Technology
Edge Computing
Low-power processors optimized for AI inference and computing at the edge, in IoT and mobile devices.
107 companies
Explainer
Edge Computing explained
Edge computing means processing data close to where it is produced, on a device or a nearby server, instead of sending everything to a distant cloud data center. In chip terms, edge processors are small, low-power chips that run software and AI models inside phones, cameras, cars and sensors.
Edge AI chips often include a neural processing unit (NPU) that handles model inference within a tight power budget, from milliwatts to a few tens of watts. TinyML is the practice of fitting small models onto microcontrollers with only kilobytes or a few megabytes of memory.
How it differs from the usual approach
Cloud chips are judged on peak throughput and are cooled by data center systems. Edge chips are judged on energy, size, cost and response time, and they must work without a steady connection. Models are shrunk through quantization and pruning to fit. Running locally also keeps raw data on the device, which helps privacy and cuts network delays.
Where it is used
- Phones, for photo processing, voice recognition and on-device assistants.
- Cameras and doorbells that detect people or vehicles.
- Driver assistance and automated driving systems.
- Factory machines that spot defects or predict failures.
- Wearables and hearing aids with always-on sensing.
What to watch
- Microsoft set a 40 TOPS NPU requirement for its Copilot+ PC branding, announced in May 2024, which pushed laptop chip makers to add larger NPUs.
- Smaller language models are being designed to run on phones and laptops.
- Microcontroller vendors are adding NPUs so that small devices can run AI.
- Memory size and software tools, more than raw speed, often decide what a device can run.
Frequently asked questions
What is edge computing?
It is processing data close to its source, on the device or a nearby machine, instead of in a faraway data center. It reduces delay, network traffic and the amount of raw data sent to the cloud.
What is an edge AI chip?
It is a small, low-power processor, usually with an NPU, that runs AI models on a device. It is designed for tight limits on power, size and cost.
Is edge computing replacing the cloud?
No. Most systems use both. Models are often trained in the cloud and then run at the edge, with the cloud used for updates and heavy jobs.
- Kontron Kontron makes embedded computers, which are small purpose-built computers that sit inside machines rather than on a desk. Its range covers boards, computer-on-modules, industrial PCs and rugged VPX components for defense use. The company traces its origins to 1959, when Branco Weiss founded it in Zurich as a supplier of scientific instruments. It is now Kontron AG, headquartered in Linz, Austria, after S&T AG took control in 2017 and later adopted the Kontron name in 2022. Edge Computing · AI Accelerators —
- IoTFlows IoTFlows sells sensors and software that show manufacturers how their machines are being used. Its SenseAi device attaches to a machine without wiring or a connection to the machine's controller and reads vibration and sound to infer when the machine is running, idle or stopped. The software turns this into utilization, downtime and overall equipment effectiveness reports, a standard measure of factory productivity. The company also offers other plant tools, including laser-based production tracking. IoTFlows is based in Atlanta, was founded in 2020 by Mahmoud Parto and Ulises Carrasco, and joined Y Combinator in the Winter 2022 batch. Edge Computing YC W22
- LeapFive LeapFive designs RISC-V processors, an open instruction set that chip designers can use without paying a licence fee to a single owner. Its products include the NB2 system on chip, which combines CPU cores, a graphics unit and a neural processing unit for AI workloads, along with development boards and the F1 core aimed at server and AI use. NB2 was announced in 2022 and rated at 4 TOPS (trillions of operations per second) for AI. The company is based in China and describes itself as about five years old. AI Accelerators · Edge Computing —
- CGCrystal Group Crystal Group builds rugged servers, embedded computers, displays, networking and storage for demanding environments, along with AI and autonomy systems. Rugged servers are standard computing hardware re-engineered to tolerate vibration, shock, extreme temperatures and moisture, so they can run aboard ships, vehicles and aircraft. The company describes itself as employee-owned and lists its address as Hiawatha, Iowa. Its product range also includes energy-series and industrial commercial servers for less severe settings. Edge Computing · AI Accelerators —
- EFEfinix Efinix makes small, low-power FPGAs, which are chips whose circuits can be reconfigured after manufacture. Its Trion and Titanium families use what the company calls the Quantum architecture, a fabric of programmable tiles aimed at edge AI and wearable devices. The name refers to this chip design and has no connection to quantum computing. The company is based in Santa Clara, California. FPGAs · Edge Computing —
- ANAFLASH ANAFLASH is a semiconductor company working on low-power AI processing placed close to sensors. Its products include Logic-EFLASH, embedded flash memory that can be built on standard logic chip processes, Legato-Logic, a time-domain neural network processor design, and Reflex-CU, an always-on computing unit. Time-domain computing represents values as signal timing rather than voltage levels. The aim is to run pre-trained AI models on the device without cloud connectivity, using less power and with lower delay. Neuromorphic Computing · Edge Computing · Memory & Storage —
- GTGreenWaves Technologies GreenWaves Technologies is a fabless chip designer, meaning it designs processors and leaves manufacturing to outside foundries. Its GAP9 processor combines ten cores based on the open RISC-V instruction set with a neural network engine, built to run small AI models on battery-powered devices such as wireless earbuds and hearing products. The design uses GlobalFoundries 22nm FDX technology, which helps keep power draw low. The company was founded in 2014 and is based in Grenoble, France. Edge Computing · AI Accelerators · CPUs & RISC-V —
- Espressif Systems Shanghai microcontroller maker founded in 2008 and listed on the STAR Market. It reported in 2022 that it was moving to RISC-V, and its ESP32-P4, announced in 2023, has dual high-performance RISC-V cores at up to 400 MHz, a low-power core, MIPI camera and display interfaces and H.264 encoding. CPUs & RISC-V · Edge Computing —
- AMAmbiq Micro Austin company founded in 2010 as Cubiq Microchip, spun out of the University of Michigan, making ultra-low-power Arm-based microcontrollers and SoCs (Apollo line) using its subthreshold design method for wearables and edge AI. It listed on the NYSE as AMBQ in July 2025, selling 4 million shares at $24 to raise $96 million. CPUs & RISC-V · Edge Computing —
- SemiDrive SemiDrive designs automotive-grade chips, including smart cockpit processors for in-car displays and infotainment, gateway processors that route data between vehicle systems, and microcontrollers for vehicle control. It also sells chips for robotics and industrial use. Its customers include Chinese carmakers such as SAIC, Chery and Changan, as well as Nissan, Honda and Volkswagen. The company was founded in 2018 and opened its global headquarters in Beijing's economic development area in 2024, after earlier operations centred on Nanjing. Edge Computing · Custom Silicon Series B-III round (total $504M raised)
