What Mythic does

Mythic is a privately held AI semiconductor company founded in 2012 that develops analog compute-in-memory (CIM) processors for AI inference across edge devices, data centers, automotive, robotics, and defense. Its core technology is the Mythic Analog Compute Engine (Mythic ACE), which performs massively parallel matrix multiplication directly inside flash memory arrays using analog computation, eliminating the memory bandwidth bottleneck inherent to Von Neumann GPU architectures. The company's products include the M1076 Analog Matrix Processor (Mythic AMP) delivering up to 25 TOPS at 3-4W, M.2 card form factors (MM1076, ME1076), and a roadmap of M2000 and Gen 2 chiplet-based processors; the May 2026 acquisition of Videantis added a unified VLIW+SIMD digital processor platform (v-MP6000UDX) with 25M+ chips shipped, creating a hybrid analog+digital AI compute architecture. Mythic generates revenue through hardware sales of its AMP chips and M.2 cards, technology licensing and joint development (notably with Honda R&D for automotive SoCs targeting 100,000+ TOPS), and IP licensing via the Videantis subsidiary into top-three European automotive OEMs and a top-three global semiconductor company. The company has raised more than $300M in venture funding across rounds led by DFJ, Lux Capital, BlackRock/HPE Pathfinder, and most recently a $125M December 2025 round led by DCVC with strategic investors Honda Motors and Lockheed Martin.

Name

Mythic

Legal name

Mythic, Inc.

Website

https://mythic.ai

Company type

Private

Founded year

2012

Operating status

Operating

Headcount range

51–100 employees

Short description

Mythic develops analog compute-in-memory AI inference chips (Mythic AMP/APUs) that perform matrix multiplication inside flash memory arrays, delivering claimed 100x energy and cost efficiency versus GPUs for edge, automotive, robotics, defense, and data center workloads.

Ownership category

akta.pro rank

Product category

AI Inference Processors

NAICS

Electronic Computer Manufacturing (334111)

SIC

Electronic Computers (3571)

akta.pro primary industry

AI Accelerators (GPUs/TPUs/NPUs/ASICs) (HDAAAAAA)

akta.pro secondary industries

Edge AI Hardware & NPUs (SoCs, modules, accelerators) (HDAAAJAC), Compute & Acceleration ICs (CPU/GPU/AI/FPGA/ASIC) (HDAHABAC)

Keywords

• Analog AI processors
• Edge AI inference
• AI accelerator chips
• Compute-in-memory
• AI semiconductor IP

Headquarters

HQ city

Austin

HQ country

United States

HQ region

North America

Offices5 records

Markets served

GTM type

B2B

Offering type

Hardware or Manufacturing

Cost components

Personnel, Technology or R&D, Operations, Marketing or Sales, Infrastructure

Revenue model

1. Hardware sales (chips and PCIe cards): Revenue from selling the M1076 Analog Matrix Processor chip, MM1076 M.2 M Key card, and ME1076 M.2 A+E Key card products for edge AI applications including surveillance, industrial machine vision, drones, AR/VR, and edge servers.
2. Technology licensing and joint development: Revenue from licensing Mythic's APU technology and co-developing application-specific SoCs with strategic partners such as Honda R&D, including cross-product deployment across the partner's future vehicle lineup.
3. Videantis IP licensing and chip integration: Acquired business from Videantis generates revenue from licensing unified digital processor IP that is already shipping in increasing major volumes inside chips from a top-three global semiconductor company, with deployment across all top three European automotive manufacturers.

Pricing tiers

Go-to-market motion1 record

Distribution channels4 records

Marketing channels7 records

Core offering

Mythic develops and sells Analog Processing Units (APUs) and M.2 cards based on analog compute-in-memory architecture for AI inference at the edge and in data centers. The company ships the M1076 Analog Matrix Processor (up to 25 TOPS at 3-4W) and its M.2 card variants, complemented by the CAMP software development kit. Mythic also licenses its APU technology for joint development of application-specific SoCs (e.g., with Honda R&D for software-defined vehicles) and, following its acquisition of Videantis, offers the v-MP6000UDX unified digital processor IP for hybrid analog+digital AI compute platforms.

