Connect your agent to akta.pro and build company lists like this across sectors and geos.
Private semiconductor startup companies, Series A or later, ranked by Q3 2026 primary-subject news volume: funding, products & 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 Series D | Funding $2.13B Last raised $700.0M Aug 2026 · Series D | Develops Sohu, an application-specific chip built exclusively for transformer models, designed to run large language model inference faster and more cheaply than general-purpose GPUs, targeting data centers and AI labs. | Hardware Sales | 130 in Q3 View signals (69) → | +7% in Q3 532 → 568 employees |
| Rebellions ↗ AI Inference Accelerators (NPU Semiconductors) 🇰🇷 Seoul, KOR Series C | Funding $918.0M Last raised $400.0M Mar 2026 · Series C | Korean fabless chipmaker designing neural processing units for AI inference, including the ATOM and REBEL accelerators, along with a software stack, for data centers and enterprises. Merged with Sapeon Korea in 2024. | Hardware Sales Hardware Sales Licensing Royalties Managed Services Professional Services | 82 in Q3 View signals (54) → | +2% in Q3 319 → 324 employees |
| OLIX ↗ AI Semiconductor / Photonic AI Accelerators 🇬🇧 London, GBR Series B | Funding $562.0M Last raised $312.0M Aug 2026 · Series B | London-based developer of the Optical Tensor Processing Unit, an AI inference chip combining photonics with on-chip SRAM instead of high-bandwidth memory to cut the cost of serving models. Formerly known as Flux Computing. | Hardware Sales Hardware Sales | 55 in Q3 View signals (22) → | n/d |
| Lumilens ↗ Photonic Interconnects for AI Data Centers 🇺🇸 San Jose, USA Series C | Funding $836.5M Last raised $700.0M Aug 2026 · Series C | Designs and manufactures optical interconnects for AI data centers, including near-package optics, co-packaged optics and pluggable transceivers for scale-up and scale-out networks. Its first scale-out products are shipping into a hyperscaler's production data centers. | Hardware Sales Professional Services | 46 in Q3 View signals (34) → | +21% in Q3 136 → 164 employees |
| iPronics ↗ Silicon Photonics Networking 🇪🇸 Valencia, ESP Series B | Funding $164.2M Last raised $125.0M Sep 2026 · Series B | Valencia-based developer of programmable photonic processors on silicon photonics that reconfigure optical connections in software, aimed at making data center, AI cluster and telecom networks more flexible and energy-efficient through software-defined optics. | Hardware Sales Subscription Recurring | 40 in Q3 View signals (16) → | +13% in Q3 106 → 120 employees |
| FuriosaAI ↗ AI Inference Accelerators 🇰🇷 Seoul, KOR Series C | Funding $274.9M Last raised $8.6M Jun 2026 · Series C | Korean chip designer making RNGD, a power-efficient AI inference accelerator for large language models and multimodal workloads, designed for high performance per watt and sold as cards and servers for data centers and enterprises. | 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 Series C | Funding $738.0M Last raised $250.0M Feb 2026 · Series C | Eindhoven-based maker of Metis AI processing units for edge inference, delivering computer vision and generative AI performance in low-power cards and systems for industrial, retail, security and automotive customers, supported by its Voyager software development kit. | Hardware Sales Licensing Royalties Subscription Recurring Subscription Recurring | 37 in Q3 View signals (22) → | n/d |
| SiFive ↗ RISC-V Processor IP 🇺🇸 Santa Clara, USA Series G | Funding $765.5M Last raised $400.0M Apr 2026 · Series G | Founded by the creators of the RISC-V architecture, SiFive licenses RISC-V processor IP cores, from embedded to high-performance, to chipmakers, and develops AI-oriented designs such as its Intelligence series for data center and edge SoCs. | Licensing Royalties Hardware Sales | 30 in Q3 View signals (26) → | n/d |