- Gemesys Bochum-based spin-off from Ruhr-University Bochum developing analog neural-network chips built on memristors, for training and inference on edge devices. It announced €8.6M in pre-seed funding on 19 November 2024, led by Amadeus APEX Technology Fund and Atlantic Labs, and had a working hardware demonstrator with silicon testing still to come. Analog Computing · Neuromorphic Computing · Edge Computing €8.6M pre-seed (Nov 2024)
- Secure-IC Secure-IC provides embedded security for chips and connected devices. Its Securyzr platform includes silicon security blocks such as physical unclonable functions, which derive a unique key from tiny manufacturing differences in each chip, and anti-tamper and cryptography cores. It also sells Laboryzr tools that test devices against side-channel attacks, which read secrets from power use or timing, and against fault injection. The company's website now states that Secure-IC is part of Cadence Design Systems, following an acquisition announced by Cadence. ASICs · Edge Computing —
- Axelera AI Axelera AI builds processors that run trained AI models, mainly for edge devices and servers. Its chips use digital in-memory computing, where arithmetic is carried out inside the memory array that stores the model weights, cutting the cost of moving data. They are paired with RISC-V processor cores. Its Europa chip is specified at up to 629 TOPS at INT8 precision, a measure of integer operations per second. Axelera was founded in 2021 and is based at the High Tech Campus in Eindhoven, the Netherlands. AI Accelerators · Edge Computing · In-Memory Computing Series C closed Feb 2026; $450M+ total raised
- Skymizer Taipei company founded in 2013 that licenses LPU (language processing unit) IP for LLM inference: EdgeThought for on-device use and HyperThought, an octa-core design on its LISA v3 instruction set. On 23 April 2026 it announced the HTX301 reference chip, with six chips and 384 GB on one PCIe card claimed to run 700B-parameter models at about 240 W. AI Accelerators · Edge Computing —
- Grayscale AI Grayscale AI is a London company that describes its work as neuromorphic computing for autonomous robots. Neuromorphic means the hardware and software are modelled loosely on how brains process signals, with small units that react to events instead of running a constant clock. The aim is to give robots low-power perception and control. The company is listed as based in London in the United Kingdom. Neuromorphic Computing · Edge Computing —
- MEMelexis Melexis is a Belgian company that designs sensor and driver chips, mainly for cars. Its parts include magnetic position sensors, which detect movement without touching, and chips that drive motors and lighting. It is a fabless designer, meaning other firms make its silicon. Its products are embedded systems components rather than computing processors. The company is headquartered in Ieper, Belgium, and its shares are listed on Euronext Brussels. Founding year and wafer sourcing details were not confirmed against a primary source in this review. Custom Silicon · ASICs · Edge Computing —
- nEye Systems nEye Systems makes programmable optical circuit switches built on silicon photonics, aimed at AI data centers. An optical circuit switch steers beams of light between fibers directly, so traffic does not need to be converted to electrical signals and back at each hop, which can reduce latency and power use in large GPU clusters. The company was founded in 2020 and is headquartered in Emeryville, California. Its technology traces to the lab of Professor Ming Wu at UC Berkeley. Photonic Computing · Edge Computing $80M Series C (April 2026, Sutter Hill-led, $152M total)
- Dover Microsystems Dover Microsystems licensed CoreGuard, a hardware security design that sits beside a processor and checks each instruction against security rules, so that memory errors and many software attacks are blocked as they happen rather than found later. It targets embedded and connected devices, including RISC-V cores. Dover was spun out of Draper in 2017, with Draper keeping a minority stake, and CoreGuard was announced that October. Qualcomm Ventures was among its early investors. Its current trading status could not be confirmed. ASICs · Edge Computing —
- Qualcomm Qualcomm designs wireless and computing chips and licenses wireless patents. Its chip business (QCT) sells Snapdragon processors, modems and connectivity products for smartphones, automotive systems and IoT devices, while its licensing business (QTL) covers 3G, 4G and 5G technology. Qualcomm is fabless, so outside foundries manufacture its chips. It supplies phone makers, carmakers and device makers rather than selling to consumers directly. The company was incorporated in 1985 and is based in San Diego, California. Custom Silicon · AI Accelerators · Edge Computing —
- RORoviero Roviero is a US startup that describes its work as neuromorphic processing for radar and sensing. Neuromorphic chips are built to work more like the brain, processing events as they happen rather than in fixed clock cycles, which can reduce power use for tasks such as interpreting radar returns. The aim is low-power AI for devices that must sense their surroundings. Details on products, customers and headquarters were not independently confirmed. Neuromorphic Computing · Edge Computing —
- Hesai Group Hesai makes lidar sensors, which measure distance by firing laser pulses and timing the reflections to build a three-dimensional map of surroundings. They are used in driver-assistance systems, robotaxis, robots and industrial equipment. Hesai was founded in 2014 by Li Yifan, Sun Kai and Xiang Shaoqing, and is headquartered in Shanghai. It is listed on the Nasdaq as HSAI and in Hong Kong as 2525. Customers include Li Auto, Aurora Innovation and Pony.ai. Lidar is a sensor business rather than a computing one. Edge Computing · Custom Silicon —
- Innodisk Innodisk makes embedded and industrial storage, DRAM modules and peripheral modules, and now positions itself as an edge AI provider that also sells cameras, networking and I/O modules, and edge AI systems built on NVIDIA, Qualcomm and Intel platforms. Industrial parts are built for harsh conditions and long product lifetimes rather than consumer use. The company was formally established in Taiwan in March 2005 and is based in Xizhi District, New Taipei City. Its group includes affiliate Aetina. Memory & Storage · Edge Computing —
- GTGyrfalcon Technology Gyrfalcon Technology is recorded as the maker of Lightspeeur AI accelerator chips, which are designed to run trained neural networks at low power on devices at the network edge instead of in a data centre. This is called inference. The company's website could not be reached during checking because of an expired security certificate, so its products, headquarters in Milpitas, California and 2017 founding year are taken from the existing record and remain unconfirmed. Edge Computing · AI Accelerators —
- ATAllwinner Technology Zhuhai fabless SoC designer founded in 2007 and listed on Shenzhen (300458) that makes Arm application processors (A-, H- and R-series). Its D1, announced April 2021 around a T-Head XuanTie C906 core, was described as the first mass-produced RISC-V application processor. CPUs & RISC-V · Edge Computing —
- Aspinity Pittsburgh-based company making analog machine-learning chips that run always-on sensor event detection at microwatt power. Its AML100 is in production, with the higher-performance AML200 following. A US$5M Series B in September 2023 included Korean automotive-chip distributor Unitrontech. Analog Computing · Edge Computing —