Product overview

Mythic offers a unified analog-AI compute platform centered on its proprietary Mythic ACE (Analog Compute Engine) tile IP and shipped as the M1076 Analog Matrix Processor (Mythic AMP) and its M.2 card variants (MM1076, ME1076) for edge inference, with the M1108 AMP as the prior-generation 35-TOPS chip and the M2000 series plus scalable Gen 2 chiplet architecture as the next-generation roadmap. The platform layers a complete software stack — the Mythic Optimization Suite, Mythic Graph Compiler, and Mythic CAMP SDK (supporting ONNX, PyTorch, TensorFlow, NVIDIA TensorRT, and NVIDIA Jetson) — around the silicon, and is now complemented by the acquired Videantis v-MP6000UDX unified digital processor IP, the Starlight sensor-embedded analog co-processor, and the joint Honda automotive-grade AI SoC program, together forming a hybrid analog+digital AI compute platform targeting edge devices, robotics, defense, automotive, and data center inference workloads.

Differentiator

Problem solved

Functional benefit

Brands

• Mythic AMP (Analog Matrix Processor): Mythic's product line of Analog Matrix Processor chips for AI inference (e.g., M1076, M1108, M2000 series).
• Mythic ACE (Analog Compute Engine)
• Starlight
• CAMP (Compute Analog in-Memory Processing) SDK

Products and services

• M1076 Analog Matrix Processor (Mythic AMP) Single-chip analog AI inference processor (Mythic AMP) built on an array of 76 Mythic ACE tiles, integrating up to 80M on-chip weights with no external DRAM. Delivers up to 25 TOPS at 3-4W typical power in a 19mm x 15.5mm BGA package, targeted at edge AI applications including surveillance, industrial machine vision, drones, AR/VR, and edge servers.
• MM1076 M.2 M Key Card Compact 22mm x 80mm M.2 M-key card based on the M1076 Mythic AMP for high-performance, power-efficient AI inference in edge devices and edge servers. Supports Ubuntu, NVIDIA L4T, and Windows; 4-lane PCIe 2.1 delivering up to 2 GB/s.
• ME1076 M.2 A+E Key Card Compact 22mm x 30mm M.2 A+E-key card based on the M1076 Mythic AMP for power-efficient AI inference in size-constrained edge devices and edge servers, with 2-lane PCIe 2.1 delivering up to 1 GB/s.
• M1108 Analog Matrix Processor (Mythic AMP) Prior-generation Analog Matrix Processor and industry-first AMP for AI inference. Integrates 108 AMP tiles each with a Mythic ACE, a 32-bit RISC-V nano-processor, SIMD vector engine, SRAM, and NOC router, plus four PCIe 2.0 control tiles. Delivers up to 35 TOPS at approximately 4W typical power on a mature 40nm process with no external DRAM, available in M.2 and PCIe card form factors.
• M2000 Series Analog Computing Solution Next-generation analog compute-in-memory product family that builds on the M1076 AMP, with a new software toolkit and streamlined architecture delivering an additional order-of-magnitude leap in power and cost efficiency for robotics, security, defense, smart cities, smart home, and consumer edge AI.
• Starlight Analog Co-Processor Sub-1W analog AI sensor co-processor designed to be embedded inside image sensors; improves sensor performance by 50x by extracting signals from noise using analog AI. Targets mission-critical applications including low-light defense, robotics, and automotive sensing.
• Mythic CAMP SDK (Compute Analog in-Memory Processing SDK) Software development kit for frictionless deployment of deep neural network applications on Mythic GEN 1 APUs. Supports ONNX, PyTorch, and TensorFlow natively as well as via NVIDIA TensorRT on platforms including NVIDIA Jetson, and is rated at the highest maturity level among compute-in-memory processor peers.
• Videantis v-MP6000UDX Unified Processor Platform Unified VLIW + SIMD digital processor IP that runs deep learning, classical computer vision, signal/image processing, video encode/decode, SLAM, attention calculations, NMS, and AI inference on one architecture. Acquired from Videantis GmbH in May 2026 and shipping in volume inside chips from a top-three global semiconductor company.

Quantifiable outcome

• 100x more energy-efficient than industry-standard GPUs (including memory transfers)
• +9 more outcomes

Named customers3 records

Ideal customer profiles5 records

Technology focussed Yes

API detail

Has API

No

API docs

API detail

Core technology

AI maturity

App detail

Integration7 records

AI capability9 records

Feature7 records

Partnerships

Four partnerships are on record, tiered flagship and core.

