NEUCHIPS

What NEUCHIPS does

NEUCHIPS Corp., founded in 2019 and headquartered in Hsinchu Science Park, Taiwan, with additional offices in Taipei and Silicon Valley, designs AI-specific ASICs (Application-Specific Integrated Circuits) for deep learning inference. The company's portfolio spans inference silicon (RecAccel N3000 for Deep Learning Recommendation Models, Raptor N3000 for Large Language Models), PCIe and M.2 inference cards (Viper series, DM2 series), licensable silicon IP (Blue Magpie Neural Processing Unit IP, NeuKompression compression IP, eFuse Controller IP), and turnkey enterprise AI solutions (ShareQA on-premise generative AI service with Multi-Agent and NL2SQL capabilities; NeuSight vision-based anomaly detection for semiconductor manufacturing). The N3000 architecture integrates a Matrix Engine, Vector Engine for feature cross operations, and an Embedding Engine that reduces off-chip memory access by 50% — a design philosophy the company describes as "Silicon-Centric DNA." Patented FFP8 quantization preserves >99.95% of FP32 accuracy at 8-bit storage, and independently validated MLPerf v3.0 results show 1.7x better performance-per-watt than Nvidia H100 on datacenter DLRM inference.

Revenue is generated through three primary streams: (1) hardware sales of AI inference accelerators and PCIe/M.2 inference cards to data centers, enterprises, and SMEs; (2) silicon IP licensing (NPU, NeuKompression, eFuse) and front-end design services to semiconductor companies building their own AI ASICs; and (3) managed AI services such as NeuSight, which processes 35,000+ videos daily for industrial quality control. Pricing is quote-based on hardware contracts and subscription-based on managed services. The company sells via direct enterprise sales, system integration partners (Vecow, ShareGuru/GSH, Netiotek), and a cloud infrastructure partnership with Taiwan's National Center for High-Performance Computing (NCHC). Cumulative funding totals approximately $75M across Series A (2019), Series B (2020/2022), and Series B3 (2025), with strategic investors including Wistron, JAFCO Asia, Powerchip Group, and Raydium Semiconductor.

The customer base spans hyperscaler-adjacent data centers (NCHC), semiconductor manufacturers (NeuSight deployments), large enterprises (offline GenAI for data sovereignty), and SMEs (MINI SERVER bundle). Operating geographies cover Taiwan, the United States, and Japan, with active participation at COMPUTEX, CES, Japan IT Week, Embedded World, and the AI Hardware & Edge AI Summit supporting demand generation.

Name

NEUCHIPS

Legal name

Neuchips Corp.

Website

https://neuchips.ai

Company type

Private

Founded year

2019

Operating status

Operating

Headcount range

51–100 employees

Short description

NEUCHIPS Corp. is a Taiwan-based AI ASIC company that designs deep learning inference accelerators (DLRM and LLM), licensable silicon IP, and on-premise GenAI solutions for data centers, enterprises, and SMEs seeking power-efficient, data-sovereign AI infrastructure.

Ownership category

akta.pro rank

Product category

AI Inference Accelerators

NAICS

Computer Systems Design Services (541512)

SIC

Services-Computer Integrated Systems Design (7373)

akta.pro primary industry

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

akta.pro secondary industries

AI Accelerators (GPUs/TPUs/NPUs/ASICs) (HDAAAAAA), Model Deployment, Serving & Inference Platforms (HDAAABAF), Model Compression & Optimization (Quantization, Distillation, Pruning) (HDAAACAL), AI Compiler, Runtime & Kernel Optimization Software (CUDA/ROCm/XLA, graph compilers) (HDAAAAAI)

Keywords

• AI inference ASICs
• DLRM acceleration
• LLM inference cards
• Silicon IP licensing
• On-premise GenAI

