system, manufactured via TSMC. | Hardware Sales | 128 in Q3 View signals (63) → | +7% in Q3 532 → 568 employees |
| SambaNova ↗ AI Inference Infrastructure 🇺🇸 San Jose, USA | Funding $2.49B Last raised $1.0B Jul 2026 · Series F | Reconfigurable Dataflow Unit AI inference chips packaged into SambaRack systems, sold as SambaStack (on-prem), SambaCloud (inference-as-a-service) and SambaManaged for sovereign AI and enterprises. | Hardware Sales Usage Based Managed Services Subscription Recurring | 114 in Q3 View signals (67) → | +4% in Q3 398 → 414 employees |
| Rebellions ↗ AI Inference Accelerators 🇰🇷 Seongnam, KOR | Funding $918.0M Last raised $400.0M Mar 2026 | Chiplet-based NPU inference chips (Rebel100, ATOM-Max) paired with rack-scale systems (RebelServer, RebelRack) and an SDK, sold to hyperscalers, telecoms and governments. | Hardware Sales Licensing Royalties Managed Services Professional Services | 81 in Q3 View signals (53) → | +2% in Q3 319 → 324 employees |
| Positron ↗ AI Inference Hardware 🇺🇸 Reno, USA | Funding $1.18B Last raised $500.0M Sep 2026 · Series C | Purpose-built transformer inference hardware: the shipping Atlas FPGA appliance and announced Asimov custom silicon, with an OpenAI-API-compatible software stack for HuggingFace models. | n/d | 61 in Q3 View signals (14) → | +26% in Q3 94 → 118 employees |
| SiMa.ai ↗ Edge AI Silicon 🇺🇸 San Jose, USA | Funding $505.0M Last raised $150.0M Sep 2026 · Series C | Modalix Machine Learning SoC (50 TOPS under 10W) for edge Physical AI, plus the Palette software suite, sold to automotive, industrial, defense, drone and healthcare OEMs. | Hardware Sales Licensing Royalties | 55 in Q3 View signals (9) → | 0% in Q3 246 → 245 employees |
| OLIX ↗ Photonic AI Accelerators 🇬🇧 London, GBR | Funding $562.0M Last raised $312.0M Aug 2026 · Series B | Optical AI accelerators, including an Optical Tensor Processing Unit and DX-1 Decode Accelerator, combining photonics with SRAM for AI inference; first deliveries planned 2027. | Hardware Sales | 54 in Q3 View signals (22) → | +6% in Q3 141 → 150 employees |
| d-Matrix ↗ AI Inference Accelerators 🇺🇸 Santa Clara, USA | Funding $429.0M Last raised $275.0M Nov 2025 · Series C | Digital In-Memory Compute inference accelerators (Corsair PCIe cards, SquadRack rack systems) with the JetStream I/O accelerator, sold to hyperscalers, neoclouds and frontier AI labs. | Hardware Sales | 51 in Q3 View signals (18) → | -1% in Q3 321 → 318 employees |
| FuriosaAI ↗ AI Inference Accelerators 🇰🇷 Seoul, KOR | Funding $274.9M Last raised $8.6M Jun 2026 | Tensor Contraction Processor-based inference chips (RNGD, Warboy) bundled into servers and an SDK, targeting data centers, defense, and sovereign AI deployments in Korea and beyond. | Hardware Sales Subscription Recurring Licensing Royalties | 38 in Q3 View signals (27) → | +7% in Q3 213 → 228 employees |
| Axelera AI ↗ Edge AI Accelerators 🇳🇱 Eindhoven, NLD | Funding $738.0M Last raised $250.0M Feb 2026 · Series C | Digital In-Memory Computing AI Processing Units (Metis, Europa, Titania) on RISC-V, sold as cards and dev kits with the Voyager SDK for defense, industrial and robotics. | Hardware Sales Licensing Royalties Subscription Recurring | 37 in Q3 View signals (21) → | +3% in Q3 290 → 300 employees |
| Syntiant ↗ Edge AI Semiconductors 🇺🇸 Irvine, USA | Funding $106.4M Last raised $15.0M Dec 2024 · Series D | Ultra-low-power Neural Decision Processors, MEMS microphones and vibration sensors for always-on edge AI in earbuds, automotive, smart home, healthcare and defense devices. | Hardware Sales Licensing Royalties Professional Services Managed Services | 35 in Q3 View signals (23) → | 0% in Q3 302 → 302 employees |
