Quadric

What Quadric does

Quadric is a privately held, venture-backed semiconductor intellectual property company founded in 2016 and headquartered in Burlingame, California. It designs and licenses the Chimera General Purpose Neural Processing Unit (GPNPU) — a hybrid Von Neumann + 2D SIMD processor IP that executes matrix, vector, and scalar operations in a single fully C++ programmable core, eliminating the graph partitioning required by heterogeneous NPU+DSP+CPU designs. The portfolio scales from the 1-7 TOPS QC-Nano through QC-Perform, QC-Ultra, and up to the QC-Multi cluster at 6,400 TOPS, with Safety Enhanced variants marketed as ASIL B/D Ready for automotive and support for transformers up to 30B parameters via cluster and UCIe chiplet configurations. The IP is synthesized into customer SoCs across 12nm to 3nm process nodes.

The company pairs the processor IP with the Chimera software stack — the Chimera Graph Compiler (ONNX ingest, multi-pass optimization, INT8 quantization), the Chimera LLVM C++ compiler, the ChiPy Python DSL, and a cycle-accurate Instruction Set Simulator — plus the cloud-based DevStudio environment that exposes a 350+ model benchmark zoo and a Model Context Protocol server for AI-agent access. Quadric's revenue comes from upfront and recurring IP licensing and development-license fees, supplemented by co-development engagements (e.g., DENSO, NSITEXE, MegaChips, ams OSRAM) and paid DevStudio services billed via Stripe.

Quadric sells primarily to semiconductor companies, ASIC design services firms, and Tier 1 automotive suppliers building AI-capable SoCs for edge LLMs, ADAS and autonomous driving, robotics, office automation, AI PCs, and sovereign AI programs. Named customers include DENSO, NSITEXE, Kyocera Document Solutions, MegaChips, TIER IV, and ams OSRAM. The company operates across North America (Burlingame HQ), Asia (Tokyo subsidiary Quadric KK, India R&D, plus offices in China, South Korea, and Taiwan) and Europe, with 51-100 employees and approximately $72M of total capital raised as of January 2026.

Name

Quadric

Legal name

Quadric, Inc.

Website

https://quadric.ai

Company type

Private

Founded year

2016

Operating status

Operating

Headcount range

51–100 employees

Short description

Quadric develops licensable Chimera GPNPU processor IP and supporting compiler tools that enable fully programmable on-device AI inference for semiconductor designers targeting automotive, edge LLM, robotics, and enterprise vision markets.

Ownership category

akta.pro rank

Product category

AI Processor IP

NAICS

Semiconductor and Related Device Manufacturing (334413)

SIC

Semiconductors & Related Devices (3674)

akta.pro primary industry

Compute & Acceleration ICs (CPU/GPU/AI/FPGA/ASIC) (HDAHABAC)

akta.pro secondary industries

Hardware/Chip Design IP Cores (Processor, Interface, Analog, Security) (HDAHAFAF), Edge AI Hardware & NPUs (SoCs, modules, accelerators) (HDAAAJAC)

Keywords

• Neural processing unit
• AI processor IP
• Semiconductor IP licensing
• Edge AI inference
• On-device AI acceleration

Headquarters

HQ city

Burlingame

HQ country

United States

HQ region

North America

Offices6 records

Markets served

GTM type

B2B

Offering type

Software

Cost components

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

Revenue model

1. Processor IP Licensing: Primary revenue stream: licensing of Chimera GPNPU processor IP cores to semiconductor companies and systems companies designing AI-capable chips. Licensees span automotive, edge LLM, office automation, and autonomous driving markets. Customers reportedly go from engagement to production-ready AI silicon in under six months.
2. Development License & Co-Development Fees: Includes development license agreements with automotive partners such as DENSO for joint IP co-development targeting specific use cases like in-vehicle NPU. Includes evaluation/SDK licenses such as with TIER IV for Autoware development.
3. DevStudio Cloud Platform Access: Cloud-based collaborative development environment providing SDK, benchmarks, model zoo, and compilation tools. Per Terms of Use, certain features may be subject to Paid Services with recurring billing via Stripe payment processor.

Pricing tiers

Go-to-market motion1 record

Distribution channels5 records

Marketing channels8 records

Core offering

Quadric licenses the Chimera General-Purpose Neural Processing Unit (GPNPU) processor IP family to semiconductor and systems companies designing AI-capable chips. The Chimera GPNPU is a fully programmable hybrid Von Neumann + 2D SIMD architecture delivering 1 to 6,400 TOPS, paired with the Chimera SDK (compiler, LLVM toolchain, instruction set simulator) and the cloud-based DevStudio environment. Together, these products let chip designers run modern AI workloads including transformers, ViTs, LLMs (up to 30B parameters), and vision models on a single unified processor core without graph partitioning between NPU, DSP, and CPU blocks.