- KIKinara Kinara developed discrete neural processing units, separate AI chips that sit beside a main processor and handle inference for edge devices. Its Ara-2 is a second-generation part rated at up to 40 TOPS and aimed at generative AI, vision-language models and similar workloads. It connects over PCIe or USB. NXP Semiconductors completed its acquisition of Kinara on 27 October 2025 for $307 million in cash before closing adjustments, and the Ara products now form part of NXP's edge AI portfolio. Edge Computing · AI Accelerators $307M all-cash acquisition by NXP (closed Oct 27, 2025)
- SSSteradian Semiconductors Steradian Semiconductors was a Bangalore, India company that designed radar chips for vehicles. Radar sends out radio waves and measures their reflections to detect the distance and speed of objects, and imaging radar aims to give a sharper picture than conventional automotive radar. The company's chips were aimed at driver assistance systems. The record in this directory states that Renesas acquired Steradian, so it should be read as a former independent company, but that acquisition date has not been independently confirmed here. ASICs · Edge Computing —
- HRHorizon Robotics Horizon Robotics designs processors and software for driver assistance and automated driving, built to run neural networks inside the vehicle rather than in the cloud. Its chips use the company's own neural network accelerator design, which it calls a BPU. The Journey series is the vehicle product line, with Journey 6 the latest generation. Horizon was founded in 2015 and is based in Beijing, and its shares have traded on the Hong Kong Stock Exchange since October 2024 under the ticker 9660. Edge Computing · AI Accelerators IPO on Hong Kong Stock Exchange Oct 2024 ($696M raised)
- Expedera Expedera licenses neural processing unit (NPU) designs, meaning blocks of circuit design that chip makers build into their own processors to run AI models, rather than selling finished chips. Its Origin family and Origin Evolution lines target smartphones, automotive, edge devices and data centers. It describes a packet-based architecture that splits each network layer into small self-contained fragments that can run in parallel, instead of processing one layer at a time. Expedera is based in Santa Clara, California, and says its technology is field-proven in over 10 million consumer devices. Edge Computing · AI Accelerators —
- Quadric Quadric licenses processor designs to chip makers instead of selling chips. Its Chimera design is a general-purpose neural processing unit, or GPNPU, which runs AI inference and conventional signal processing and control code on a single core with one software toolchain. Most competing designs split these tasks across separate units. The multicore versions scale up to 864 TOPS, a measure of trillions of operations per second. Kyocera licensed Chimera in 2025 for office equipment chips. The company is based in Burlingame, California. Edge Computing · AI Accelerators $30M Series C (Jan 2026, BEENEXT led)
- Quintauris RISC-V company incorporated in Munich on 22 December 2023 by Bosch, Infineon, Nordic Semiconductor, NXP and Qualcomm. It builds reference architectures and compatible RISC-V platforms, starting with automotive and later IoT and consumer; its RT-Europa platform is described as the first RISC-V real-time platform for automotive. CEO is Alexander Kocher. CPUs & RISC-V · Edge Computing —
- MTMindgrove Technologies Mindgrove Technologies is a fabless chip company based in Chennai, meaning it designs chips but does not make them. It was founded in 2021 and was incubated at IIT Madras. Its chips build on the Shakti processor, an open-source RISC-V design from IIT Madras, where RISC-V is a freely licensable instruction set. Its Secure IoT chip, launched in May 2024, runs at 700 MHz and includes an AES-256 accelerator, one-time programmable memory and a hardware random number generator. It targets wearables, smart meters and connected home devices. Edge Computing · Custom Silicon · CPUs & RISC-V $8M Series A (December 2024)
- ECESWIN Computing Beijing RISC-V chip company, part of Beijing ESWIN Technology Group. Its EIC7700X edge SoC pairs four SiFive P550 cores with an Imagination GPU and an in-house 19.95 INT8 TOPS NPU, and it showed automotive chips on its own RISC-V cores in April 2026. Its site lists a Hong Kong Stock Exchange listing on 9 October 2026. CPUs & RISC-V · Edge Computing —
- Bosch Semiconductors This record covers the semiconductor business of Robert Bosch GmbH, a German engineering and technology group, rather than a separate company. The business develops and makes chips for vehicles and other products, including MEMS sensors (tiny mechanical sensing structures built on silicon) and power devices. In 2023 Bosch acquired TSI Semiconductors, a California maker of silicon carbide power devices. The Bosch group was founded in Stuttgart on 15 November 1886 and is 94 percent owned by the charitable Robert Bosch Stiftung. ASICs · Custom Silicon · Edge Computing —
- NXP Semiconductors NXP Semiconductors makes chips for automotive, industrial and mobile uses, including digital signal processors, which are specialised for fast repetitive arithmetic on audio and sensor data. Its SAF9xxx audio processors serve car infotainment systems. NXP began in 2006 as a spin-off of Philips, whose semiconductor unit dates to the 1950s, and in December 2015 it completed a merger with Freescale, a company with Motorola roots. NXP is headquartered at the High Tech Campus in Eindhoven, Netherlands. DSPs · Edge Computing · CPUs & RISC-V —
- ISIngenic Semiconductor Ingenic Semiconductor designs processors and system-on-chip products, mainly for embedded and consumer devices. Its XBurst CPU cores are based on the MIPS instruction set, and the second generation, XBurst2, appeared in 2020 in the X2000. The company licenses MIPS, designs its own core microarchitecture and has chips made by outside foundries. It was founded in 2005 by Liu Qiang, who is chairman, and is headquartered in Beijing, China. A 2011 prospectus covered its stock market listing. CPUs & RISC-V · Edge Computing —
- ALAlphaICs AlphaICs is an Indian chip company developing processors for running trained AI models at the edge, meaning on devices near the data source rather than in a remote data centre. Its architecture is branded the Real AI Processor, or RAP. The company is based in Bangalore, India. Founding year, funding and customer details could not be independently confirmed from the sources checked, so none are stated here. Edge Computing · AI Accelerators —
- Collins Aerospace Collins Aerospace supplies systems and components for commercial and business aviation, military and defense, helicopters and space. Its products include avionics, which are the electronics that fly and monitor an aircraft, and mission computers that run onboard software for flight and military tasks. It was formed in 2018 by the merger of Rockwell Collins and UTC Aerospace Systems and is a subsidiary of RTX Corporation. It is headquartered in Charlotte, North Carolina. It is included here for its aircraft computing rather than as a chip designer. ASICs · Edge Computing —