| Ayar Labs ↗ Optical I/O semiconductors for AI infrastructure 🇺🇸 San Jose, USA Series E | Funding $1.25B Last raised $225.0M Sep 2026 · Series E | Develops optical I/O chiplets, including TeraPHY, and SuperNova laser light sources that replace electrical links between chips, increasing bandwidth and cutting power in AI scale-up networks and high-performance computing systems built by chipmakers and system builders. | Hardware Sales Licensing Royalties | 29 in Q3 View signals (20) → | n/d |
| Fractile ↗ 🇬🇧 London, GBR Series B | Funding $263.8M Last raised $220.0M May 2026 · Series B | London-based AI chip startup building hardware that performs computation inside memory to accelerate large language model inference with lower latency and power use than GPU-based systems, aimed at easing the memory-bandwidth bottleneck that limits inference speed. | Hardware Sales Managed Services | 28 in Q3 View signals (19) → | n/d |
| Velaura AI ↗ Semiconductor IP and Systems 🇺🇸 Santa Clara, USA Series C | Funding $424.0M Last raised $110.0M Aug 2026 · Series C | Santa Clara chip company, formerly Auradine, offering Titan Core, a silicon design and licensable IP platform with low-voltage libraries that cut power in AI accelerators, plus compute platforms for physical AI such as robots and drones. | Licensing Royalties Hardware Sales Hardware Sales | 26 in Q3 View signals (13) → | n/d |
| Xsight Labs ↗ Networking Semiconductors 🇮🇱 Kiryat Gat, ISR Series E | Funding $400.0M Last raised $300.0M Jul 2026 · Series E | Israeli fabless company based in Kiryat Gat designing Ethernet switch and DPU chips, including X-Series programmable switches and the E1 DPU, for hyperscale data centers, AI networking and edge infrastructure. | Hardware Sales Hardware Sales Hardware Sales | 23 in Q3 View signals (15) → | +3% in Q3 260 → 269 employees |
| Lightmatter ↗ Silicon Photonics Interconnect Hardware 🇺🇸 Mountain View, USA Series D | Funding $822.0M Last raised $400.0M Sep 2025 · Series D | Builds photonic interconnect for AI infrastructure, notably Passage, a 3D photonic engine for co-packaged optics that links large numbers of accelerators in AI clusters, along with Guide laser products, aimed at hyperscalers and chipmakers. | Hardware Sales Professional Services Licensing Royalties | 21 in Q3 View signals (16) → | -2% in Q3 382 → 375 employees |
| SiPearl ↗ High-Performance Server Processors 🇫🇷 Maisons-laffitte, FRA Series A | Funding $154.0M Last raised $37.5M Jul 2025 · Series A | French designer of Rhea, a high-performance, low-power Arm-based processor with high-bandwidth memory for European supercomputers and HPC and AI servers, developed within the European Processor Initiative and based in Maisons-Laffitte, near Paris. | Hardware Sales Subscription Recurring | 20 in Q3 View signals (12) → | -1% in Q3 168 → 166 employees |
| Celero Communications ↗ Coherent DSP Semiconductors 🇺🇸 Irvine, USA Series C | Funding $451.8M Last raised $275.0M Sep 2026 · Series C | Develops coherent DSP chips for AI infrastructure, including what it says is the industry's first 2nm coherent DSP, carrying high-bandwidth optical links between and inside data centers. Based in Irvine with design centers in Canada and Argentina. | Hardware Sales | 19 in Q3 View signals (1) → | +11% in Q3 128 → 142 employees |
| Efficient Computer ↗ Edge AI Processors 🇺🇸 Pittsburgh, USA Series B | Funding $379.8M Last raised $97.0M Sep 2026 · Series B | Pittsburgh-based developer of Fabric, an energy-efficient general-purpose processor architecture founded on Carnegie Mellon research, and its Electron E1 chip for edge AI and embedded devices where power budgets are tight, such as sensors, wearables and drones. | Hardware Sales Licensing Royalties | 18 in Q3 View signals (6) → | +8% in Q3 66 → 71 employees |
| Morse Micro ↗ Wireless IoT semiconductors 🇦🇺 Surry Hills, AUS Series C | Funding $205.8M Last raised $58.1M Sep 2025 · Series C | Australian fabless chipmaker of Wi-Fi HaLow (802.11ah) system-on-chips, which operate in sub-1 GHz bands to offer longer range and better penetration than conventional Wi-Fi for low-power IoT, industrial, smart city and consumer devices. | Hardware Sales Licensing Royalties One Time License | 17 in Q3 View signals (9) → | +1% in Q3 198 → 200 employees |