Product overview

Quadric offers a unified AI inference platform centered on the Chimera General-Purpose Neural Processing Unit (GPNPU) processor IP family — a single fully programmable hybrid Von Neumann + 2D SIMD architecture that scales from the 1–7 TOPS QC-Nano single core, through the 4–28 TOPS QC-Perform and 16–108 TOPS QC-Ultra single cores, up to the 32–6,400 TOPS QC-Multi multi-core/multi-instance cluster. The hardware IP is paired with the Chimera SDK — comprising the Chimera Graph Compiler (CGC), the Chimera LLVM C++ Compiler, the ChiPy Python DSL, and the cycle-accurate Instruction Set Simulator (ISS) — and with the cloud-based DevStudio environment, where customers compile ONNX models from PyTorch/TensorFlow, run pre-built model demos, profile thousands of parametric Chimera configurations, and access live MCP-exposed resources for connected AI agents. Together these products enable customers to go from engagement to production-ready LLM-capable silicon in under six months across automotive, edge-server LLM, robotics, consumer, and enterprise vision applications.

Differentiator

Problem solved

Functional benefit

Brands

• Chimera: Chimera GPNPU — general-purpose neural processing unit semiconductor IP product family offered by Quadric.
• DevStudio
• Chimera SDK

Products and services

• Chimera GPNPU Processor IP Family Licensable General-Purpose Neural Processing Unit (GPNPU) semiconductor IP with a hybrid Von Neumann + 2D SIMD architecture, 7-stage in-order pipeline, and 64-bit instruction word. Sold as the QC-Nano (1-7 TOPS), QC-Perform (4-28 TOPS), QC-Ultra (16-108 TOPS), and QC-Multi (32-6,400 TOPS) variants synthesized for 12 nm to 3 nm process nodes, with optional ASIL B/D Ready safety-enhanced configurations. Target customers are semiconductor and SoC design companies building AI-capable chips for automotive, edge LLM, robotics, consumer, and enterprise vision.
• Chimera SDK End-to-end software toolkit for compiling AI graphs and C++ code to the Chimera GPNPU. Includes the Chimera Graph Compiler (ONNX import, INT8 auto-quantization, multi-pass optimization, CCL output), the Chimera LLVM C++ Compiler with Quadric-specific back end, the cycle-accurate Instruction Set Simulator (ISS), and Docker-based deployment for on-premise/private-cloud compilation and profiling. For chip designers and ML engineers developing Chimera-based SoCs.
• DevStudio Cloud-based collaborative development environment for compiling, simulating, profiling, and benchmarking AI models on Chimera GPNPU configurations. Hosts a 350+ curated model zoo, parametric configuration sweeps across QC-Nano/QC-Perform/QC-Ultra product families, cycle-accurate job logs, downloadable compiled artifacts, and serves as the data backend for Quadric's MCP server used by connected AI agents. Free registration required; certain features subject to Paid Services with recurring billing via Stripe.
• ChiPy Python DSL Quadric's Python domain-specific language for building end-to-end on-chip AI pipelines that compile directly to native Chimera GPNPU binaries without host CPU intervention or DDR round-trips, leveraging the same C++ kernel extension path as the rest of the SDK. For ML engineers and embedded developers building on Chimera.
• DevStudio MCP Server Quadric's Model Context Protocol server that exposes DevStudio documentation search, model catalog search, and live resources to connected AI agents (e.g., Claude) so they can understand the Chimera GPNPU stack. For developers and partners integrating AI agents into the Chimera evaluation workflow.

Named customers6 records

Segments7 records

Ideal customer profiles4 records

Technology focussed Yes

API detail

Has API

Yes

API docs

API detail

Core technology

AI maturity

App detail

AI capability14 records

Feature11 records

Partnerships

Four partnerships are on record, tiered flagship automotive customer and co-development partner, ecosystem partner for startup customer acquisition, joint ecosystem marketing partner for intelligent image sensors and flagship distribution and co-development partner in japan/asia.

• DENSO Corporationflagship automotive customer and co-development partnerStrategic or Co-development Partner · 29 October 2024Signed a development license agreement (October 2024) for an NPU, under which DENSO acquires the Chimera GPNPU IP core license and the two companies co-develop IP for an in-vehicle semiconductor targeting advanced driver assistance and intelligent vehicle applications. Builds on the earlier NSITEXE/Quadric joint development and represents DENSO's direct commitment to the GPNPU architecture for production ADAS chips.
• Silicon Catalystecosystem partner for startup customer acquisitionGTM or Marketing Partner · 29 June 2023Joined as a member of Silicon Catalyst's In-Kind Partner (IKP) ecosystem in June 2023, making Chimera GPNPU IP available to Silicon Catalyst's portfolio companies in the incubator. Provides technical and business support to early-stage semiconductor startups, accelerating Quadric's reach into emerging chip designers.
• ams OSRAMjoint ecosystem marketing partner for intelligent image sensorsTechnology or Integration · 5 January 2023Collaboration announced for CES 2023 (January 2023) demonstrating Chimera GPNPU processors alongside ams OSRAM Mira CMOS image sensors to deliver low-power, small-form-factor smart sensing demos for face detection, human pose, motion and occupancy detection. Showcases the integration of Quadric IP with industry-leading image sensor technology.
• MegaChipsflagship distribution and co-development partner in japan/asiaStrategic or Co-development Partner · 16 March 2022Strategic partnership (May 2022) with this leading Japan-based ASIC and SoC services company to deliver ASIC and SoC solutions built on Quadric's GPNPU architecture. Also participated as a major investor in the Series B funding round. Combines capital, co-development, and channel reach into the Japanese ASIC market.