- Mentium Technologies Mentium Technologies designs AI co-processors meant to keep working in the radiation found in orbit, where charged particles can upset ordinary chips. Its Luna family includes the Luna-R1 and Luna-R2, and it describes the work as neuromorphic, meaning loosely modelled on how the brain processes signals. The company is based in the Santa Barbara area of California, and most listings give Goleta. It holds a NASA Small Business Innovation Research Phase II contract for a chip for radiation environments, and Luna-R1 has been selected for the ET-01 satellite constellation mission. Edge Computing · AI Accelerators —
- Bestechnic Bestechnic (Shanghai) designs smart audio system-on-chip products, which combine a processor, radio and audio circuits on one die. Its chips are used in Bluetooth earphones, Type-C earphones and smart speakers, and press coverage names Google Pixel Buds as one customer. It is a fabless company, so it designs chips and relies on outside foundries to make them. Bestechnic was founded in Shanghai in June 2015 and listed on the Shanghai Stock Exchange STAR market in December 2020 under code 688608. Custom Silicon · Edge Computing —
- Egis Technology Egis Technology is a Taiwanese fabless semiconductor company known for capacitive and optical fingerprint sensors, including under-display and side-mount designs for phones. It also offers FIDO-based authentication products, ASIC design services and silicon IP for AI, high-performance computing and automotive uses through its wholly owned subsidiary InPsytech. In the supply chain it supplies sensing and interface chips to device makers rather than manufacturing wafers itself. The company is listed on the Taipei Exchange under the code 6462. ASICs · Edge Computing —
- Realtek Realtek is a fabless chip company, meaning it designs integrated circuits and outsources their manufacture. Its products include Ethernet and Wi-Fi network controllers, audio codecs, which convert between digital data and sound, and display controllers for monitors and televisions. Its low-cost parts appear in many personal computers and consumer devices. Realtek was founded in October 1987, is headquartered in Hsinchu Science Park in Taiwan, and has been listed on the Taiwan Stock Exchange since 1998. Custom Silicon · Edge Computing —
- Nordic Semiconductor Nordic Semiconductor designs low-power wireless chips for connected devices. Its products include Bluetooth Low Energy system-on-chips in the nRF52, nRF53 and nRF54 series, the nRF91 cellular IoT module for LTE-M and NB-IoT networks, the nRF70 Wi-Fi companion chip and nPM power management chips. It is fabless, meaning it designs chips and has others manufacture them. The company was founded in Trondheim, Norway, in 1983 as Nordic VLSI and has been listed on the Oslo Stock Exchange since 1996. Custom Silicon · Edge Computing —
- Intellifusion Intellifusion is a Shenzhen joint-stock company that calls itself an AI inference chip maker, where inference means running a trained AI model rather than training one. Its chips come in three lines, DeepEdge for edge gateways and boxes, DeepVerse for large-model inference in the cloud, and DeepXBot for robots, drones and vehicles. It also sells AI cameras, wearable modules, compute services and IP licensing, and offers city solutions for policing, transportation and urban governance. Its DeepEdge10 chip has been adapted to the DeepSeek R1 model series. AI Accelerators · Edge Computing —
- Neurobus Neurobus is a French company working on onboard AI for space and defence. It applies neuromorphic processors, which handle data as sparse events and use little power, so that satellites can analyse imagery and signals in orbit instead of sending everything to the ground. Its stack includes hardware integration, a neuromorphic operating system called NeurOS, and real-time perception algorithms. The company was founded by Florian Corgnou, who gives 2023 as the founding year, although one database lists 2022. Sources differ on whether it is based in Paris or Toulouse. Neuromorphic Computing · Edge Computing €5M-€10M Seed (Q4 2025/Q1 2026 in progress)
- HOHOUMO.AI Houmo AI, known in China as Houmo Intelligent, designs AI chips that use compute-in-memory, in which calculations happen inside or next to the memory cells instead of moving data to a separate processor. This aims to cut power use. Its first product, the Hongtu H30, was launched in 2023 for intelligent driving and is built on a 12nm process. The company was founded in 2020 in Nanjing, Jiangsu province, and is led by chief executive Wu Qiang. AI Accelerators · Edge Computing Strategic round of 'hundreds of millions of yuan' (2024, led by China Mobile Capital)
- Aitech Defense Systems Aitech Defense Systems makes rugged, military and space-qualified computing products built from commercial off-the-shelf parts, meaning standard components adapted for harsh conditions. Its customers are in defense, aerospace and space programs. The company was founded in 1983 and is headquartered in Chatsworth, California, and its Israeli operation, Aitech Systems Ltd, is a subsidiary based in Herzliya. AI Accelerators · Edge Computing —
- femtoAI femtoAI, formerly Femtosense, designs the Sparse Processing Unit (SPU), a low-power AI chip for speech, audio and other one-dimensional sensor data in devices such as hearing aids and earbuds. Sparse processing means the chip skips the zero-valued weights and activations in a neural network, which saves memory and energy. The SPU-001 has 1 MB of on-chip memory and is made on a 22 nanometer process, and the company says such workloads can run below one milliwatt. It was founded in 2018 and is based in San Bruno, California. Edge Computing · AI Accelerators —
- HMHopeRF Microelectronics HopeRF designs wireless radio chips and ready-made radio modules, mainly for sub-1GHz bands, along with sensor chips. Its range includes LoRa modules such as the RFM95, which use a long-range, low-power radio technique suited to sensors and meters that send small amounts of data over long distances. It sells components for other makers to build into products rather than finished consumer devices. The company says it was registered in Shenzhen, China in 2004 and reports more than 2,500 cooperating customers. Edge Computing · ASICs —
- Eta Compute Eta Compute developed ultra-low-power chips for running machine learning on small battery-powered devices, including the ECM3532 processor. Running models on the device itself, rather than sending data to a server, is called edge or tinyML computing. The company was based in Westlake Village, California. Founding year, ownership and whether it was acquired could not be independently confirmed from the sources checked, so the earlier acquisition claim is left out. Edge Computing · AI Accelerators —
- Nuclei System Technology Chinese RISC-V core IP vendor founded in 2018 by Bob Hu, creator of the Hummingbird E203 open-source core. It sells the N-series cores plus NS security and NA functional-safety variants; reports cite over 150 customers and investment from CETC and Xiaomi, and a later round led by China Fortune Innovation Capital. CPUs & RISC-V · Edge Computing —
- Atmosic Technologies Atmosic is a fabless chip company that designs very low power Bluetooth system on chip devices for sensors, tags and other small connected products. Its chips can run on energy harvested from radio waves, light or motion, which lets some devices work with a much smaller battery or none at all. Atmosic was founded in 2016 and is based in the San Jose area of California. Its current product lines include the ATM33 and ATM34 series, and the ATM34 family supports Bluetooth 6.0. Edge Computing · ASICs $40M Series D (Mar 2025)
- PTPOLYN Technology NASP neuromorphic analog signal processors; world-first silicon-proven NASP chip announced Oct 2025; demoed ~34µW voice-detection chip at CES 2026. Neuromorphic Computing · Analog Computing · Edge Computing —