| Point2 Technology ↗ Data Center Interconnect Semiconductors 🇺🇸 San Jose, USA Series B | Funding $187.2M Last raised $80.0M Aug 2026 · Series B | Develops e-Tube interconnect technology, which sends data over plastic waveguides using radio-frequency signals, and related chips for short-reach links between servers, switches and accelerators in AI data center clusters, as an alternative to copper and optics. | Hardware Sales Licensing Royalties | 11 in Q3 View signals (7) → | +2% in Q3 55 → 56 employees |
| SemiDrive ↗ Automotive-grade semiconductors (cockpit, ADAS, gateway and zonal SoCs and MCUs) 🇨🇳 Nanjing, CHN Series C | Funding $654.1M Last raised $50.0M Sep 2026 · Series C | Nanjing-based designer of automotive-grade chips including cockpit, ADAS, gateway and zonal SoCs and MCUs, supplying automakers and Tier 1 suppliers with integrated vehicle computing platforms for next-generation electronic architectures, from the digital cockpit to driver assistance. | Hardware Sales Professional Services Licensing Royalties Hardware Sales | 9 in Q3 View signals (5) → | -2% in Q3 49 → 48 employees |
| Cornelis Networks ↗ HPC / AI Networking Infrastructure 🇺🇸 Wayne, USA Series C | Funding $615.3M Last raised $205.0M Sep 2026 · Series C | Provides high-performance networking for HPC and AI clusters, led by its CN5000 Omni-Path switches and adapters, designed for low latency and high message rates; spun out of Intel's Omni-Path business. | Hardware Sales Subscription Recurring | 8 in Q3 View signals (3) → | +2% in Q3 241 → 245 employees |
A semiconductor startup designs, and sometimes also fabricates or packages, integrated circuits or chip-adjacent hardware, rather than building software that runs on chips someone else makes. Most are fabless: they design the chip and license a foundry such as TSMC or Samsung to manufacture it, which keeps capital needs lower than an integrated device manufacturer that owns its own fab, but still far higher than a software company, since tape-out, packaging and first silicon all cost real money before a single unit ships.
The table splits into a few distinct segments. AI inference accelerator makers, including Etched.ai, Rebellions and FuriosaAI, design custom silicon to run trained AI models faster or cheaper than general-purpose GPUs. A photonics and optical interconnect cluster, spanning Lumilens, Ayar Labs, Lightmatter, iPronics and OLIX, uses light instead of electrical signals to move data between or within chips, aimed at the power and bandwidth limits general-purpose GPUs hit in large AI data centers. A semiconductor IP and design segment, led by SiFive's RISC-V processor cores and Velaura AI's low-power chip designs, sells the blueprint rather than the physical chip. A networking and interconnect silicon group, including Xsight Labs, Point2 Technology, Cornelis Networks and Celero Communications, builds the chips that move data between servers inside a data center. SemiDrive represents a smaller automotive silicon segment, building SoCs and MCUs for vehicle cockpits and driver-assistance systems.
Semiconductor startups differ from the incumbent chipmakers they are challenging, primarily Nvidia in AI accelerators and the legacy networking silicon vendors, in that most are still pre-revenue or early-revenue, financing years of design and tape-out cycles on venture capital rather than selling into an existing customer base. That is why valuations in this list can run well ahead of shipped product: 8 of 20 companies already carry $1B+ valuations despite most chips not yet reaching volume production, a bet that the architecture, not just the balance sheet, will out-perform the incumbent once silicon actually ships.
The table also shows a cohort still selling the traditional way: none of the 20 companies price on usage-based or consumption terms, the model common in AI software. Chips are sold as hardware, licensed as IP, or bundled into long-term supply contracts, such as Etched.ai's reported chip contracts ahead of its first shipments, rather than metered by the compute cycle the way an API-based AI company would charge.
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.