Scale indicators16 records

Recent moves8 records

Expansion highlights6 records

Broad incumbents

• Arm: Arm's Ethos NPU IP family (Ethos-U55, U65, N78) is the most widely deployed licensable NPU IP in the world, integrated across billions of edge devices. As the dominant processor IP incumbent with established customer relationships at every major SoC designer, Arm is the most direct competitive threat to Quadric's Chimera GPNPU in edge AI inference, though Arm's NPU IP tends to be more fixed-function and less programmable than Chimera's single-core C++ approach.
• Synopsys: Synopsys' ARC NPX and EV7x processor IP families target the same licensable AI/ML inference IP market as Quadric, with the IP group scaled to $1.9B in annual revenue under former EVP/GM Joachim Kunkel (now on Quadric's board). Synopsys bundles NPU IP with its dominant EDA toolchain, creating both a formidable competitor and a potential long-term partnership vector; its ARC NeuroProcessor and DesignWare EV cores overlap directly with Chimera's programmable inference positioning.
• Cadence Design Systems: Cadence's NeuroWeave and Tensilica AI IP families offer licensable neural network and DSP processors to SoC designers, bundled with Cadence's EDA flow and digital implementation tools. Cadence competes with Quadric for the same automotive, vision, and edge AI IP sockets, and previously employed Quadric's VP Engineering (Dhanendra Jani) — making it both a direct competitor and a recruiting source for Quadric's leadership bench.

Direct peers

• Ceva: Ceva's NeuPro family of NPU IP and NeuPro-M/Nano edge AI cores target the same licensable on-device AI inference market as Quadric, with comparable performance-per-watt claims and design wins in consumer, IoT, and automotive. Ceva's installed base in DSP-based SoCs gives it strong incumbency with voice, audio, and vision customers that overlap with Quadric's edge AI pipeline, making it one of the closest direct comparables in business model and target market.
• VeriSilicon: VeriSilicon's VIP Series (VIPNano, VIP7000, VIP9000) NPU IP cores offer licensable AI inference processors for edge and embedded applications, serving similar automotive, IoT, and consumer SoC customers. As a China-headquartered IP vendor with strong presence in the Chinese SoC ecosystem, VeriSilicon competes with Quadric in some of the same design wins — particularly in image processing, vision, and consumer AI applications — while also serving as a regional alternative.
• Imagination Technologies: Imagination's PowerVR GPU and NNA (Neural Network Accelerator) IP families target licensable edge AI inference with comparable TOPS-per-watt positioning to Chimera. With a heritage in GPU IP and a renewed push into automotive and consumer AI inference, Imagination competes with Quadric for the same SoC design wins, particularly in vision-based automotive and consumer applications where both vendors are well-positioned.

Emerging players

• Lightmatter: Lightmatter builds photonic AI compute hardware and interconnects targeting data center and edge inference workloads, and is expanding into licensable photonic IP for AI accelerators. While Lightmatter's primary product is silicon photonics-based hardware rather than a programmable NPU core IP, it competes for the same SoC designer mindshare and capital on the 'next generation of AI compute' narrative that Quadric is building its GPNPU positioning around.
• Tenstorrent: Tenstorrent develops RISC-V-based AI accelerator IP and licensable processor IP targeting edge and data center inference, with a comparable focus on programmable AI architectures and an emphasis on supporting modern transformer models natively. While Tenstorrent is more vertically integrated (selling its own AI accelerator cards in addition to IP), its RISC-V-based Wormhole/Blackhole IP overlaps with Quadric's positioning for customers seeking an open, programmable AI compute foundation.
• Esperanto Technologies: Esperanto Technologies develops RISC-V-based AI accelerator IP and inference SoCs targeting data center and edge applications, with an architecture emphasizing energy-efficient, programmable AI compute. As a venture-backed NPU IP startup with comparable scale to Quadric (private, post-Series stages), Esperanto serves as a useful reference point for valuation, headcount, and design-win benchmarks in the emerging RISC-V NPU IP category.

Others

• Edge AI and Vision Alliance member ecosystem (e.g., Blaize, Syntiant): Companies like Blaize and Syntiant operate in the adjacent edge AI chip space, often building their own NPU architectures for low-power inference SoCs rather than licensing IP. They are ecosystem participants alongside Quadric in the Edge AI and Vision Alliance and compete for some of the same end-customer design wins (e.g., always-on vision, audio, IoT), but their business model (selling chips vs. licensing IP) is fundamentally different from Quadric's licensable GPNPU approach.

Market position

Strengths5 records

Weaknesses5 records

Competitive moat6 records

Key risks5 records

Key highlights7 records

Customer concentration

• Xx.com/quadric_io

Compliance4 records

Revenue estimate

Valuation estimate

Number of profiles

Profiles9 records

Subsidiaries1 record

Funding overview

Funding rounds7 records

Investors22 records

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

M&A

Investments

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

Frequently asked questions about Quadric