- VEVersaLogic VersaLogic makes embedded computers for demanding industrial and defense uses, including PC/104 single-board computers, embedded processing units and edge AI computers. A single-board computer places the processor, memory and input and output on one small card, so it can be built into a larger machine such as a drone, vehicle or instrument. PC/104 is a compact stacking format for such boards. The company describes a 50-year history in this field and offers US-based technical support. Edge Computing · AI Accelerators —
- Embedder Embedder makes an AI coding agent for firmware, the low-level software that runs directly on microcontrollers and other embedded chips. Unlike general coding assistants, it is built to be hardware-aware, meaning it works with the specific chip and board in use and can write, test and debug code against real devices. It has a partnership with SEGGER that lets the agent drive J-Link debug probes to flash builds, set breakpoints and read registers. Embedder is based in San Francisco, was founded in 2025 by Ethan Gibbs and Bob Wei, and went through Y Combinator in the Summer 2025 batch. Edge Computing YC S25; backed by CRV, Box Group, Emerson Collective
- XMXMOS XMOS designs xcore multicore microcontrollers, small processors that put several cores on one chip so that audio, control and AI tasks can run together without extra dedicated hardware. Its products cover USB and multichannel audio, voice processing such as the XVF3800 and xcore.ai for edge AI, and in 2025 it began calling the line a Generative System-on-Chip. XMOS is a private company founded in Bristol, UK, in 2005, where it is still headquartered, and its investors include Infineon and Bosch. Edge Computing · DSPs —
- LDLeonardo DRS Leonardo DRS supplies defense electronics to the US military and allies, including sensors, networked computing and power systems. It is majority owned by the Italian group Leonardo S.p.A., which bought the business in 2008 while it was known as DRS Technologies, and it trades on Nasdaq as DRS. Two engineers founded it in 1968 as Diagnostic/Retrieval Systems, and it is headquartered in Arlington, Virginia. In 2022 it agreed to combine with the US subsidiary of the Israeli firm RADA Electronic Industries. AI Accelerators · Edge Computing —
- MEMediaTek MediaTek is a fabless chip company, meaning it designs processors but leaves manufacturing to outside foundries. It is known for smartphone system-on-chip processors such as the Dimensity series, and also makes modems, Wi-Fi chips and chips for televisions. A system-on-chip puts the processor cores, graphics, modem and other parts on one piece of silicon. MediaTek was spun off from United Microelectronics Corporation in 1997, is headquartered in Hsinchu, Taiwan, and has been listed on the Taiwan Stock Exchange since 2001. Edge Computing · AI Accelerators —
- SYSyntiant Syntiant, based in Irvine, California, makes Neural Decision Processors, very low power chips that run neural networks for always-on tasks such as voice commands and sensor monitoring. Its Core 2 design couples memory with multiply-accumulate units, the circuits that perform the basic arithmetic of neural networks, so less data has to travel between them. The NDP200, introduced in September 2021, was its first chip aimed at vision tasks such as person detection and object classification. Syntiant says it runs these workloads at about a milliwatt or two, a vendor claim. Edge Computing · Neuromorphic Computing · In-Memory Computing Equity from Khazanah Nasional & Boardman Bay (Dec 2024) financing $150M Knowles CMM acquisition
- Lumotive Lumotive designs chips that steer a laser beam without moving parts, which lets a LiDAR sensor scan its surroundings electronically. Its Light Control Metasurface uses a CMOS chip patterned with tiny liquid-crystal structures that bend light, so scan patterns can be changed in software. Lumotive sells the beam-steering chips to sensor makers rather than building complete LiDAR units. Its LM10 chip is used in the Hokuyo YLM-10LX 3D sensor. The company is based in Redmond, Washington and came out of Intellectual Ventures, with sources dating its founding to 2017 or 2018. Photonic Computing · Edge Computing $59M Series B (July 2025)
- Silicom Silicom designs and sells server network adapters and connectivity products, including FPGA-based SmartNICs, for equipment makers, cloud providers and telecom operators. Its cards are built into customer systems such as security, WAN optimization and network appliances, usually through a design win model. The company was founded in 1987, is based in Kfar Sava, Israel, and is part of the RAD Group. Its shares are listed on Nasdaq under SILC and on the Tel Aviv Stock Exchange. FPGAs · Edge Computing · Interconnect & Networking —
- Black Sesame Technologies Black Sesame Technologies designs system-on-chip processors for driver assistance and autonomous driving, along with the algorithms and toolchain that run on them. Its Huashan series includes the A1000 and A2000 families, and its Wudang series includes the C1236 and C1296 chips for cross-domain vehicle computing, where one chip handles several functions. It also licenses image signal processor and neural processing unit IP. The company was founded in 2016 and listed on the Hong Kong Stock Exchange under code 2533 in August 2024. Edge Computing · AI Accelerators —
- DEEPX DEEPX is a South Korean fabless company that designs NPUs, chips specialised for running trained AI models efficiently on local devices such as cameras, robots and appliances. Its first product, the DX-M1, is built on Samsung Foundry's 5nm process, and the next chip, the DX-M2, is planned on Samsung's 2nm process. The company was founded in December 2018 and is based in Pangyo, Seongnam, in Gyeonggi Province near Seoul, rather than in Seoul itself. Edge Computing · AI Accelerators Series C led by SkyLake Equity Partners (2025); ~$183M total
- Alif Semiconductor Fabless maker of Arm Cortex-M55 microcontrollers with Ethos-U55 NPUs (Ensemble) and a wireless variant (Balletto). It opened a European R&D site in Oulu, Finland in May 2025. On 10 September 2026 Analog Devices announced an agreement to acquire Alif for $1.35 billion plus up to $200 million contingent, closing expected before end of 2026. CPUs & RISC-V · Edge Computing —
- Himax Technologies Himax Technologies is a fabless chip company, meaning it designs semiconductors and outsources manufacturing. It makes display driver ICs and timing controllers, which send image data to screens, along with touch controllers, CMOS image sensors, LCoS microdisplays for augmented reality, and WiseEye, a low-power AI sensing product that processes images on the device itself. Himax was founded in 2001 and is headquartered in Tainan, Taiwan. Its shares trade on the Nasdaq under the ticker HIMX. Custom Silicon · Edge Computing —
- Sequans Communications Sequans is a fabless semiconductor company that designs chips and modules for 4G and 5G cellular IoT devices. Its portfolio includes baseband, RF transceiver and highly integrated SoC products, among them the Monarch LTE-M and NB-IoT platform and the Calliope Cat 1 platform. The company began with WiMAX and moved to LTE in 2009. Sequans was incorporated in 2003, is based in Paris, France, and its American depositary shares trade on the NYSE under SQNS. ASICs · Edge Computing —