Headquarters

HQ city

Hsinchu

HQ country

Taiwan

HQ region

Asia

Offices3 records

Markets served

GTM type

B2B

Offering type

Hardware or Manufacturing

Cost components

Technology or R&D, Personnel, Supply Chain, Operations, Marketing or Sales

Revenue model

1. AI ASIC Hardware Sales: Sale of AI inference accelerators including RecAccel N3000 for DLRM and Raptor N3000 for LLM workloads. PCIe accelerator cards (Viper Series, DM2 Series) for enterprise AI deployments. Revenue from data center and enterprise customers.
2. Silicon IP Licensing: Licensing of silicon-proven technology including Neural Processing Unit (NPU) IP, NeuKompression IP, and eFuse Controller IP to help partners build their own application-specific SoCs.
3. Front-End Design Services: Front-End Design-as-a-Service including architecture & RTL design, verification & power analysis, emulation & validation. Functions as extension of customer engineering teams for AI hardware development.
4. NeuSight AI Training Service: Vision-based anomaly detection service offering customized AI model training for industrial quality control applications. Processes over 35,000 videos daily.

Pricing tiers

Go-to-market motion3 records

Distribution channels4 records

Marketing channels7 records

Core offering

NEUCHIPS designs and sells purpose-built AI ASICs (Application-Specific Integrated Circuits) for deep learning inference workloads. Its product portfolio includes the RecAccel N3000 DLRM inference accelerator, Raptor N3000 LLM inference ASIC, PCIe inference cards (Viper and DM2 series), licensable silicon IP (Blue Magpie NPU, NeuKompression compression, eFuse Controller), front-end design services, and managed AI solutions (ShareQA on-premise GenAI service and NeuSight anomaly detection). The company targets data centers, hyperscalers, enterprises, and industrial manufacturers with energy-efficient, workload-specific AI silicon.

Product overview

Neuchips offers a comprehensive AI silicon and solutions portfolio centered on energy-efficient ASICs for inference workloads. The product family spans AI Inference ASICs (RecAccel N3000 for DLRM, Raptor N3000 for LLM), Silicon IP (Blue Magpie NPU, NeuKompression compression, eFuse Controller), Generative AI Inference Cards (Viper and DM2 series PCIe/M.2 modules), and enterprise AI Solutions (ShareQA on-premise GenAI service with customer service and knowledge management systems, NeuSight anomaly detection). The offerings are unified by Neuchips' software stack including proprietary compiler, SDK with PyTorch extension, and support for popular pre-trained models like Llama and Mistral, enabling deployment from edge to data center.

Differentiator

Problem solved

Functional benefit

Brands

• RecAccel: DLRM Inference Accelerator product line including RecAccel N3000
• Raptor
• Viper
• DM2
• ShareQA
• NeuSight
• NeuKompression
• Blue Magpie

Quantifiable outcome

• 2.2x better performance-per-watt for DLRM inference compared to alternatives
• +6 more outcomes

Named customers3 records

Segments4 records

Ideal customer profiles4 records

Technology focussed Yes

API detail

Has API

Yes

API docs

API detail

Core technology

AI maturity

App detail

AI capability9 records

Feature7 records

Partnerships

Six partnerships are on record, tiered core, flagship and minor.