• Videantis GmbHflagshipStrategic or Co-development Partner · 19 May 2026Acquired by Mythic to combine analog compute-in-memory with Videantis' unified VLIW+SIMD digital processor architecture (v-MP6000UDX) into a hybrid AI compute platform with 100x energy efficiency advantage over conventional GPU-based systems. Videantis' processor IP has powered more than 25 million chips shipped worldwide with zero field defects, deployed across all top three European automotive manufacturers. All five founders join Mythic, with Videantis continuing as a wholly owned subsidiary.
• Microchip Technology (Silicon Storage Technology / SST)coreTechnology or Integration · 17 March 2026Mythic selected SST's memBrain neuromorphic hardware IP (built on SuperFlash embedded non-volatile memory, 150B+ units shipped) for its next-generation Analog Processing Units, enabling up to 120 TOPS/watt inference and APUs up to 100x more energy-efficient than conventional digital GPUs. Technology deployed in 40nm and 28nm processes, with 22nm development planned.
• Honda R&D Co., Ltd.flagshipStrategic or Co-development Partner · 6 February 2026Joint development agreement in which Honda R&D licenses Mythic's Analog Processing Unit (APU) technology and the companies co-develop an automotive-grade AI SoC for Honda's next-generation software-defined vehicles (SDVs). Targeted deployment by the late 2020s/early 2030s with over 100,000 TOPS of AI performance for advanced driver-assist and autonomous features, supporting Honda's goal of zero traffic collision fatalities by 2050. Relationship expected to expand across Honda's future vehicle lineup.
• ModalAIcoreTechnology or Integration · 3 May 2022Mythic and ModalAI partnered to integrate the M1076 AMP (up to 25 TOPS of additional AI performance) into ModalAI's VOXL 2 UAV platform, enabling SWAP-optimized smaller, smarter, and safer drones built in the U.S. for security, agriculture, and other commercial applications.

Scale indicators12 records

Recent moves6 records

Expansion highlights6 records

Broad incumbents

• NVIDIA: Dominant GPU and AI accelerator incumbent. Mythic explicitly positions its analog compute-in-memory APUs as a 100x energy-efficiency challenger to NVIDIA's GPUs across data center and edge inference. NVIDIA is the benchmark Mythic competes against.

Direct peers

• Groq: AI inference accelerator company using a deterministic LPU architecture. Competes with Mythic directly in the high-throughput, low-latency AI inference market, particularly for LLM serving where Mythic claims its 750x tokens/sec/watt advantage.
• Cerebras Systems: AI accelerator company building wafer-scale compute engines for training and inference. Competes with Mythic in the high-performance AI compute market, addressing similar customer demand for alternatives to NVIDIA GPUs.
• SambaNova Systems: AI accelerator company offering reconfigurable dataflow units for enterprise AI inference and training. Targets the same enterprise inference workloads that Mythic's APUs are designed for, with similar positioning around TCO and energy efficiency.
• Graphcore: AI accelerator company whose Intelligence Processing Unit (IPU) targets data center and enterprise AI workloads. Competes with Mythic on the promise of non-GPU architectures delivering better economics for AI inference.
• Tenstorrent: AI accelerator company developing RISC-V based processors for AI training and inference. Comparable to Mythic as an alternative-architecture AI chip startup targeting both data center and edge markets.
• Lightmatter: AI compute company using photonic (light-based) processors for AI inference and interconnect. Closest analog compute-in-memory competitor to Mythic, claiming similar energy-efficiency advantages over GPUs using silicon photonics instead of analog flash.
• Hailo: Edge AI processor company producing low-power AI accelerators for computer vision at the edge. Most direct competitor to Mythic's M1076 AMP in the edge inference market for surveillance, industrial machine vision, and IoT applications.

Emerging players

• BrainChip: Neuromorphic processor company focused on ultra-low-power edge AI inference using spiking neural networks. Comparable to Mythic in targeting sub-watt edge AI workloads with fundamentally non-Von-Neumann architectures.
• Syntiant: Low-power AI processor company for always-on edge AI in battery-powered devices. Overlaps with Mythic's Starlight analog co-processor vision for sub-1W edge AI inference in sensors and IoT.

Market position

Strengths5 records

Weaknesses5 records

Competitive moat5 records

Key risks5 records

Key highlights7 records

Customer concentration

• Xx.com/mythicinc
• Facebookfacebook.com/mythicinc
• LinkedInlinkedin.com/company/mythic-ai

Compliance1 record

Revenue estimate

Valuation estimate

Number of profiles

Profiles9 records

Subsidiaries2 records

Funding overview

Funding rounds8 records

Investors25 records

Funding detail is available on the Subscription and Enterprise plan.Contact sales →

M&A1 record

Investments

M&A and investment is available on the Subscription and Enterprise plan.Contact sales →

Frequently asked questions about Mythic