- u-blox u-blox supplies positioning and wireless communication products, including GNSS chips and modules for satellite navigation, cellular modules, and Bluetooth and Wi-Fi modules for IoT and machine-to-machine devices. GNSS is the general term for satellite systems such as GPS, Galileo and BeiDou. It sells to automotive and industrial customers rather than to computing buyers. The company was founded in 1997 as a spin-off from ETH Zurich and is headquartered in Thalwil, Switzerland. It has been listed on the SIX Swiss Exchange since October 2007, and Advent International has agreed to acquire it. Custom Silicon · Edge Computing —
- VSORA VSORA is a French fabless semiconductor company that designs chips for AI inference, which means running trained models rather than training them. Its Jotunn8 chip is built on TSMC's 5 nm process with advanced multi-chip packaging, and the company says it targets data center workloads at lower power than GPUs. These performance claims are the company's own and are not independently benchmarked. VSORA is based in Paris and was founded in 2015 by Khaled Maalej, who earlier co-founded the mobile TV chip maker DiBcom. AI Accelerators · Edge Computing $46M (Apr 2025, led by Otium with EIC Fund)
- Apple (Neural Engine) The Neural Engine is a block of silicon inside Apple's own chips, built to run machine learning tasks on the device rather than in a data centre. Such blocks are often called NPUs, short for neural processing units, and they handle the repetitive matrix arithmetic of neural networks using far less power than a general processor. Apple first included one in the A11 Bionic chip in 2017, which powered the iPhone 8, iPhone 8 Plus and iPhone X. Later versions appear in Apple's A-series and M-series chips. Apple is headquartered in Cupertino, California. Custom Silicon · Edge Computing —
- Blaize Blaize is a fabless semiconductor company that builds programmable AI inference chips and software for edge and hybrid AI deployments. Its chips use a Graph Streaming Processor architecture, and its AI Studio software is a low-code tool for building and deploying models. The company targets workloads such as video analytics, smart infrastructure and defense, where power efficiency matters. Blaize is based in El Dorado Hills, California, and became publicly listed on Nasdaq under the ticker BZAI in January 2025 through a business combination. Edge Computing $35M registered equity offering Q1 2026
- Telechips Seongnam-based fabless designer (KOSDAQ: 054450) of Arm-based automotive application processors: Dolphin+ (Cortex-A53/R5), Dolphin3 and Dolphin5 (Cortex-A76 with Mali-G78AE GPU and Ethos-N78 NPU, built with Samsung Foundry). 2024 revenue was about 186.6 billion won. CPUs & RISC-V · Edge Computing —
- ADLINK Technology ADLINK Technology makes embedded computing hardware, including computer-on-modules, which are small boards that hold a processor and memory and plug into a customer's own device, as well as industrial motherboards, data acquisition modules and complete systems. It sells mainly to equipment makers and system integrators for edge uses such as factory automation, where AI runs on the device rather than in a data centre. ADLINK was founded in 1995, is headquartered in Taipei, and has been listed on the Taiwan stock exchange under code 6166. FPGAs · Edge Computing —
- Ambient Scientific Edge AI chip company whose GPX10 Pro is a programmable always-on AI processor for wearables, hearables and sensors; GPX40, GPX80 and GPX Ultra are listed on its site. On 10 March 2026 it announced the MAI women's safety wearable built on its chip, and in June 2026 it outlined plans to scale the architecture to laptops, servers and robotics. AI Accelerators · Edge Computing —
- SiMa.ai SiMa.ai makes machine learning system-on-chip products for AI at the edge, meaning devices that process data locally such as robots, cameras and vehicles. Its MLSoC puts the neural accelerator, CPU cores and vision processing on one chip, so a whole application can run there and not just the neural network portion. The second generation, Modalix, targets multi-modal and generative AI and comes in several performance tiers. The company was founded in 2018 by Krishna Rangasayee, a former Groq executive, and is based in San Jose, California. Edge Computing · AI Accelerators $85M (Maverick Capital, 2025) + Micron strategic investment (April 2026)
- PEPerceive Perceive develops low-power chips for running neural networks on devices at the edge, with its Ergo processors aimed at tasks such as audio and video analysis without sending data to the cloud. The company was founded in 2018 and was formed within Xperi, the licensing and technology group. It is based in San Jose, California. Edge chips of this kind aim to keep power use low enough for battery-powered products. Edge Computing · AI Accelerators —
- Silicon Labs Silicon Labs is a fabless company that designs microcontrollers, wireless system-on-chips and modules, and sensors for Internet of Things devices. Its parts support wireless protocols such as Bluetooth, Zigbee, Z-Wave, Thread, Wi-Fi and Matter, and it also supplies software tools such as the Simplicity Studio development platform. The company is focused on low-power connected devices rather than data center compute. It was founded in 1996 by Nav Sooch, Dave Welland and Jeff Scott and is based in Austin, Texas. Texas Instruments announced in February 2026 that it intends to buy the company for $7.5 billion. Custom Silicon · Edge Computing —
- Cortus French fabless company near Montpellier selling RISC-V microcontrollers and AI inference chips, mainly for automotive and avionics. In February 2025 it reported first silicon of the MINERVA 64-bit out-of-order RISC-V processor, made on GlobalFoundries 22FDX, booting Linux. Its ULYSS family targets ISO 26262 automotive use; a partnership with CTAG on AUTOSAR was announced in February 2026. CPUs & RISC-V · Edge Computing —
- Napatech Napatech makes SmartNICs and data processing units, which are network cards with their own programmable chips so that packet capture, filtering and monitoring happen on the card instead of the server processor. Its products are built on FPGAs, chips that can be reconfigured after manufacture. The company is headquartered in Søborg, north of Copenhagen, Denmark, and its shares trade on the Oslo Stock Exchange under the ticker NAPA. Napatech states a founding year of 2001 on its own site, while other sources give 2003. FPGAs · Edge Computing · Interconnect & Networking —
- Seiko Epson Semiconductor This record covers the semiconductor activities of Seiko Epson Corporation, a Japanese group better known for printers, projectors and watches. Its chip business has centred on low-power microcontrollers, which are small processors built into devices, along with display driver chips and timing devices such as crystal oscillators. The emphasis on very low power draw suits wearables and sensors rather than data centre computing. Seiko Epson is headquartered in Suwa, Nagano, and the chip operation is part of the group rather than a separate listed company. Custom Silicon · Edge Computing —
- DEDeGirum DeGirum makes ORCA, an accelerator chip for running trained neural networks on edge devices such as cameras and small computers, together with PySDK, a Python toolkit meant to work across different hardware. It first presented ORCA publicly in late 2021 after operating quietly. The company also runs a cloud service that lets developers test models on its hardware before buying. DeGirum was founded in 2017 and is generally listed as based in Santa Clara, California. AI Accelerators · Edge Computing · ASICs $42M