| MatX ↗ AI Accelerator Chips 🇺🇸 Mountain View, USA | Funding $625.0M Last raised $500.0M Feb 2026 · Series B | MatX One, an LLM-first accelerator on a splittable systolic array with hybrid SRAM/HBM memory, built by ex-Google TPU engineers for frontier labs; shipping targeted 2027. | Hardware Sales | 30 in Q3 View signals (6) → | +11% in Q3 136 → 151 employees |
| Acrab ↗ Edge AI Silicon 🇸🇬 Singapore, SGP | Funding $700.0M Last raised $130.0M Jun 2026 · Series B | Agentic Edge SoC and compute platform delivering on-device LLM inference, long-term memory and cross-device orchestration, productized into creation, assistant and sales-enablement apps. | n/d | 28 in Q3 View signals (11) → | +15% in Q3 20 → 23 employees |
| Fractile ↗ AI Memory & Storage 🇬🇧 London, GBR | Funding $263.8M Last raised $220.0M May 2026 · Series B | SRAM-based in-memory compute inference processors interleaving memory and compute on-die, paired with a CUDA-rival software stack; first commercial chips targeted for 2027. | Hardware Sales Managed Services | 28 in Q3 View signals (19) → | +18% in Q3 118 → 139 employees |
| Lightmatter ↗ Silicon Photonics Interconnect Hardware 🇺🇸 Mountain View, USA | Funding $822.0M Last raised $400.0M Oct 2024 · Series D | Photonic interconnect chips (Passage, Guide light engines, vClick/eClick fiber attach) delivering 32 to 114 Tbps of optical I/O for hyperscale AI data centers. | Hardware Sales Professional Services Licensing Royalties | 21 in Q3 View signals (16) → | -2% in Q3 382 → 375 employees |
| Efficient Computer ↗ Edge AI Processors 🇺🇸 Pittsburgh, USA | Funding $379.8M Last raised $97.0M Sep 2026 · Series B | Electron E1, a general-purpose processor on a spatial dataflow Fabric architecture delivering up to 100x the energy efficiency of conventional CPUs for power-constrained edge AI. | Hardware Sales Licensing Royalties | 18 in Q3 View signals (6) → | +8% in Q3 66 → 71 employees |
An AI chip company designs processors built to train or run AI models faster or more efficiently than general-purpose GPUs and CPUs. The category is also called AI accelerators, AI ASICs or neural processing units.
Data center inference chips are the largest segment in the table: Etched.ai, Positron, d-Matrix, MatX, Rebellions, FuriosaAI and Fractile target the cost of serving large language models. SambaNova pairs its chips with full systems and a cloud service.
Edge AI processors from SiMa.ai, Axelera AI, Syntiant and Efficient Computer run models inside cameras, vehicles, robots and devices. Photonic and novel architectures from Lightmatter, OLIX and Acrab use light or new designs to move and compute data.
The category is defined against incumbent GPU and CPU makers such as Nvidia, AMD and Intel, whose general-purpose parts still power most AI workloads. Each startup here bets on a narrower design that wins on cost, power or latency.
Revenue models in the table follow hardware: 13 of 15 list hardware sales, often with licensing of chip IP or managed services, and only one prices on usage.
Companies are tagged against standard industry codes and against akta.pro's internal taxonomy, so the same set can be rebuilt by code, by segment, or by a combination of both.
NAICSIndustry codes are provided for cross-referencing against public filings and procurement records. akta.pro classification is independent of them, so a company may carry an internal segment without a matching NAICS or SIC registration.