• NetiotekcoreTechnology or Integration · 1 April 2026Collaboration to launch integrated on-premise AI solution. Netiotek provides NERMPC-265K industrial-grade edge computing platform with superior stability and thermal design. Combined with Neuchips' AI inference acceleration cards optimized for Transformer architectures for Japan IT Week showcase.
• ShareGuru Intelligence TechnologiesflagshipStrategic or Co-development Partner · 1 October 2025Deep collaboration to deliver on-premise Generative AI solutions. Combined Neuchips' Viper LLM Inference Card with ShareGuru's Multi-Agent AI Application Platform and NL2SQL technology. Jointly developed MINI SERVER All-in-One Host for enterprise customer service and knowledge management. Partnership selected for TAIWAN AI RAP program trial operation.
• VecowcoreStrategic or Co-development Partner · 1 March 2025Partnership to create compact RAG-enabled LLM computing platforms. Vecow integrates Neuchips Viper AI accelerator cards into ECX-3100 RAG Edge AI Inference Workstation. Collaboration for industrial AI deployment with private, secure, and power-efficient AI solution for natural language SQL queries.
• National Center for High-Performance Computing (NCHC)coreTechnology or Integration · 1 January 2025Collaboration with Taiwan's NCHC under NIAR to bring cloud-native customers power-efficient AI solutions. Neuchips AI accelerators combined with NCHC infrastructure to reduce cloud service costs while ensuring data security through hybrid architecture. Partnership for TAIWAN AI RAP program.
• MAPLE LEAF INFORMATION AND TECHNOLOGYminorStrategic or Co-development Partner · 1 January 2025Collaboration to create compact systems for complex generative AI applications without requiring additional power configurations.
• Golden Smart Home (GSH) Technology Corp.coreStrategic or Co-development Partner · 1 January 2025GSH's ShareGuru division combines Multi-Agent technology with NL2SQL Agentic AI for high-precision on-premise AI meeting analysis and intelligent customer service solutions. Integrated with Neuchips Viper series for 45W power efficiency with 12B parameter models.

Scale indicators10 records

Recent moves6 records

Expansion highlights6 records

Direct peers

• Tenstorrent: Tenstorrent designs RISC-V-based AI accelerators and licenses its IP to other chip makers, directly comparable to Neuchips' dual ASIC-plus-IP licensing strategy (Raptor N3000 + Blue Magpie NPU IP).
• Groq: Groq designs LPU inference ASICs purpose-built for low-latency LLM serving, directly competing with Neuchips' Raptor N3000 and Viper cards in the AI inference accelerator market. Both companies target the same emerging custom-silicon inference niche.
• SambaNova Systems: SambaNova develops RDU (Reconfigurable Dataflow Unit) AI accelerators for enterprise and data center inference. Like Neuchips, it targets on-premise and sovereign AI deployments with a full-stack hardware-plus-software approach.
• Cerebras Systems: Cerebras builds wafer-scale AI accelerator chips (CS-3/WSE) for training and inference workloads. It competes with Neuchips in the broader custom AI silicon market and pursues similar power-efficiency narratives for data center AI.
• Hailo: Hailo makes edge AI inference processors (Hailo-8/Hailo-15) for industrial, automotive, and IoT applications, overlapping with Neuchips' Blue Magpie NPU IP and edge/industrial positioning (NeuSight, Vecow partnerships).

Emerging players

• Graphcore: Graphcore's IPU (Intelligence Processing Unit) targets AI training and inference, offering a specialized alternative to Nvidia similar to Neuchips' ASIC approach but with a different architecture and target market.
• Lightmatter: Lightmatter develops photonic AI processors and interconnect for data center AI workloads. It competes in the same energy-efficient inference accelerator space as Neuchips with a differentiated photonic compute approach.

Broad incumbents

• Nvidia: Nvidia dominates AI compute with H100/H200/B200 GPUs and full-stack CUDA software. Neuchips benchmarks specifically against Nvidia (MLPerf DLRM perf/W) and competes for the same hyperscaler and enterprise inference workloads.
• AMD: AMD's Instinct MI300/MI325 accelerators and ROCm stack compete with Neuchips in data center AI inference, providing the GPU alternative that Neuchips' ASICs aim to displace on power-efficiency grounds.
• Intel (Habana Labs): Intel's Habana Gaudi and Greve accelerators compete in data center AI training/inference. As a large incumbent, Intel represents the alternative path for customers considering custom silicon vs general-purpose accelerators.

Market position

Strengths5 records

Weaknesses5 records

Competitive moat5 records

Key risks5 records

Key highlights7 records

Customer concentration

• GitHubgithub.com/neuchips-support
• YouTubeyoutube.com/channel/UC0A8_BqWjftqhSFNFuJicAQ
• LinkedInlinkedin.com/company/neuchips

Revenue estimate

Valuation estimate

Number of profiles

Profiles7 records

Funding overview

Funding rounds4 records

Investors4 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 NEUCHIPS