- Recogni Recogni began as a developer of AI chips for vision tasks such as automotive perception, and in September 2025 it renamed itself Tensordyne. The company now targets generative AI inference, meaning running trained models to answer requests. Its approach uses a logarithmic number system, which turns multiplications into additions, since a multiply costs far more silicon area and energy than an add. Tensordyne says its first chip is built on TSMC 3nm and co-designed with Broadcom. Reuters describes it as based in Sunnyvale, California, while the record lists San Jose. AI Accelerators · Edge Computing $102M Series C
- Nordic ID Nordic ID is a Finnish company that makes RFID tracking solutions, which use radio tags and readers to locate and count items without line of sight. Its work covers handheld readers and related software for inventory, retail and asset tracking rather than chip design. It began in 1986 as a small team focused on item tracing in Finland. The company joined the Brady Corporation group in 2021 and was rebranded as Nordic ID Technology in 2024. It sits outside compute hardware and is included as an edge data capture business. Custom Silicon · Edge Computing —
- Mobilint Mobilint is a South Korean company that designs neural processing units, or NPUs, which are chips specialised for running trained AI models efficiently on local devices. It has two main products. ARIES is a higher-power accelerator offered on PCIe cards and in a compact AI box, and REGULUS is a low-power system on chip for embedded devices that combines an NPU, CPU and image processor. Mobilint was founded in 2019 and is based in South Korea. Edge Computing · AI Accelerators 70B KRW (~$46M) Series C (Apr 2026)
- InCore Semiconductor InCore Semiconductors designs processor cores based on RISC-V, an open instruction set that anyone can use without paying a license fee, and licenses them to chip makers. Its cores include Azurite, a small low-power embedded design, and Calcite, which targets higher performance. The company grew out of the Shakti processor project at IIT Madras and was founded in Chennai in 2018. It raised $3 million in seed funding from Peak XV in 2023. ASICs · CPUs & RISC-V · Custom Silicon $3M seed (Peak XV)
- Systel Systel designs and builds rugged computers for defense and industrial customers, sold as small embedded units under the STRIKE name and as rackmount servers under the CHARGE name. Rugged here means built to military specifications for shock, vibration, heat and dust, so the machines keep running in vehicles, aircraft and field shelters rather than in a data centre. Its newer Hawk-Strike AI platform is built on the NVIDIA Jetson AGX Thor module for edge AI. Systel has operated since 1988 and builds its products at its headquarters in Sugar Land, Texas. AI Accelerators · Edge Computing —
- Rockchip Fuzhou SoC designer founded in 2001 and listed on the Shanghai exchange (603893) that makes Arm application processors from the RK3399 to the 8-core RK3588 and RK3576. Preliminary specs for the RK3668 (10 Cortex-A730/A530 cores, 16 TOPS NPU) were announced at RKDC 2025. CPUs & RISC-V · Edge Computing —
- SPSpacemiT SpacemiT designs processors based on RISC-V, an open instruction set, aimed mainly at AI workloads. Its products include the Key Stone K1 eight-core chip, the VitalStone V100 server processor and the K3 chip built on its X100 core. It also sells Muse-branded consumer devices, including a laptop running its Bianbu operating system. SpacemiT is privately held and based in Hangzhou, China. ASICs · AI Accelerators · Edge Computing $86M Series B (2025); Series A+ in hundreds of millions CNY
- PQShield PQShield sells post-quantum cryptography, which means encryption designed to resist attacks from future quantum computers. Its range includes software libraries for small embedded devices and OpenSSL applications, secure coprocessors, a root-of-trust system, and hardware accelerators that chip designers license. Unlike many security vendors, it supplies both the hardware blocks and the software, along with advisory services. PQShield was founded in 2018 as a spin-out from the University of Oxford and is headquartered in Oxford, United Kingdom. ASICs · Edge Computing $37M Series B (June 2024)
- Bouffalo Lab Bouffalo Lab is a fabless chip designer (it designs chips but does not own factories) that makes low-power wireless system-on-chip products for smart IoT and edge devices. Its product families combine Wi-Fi, Bluetooth Low Energy and Zigbee radios with processors, and also cover Matter, Thread and Bluetooth Mesh standards. It is known for building these parts around the open RISC-V instruction set. The company was established in Nanjing, China, in 2016 and has its headquarters there, in the Jiangbei New Area. CPUs & RISC-V · Edge Computing —
- Ambarella Ambarella designs AI vision processors for edge devices, meaning chips that analyse video inside the camera or vehicle instead of sending it to a data centre. Its CVflow architecture combines image processing, video encoding and neural network acceleration on one system-on-chip. The company sells into automotive driver assistance, security cameras, consumer devices, and industrial and robotics products. It is a fabless company, so it designs chips and relies on outside foundries to make them. Ambarella is headquartered in Santa Clara, California, and trades on Nasdaq as AMBA. Edge Computing · AI Accelerators —
- Lattice Semiconductor Lattice Semiconductor designs low-power field-programmable gate arrays, or FPGAs, which are chips whose logic the customer can rewire after manufacture. Its Nexus platform covers small devices and Avant covers mid-range ones, and it also sells design software such as Radiant and Propel. The focus is small, power-frugal chips for edge uses such as sensing, vision and control rather than large data-centre parts. Founded in 1983, the company is headquartered in Hillsboro, Oregon, and trades on Nasdaq under the symbol LSCC. FPGAs · Edge Computing —
- Kneron Kneron designs neural processing units, chips that run AI models efficiently on devices rather than in the cloud. Its KL730, announced in August 2023, is an automotive-grade chip that combines a fourth-generation NPU with an image signal processor and Arm Cortex-A55 CPU cores, and is built on TSMC 12-nanometre technology. Target uses include driver assistance, medical devices and industrial automation. Kneron was founded in 2015 by Albert Liu, who is chief executive, and is headquartered in San Diego with a development team in Taipei. Edge Computing · AI Accelerators $2M NTDP grant (Saudi Arabia, 2025); pursuing Nasdaq SPAC listing at $1B valuation
- GMGrAI Matter Labs GrAI Matter Labs developed NeuronFlow, a neuromorphic processor design that does work only on data that has changed, an approach borrowed from how brains handle signals, to run AI at low power on edge devices such as cameras. Its GrAI VIP chip was aimed at vision and audio. Snap Inc. acquired the company in 2023, so it no longer appears to sell chips as an independent business. Neuromorphic Computing · Edge Computing —
- e-peas e-peas is a fabless Belgian company that makes ultra low power chips for energy harvesting, which means collecting small amounts of power from light, heat, vibration or radio waves. Its main products are power management ICs in the AEM family, which condense that trickle of energy into a stable supply for wireless sensors and similar devices. It also develops microcontrollers for the same low power market. The company was founded in 2014, is headquartered in Mont-Saint-Guibert, Belgium, and has offices in Switzerland and the United States. Edge Computing · ASICs €17.5M Series C (2025)
- Semtech Semtech is an American maker of analog and mixed-signal semiconductors, chips that handle real-world signals such as radio, power and touch. It developed LoRa, a low-power, long-range wireless networking technology for Internet of Things devices, and also sells proximity sensors, wireless charging chips and power management products. In January 2023 it completed its purchase of Sierra Wireless, a Canadian maker of cellular IoT modules, for about US$1.2 billion including debt. Semtech was founded in 1960 and is based in Camarillo, California. Custom Silicon · Edge Computing · Interconnect & Networking —
- NENetrasemi Netrasemi is an Indian fabless chip startup that designs processors for edge AI. Its A2000 is a system-on-chip for AI video analytics, and its R1000 is a microcontroller for smart IoT sensors, with both targeting TSMC's 12-nanometre process. In July 2025 it closed a Series A round of about 107 crore rupees, co-led by Zoho and Unicorn India Ventures. Zoho had earlier shelved its own plan to enter semiconductor manufacturing. Reports describe the company as Kerala-based. Edge Computing · AI Accelerators $12M Series A (Jul 2025, Zoho-led)
- EdgeCortix EdgeCortix is a fabless chip company that designs AI accelerators for edge devices. Its SAKURA-II chip, announced in May 2024, uses the company's second-generation Dynamic Neural Accelerator, a processing engine that can be reconfigured at run time to suit the model being run. It targets generative AI, including language and vision models, and ships as a chip, M.2 modules and PCIe cards. A software suite called MERA compiles models from PyTorch, TensorFlow Lite and ONNX. The company is based in Tokyo, Japan. Edge Computing · AI Accelerators Series B oversubscribed close, total >$110M (Nov 2025)
- Hailo Hailo designs low-power AI processors for edge devices such as cameras, vehicles and industrial equipment, where models run locally rather than in the cloud. Its Hailo-8 chip is rated at up to 26 tera operations per second. The company uses a data flow architecture, in which software maps each layer of a neural network onto separate regions of the chip so data moves less between memory and compute. Hailo is based in Israel and was founded in 2017 by former members of an Israel Defense Forces technology unit. Edge Computing · AI Accelerators SPAC merger MOU (March 2026); $9M emergency loan from Delek (Jan 2026)
- Axera Semiconductor Axera, known in Chinese as Aixin Yuanzhi, designs system-on-chip processors for AI inference, which means running trained models rather than training them. Its products cover vision chips for cameras, automotive chips for intelligent driving and smart cockpits, and edge computing chips in high, mid and low performance tiers. Two in-house technologies are a mixed-precision NPU (neural processing unit) and an AI-ISP, an image signal processor that uses neural networks to improve color in low-light images. It publishes investor relations material on its website. AI Accelerators · Edge Computing —
- OCOculii Oculii developed software that makes radar sharper, using adaptive waveforms and AI to give a small radar sensor the angular detail of a much larger antenna array, which the company called virtual aperture imaging. Its main market was automotive driver assistance. Oculii was acquired by Ambarella, a vision processor company, and is now part of that business rather than independent. The Ohio base is as recorded but not confirmed here. AI Accelerators · Edge Computing —
- EdgeQ EdgeQ designs a single chip that combines the radio processing of a 5G base station with AI processing, which it calls a Base Station-on-a-Chip. A base station is the equipment that links phones to a mobile network, and putting 5G and AI on one chip lets operators run machine learning at the network edge without separate hardware. The design uses RISC-V and Arm processor cores and targets small cells, macro cells and private 5G networks. EdgeQ was founded in 2018 and is based in Santa Clara, California, with offices in San Diego and Bangalore. Edge Computing · AI Accelerators —
- Autumn Labs Autumn Labs builds a monitoring platform for factory production lines, often described as Datadog for industrial robots. Datadog is a widely used tool for watching the health of software systems, and Autumn applies a similar idea to robotic stations, automated test stations and industrial arms. The platform collects live data, records traceability so each unit can be tied to how it was built, and moves that data securely off the line. The company was founded in 2023 by Prathik Muppidi, Harsh Bhatt and Sam Monga, is based in San Francisco, and joined Y Combinator in the Summer 2024 batch. Edge Computing YC S24
- Renesas Electronics Renesas Electronics makes microcontrollers, system-on-chip products, analog and power devices, with a large presence in automotive and industrial markets and a growing range of edge AI processors. It began operating in April 2010 when NEC Electronics absorbed Renesas Technology and took the Renesas name. Renesas Technology had itself been formed in 2003 from the semiconductor divisions of Hitachi and Mitsubishi Electric. The company is headquartered in Tokyo, Japan. ASICs · AI Accelerators · Edge Computing —
- BrainChip BrainChip develops the Akida processor, a neuromorphic design for low-power AI at the edge in devices such as sensors and cameras. Neuromorphic means the chip is modelled loosely on the brain and processes data as sparse events rather than continuously, which can cut energy use compared with conventional processors. The company sells Akida both as licensable IP and as a reference chip, and it supplies a software toolkit called MetaTF for building models. BrainChip is headquartered in Sydney, Australia, and its shares trade on the Australian Securities Exchange under the ticker BRN. Neuromorphic Computing · Edge Computing $25M (Dec 2025); LDA put option extended to AUD $140M
- HYHyperCIM HyperCIM is a small London-based start-up developing a Latency Processing Unit, an add-in device meant to take data preprocessing and protocol handling off CPUs and GPUs. It is built on compute-in-memory, which does calculations inside non-volatile memory cells instead of shuttling data to a separate processor, a design the company says cuts power use. Its first target is financial services. The company was incorporated in December 2023 and is incubated through Deep Science Ventures and Silicon Catalyst UK. No shipped product or customers have been confirmed. AI Accelerators · In-Memory Computing · Edge Computing —
- LELeapMind LeapMind was a Tokyo company that developed AI inference technology based on extremely low bit quantization, which shrinks a neural network by storing its numbers with very few bits so it runs on small, low power chips. Its main product was Efficiera, an accelerator design licensed for use in FPGAs or custom chips. In 2022 Maxell Frontier licensed Efficiera for an image recognition product line. LeapMind was founded in 2012, and the company reported total investment of 4.99 billion yen as of May 2021. AI Accelerators · FPGAs · Edge Computing ¥5B+
- QuickLogic QuickLogic sells embedded FPGA, or eFPGA, intellectual property, which is a block of reprogrammable logic that other chipmakers place inside their own system-on-chip designs. It also offers antifuse and SRAM-based FPGAs, FPGA microcontrollers and custom rad-hard eFPGA IP, meaning versions built to tolerate radiation for aerospace and defence use. The company is based in San Jose, California, and was founded in 1988. FPGAs · Edge Computing —
