Skip to main content

Trace Intelligence

Category: AI Chips / Semiconductors

A Shanghai-based AI inference chip startup developing agent-dedicated silicon using an AI-driven Agentic EDA platform, founded by a Fudan University PhD in May 2026. Trace Intelligence was founded in 2026. The company is led by 潘鸿洋 (Pan Hongyang). Based in Shanghai, China. Team size: 11-50. Total funding raised: 10571300. Latest round: Seed. Key investors include 砺思资本 (Monolith), 启盈同创基金 (Qiyin Tongchuang).

Founded
2026
Headquarters
Shanghai, China
Team size
11-50
Total funding
10571300

Value proposition

Trace Intelligence builds custom AI inference chips for agent applications using an AI-driven EDA platform (Agentic EDA) that compresses chip design cycles from 12-18 months to 3-6 months, enabling rapid adaptation to fast-evolving model architectures and fragmented inference scenarios.

Products and solutions

Agentic EDA Platform (AI-driven chip design automation from spec to GDS), First-gen Agent Inference Chip (targeting tapeout end of 2026, entirely AI-designed)

Unique value

AI-native chip design that compresses traditional 12-18 month design cycles to 3-6 months, with each tapeout improving the next through a data flywheel — enabling custom inference chips that keep pace with rapidly evolving AI models.

Target customer

AI hardware companies, cloud data center operators, edge device manufacturers (smart glasses, AI PCs, robots, autonomous driving systems), and enterprises deploying agent-based AI workloads.

Industries served

AI Hardware, Cloud Computing, Edge Computing, Consumer Electronics, Automotive (Autonomous Driving), Robotics

Technology advantage

Proprietary Agentic EDA platform that uses AI agents across the full chip design flow — from design specification to layout (GDS). The platform has been validated through real tapeouts and already generates commercial revenue. Each tapeout creates a data flywheel, making subsequent designs faster. The methodology was validated by winning the NVIDIA-sponsored MLCAD 2025 international competition.

How they differentiate

Unlike competitors who build static chip architectures, Trace Intelligence uses its AI-driven Agentic EDA platform to enable rapid custom chip iteration, compressing design cycles from 12-18 months to 3-6 months. Each tapeout feeds data back into the AI system, creating a compounding advantage — the more chips delivered, the faster the next one ships. The company targets agent-specific inference (serial, low-latency workloads) that GPUs handle inefficiently.

Main competitors

Ricursive Intelligence (AI-driven chip design, $335M raised, $4B valuation), NVIDIA (with Groq IP acquisition and Vera CPU for agents), Cambricon Technologies (Chinese AI inference accelerators)

Key partnerships

Fudan University State Key Lab of Integrated Circuits and Systems (academic support), Multiple smart hardware manufacturers (joint R&D partnerships mentioned in articles)

Major milestones

Founded May 2026, MLCAD 2025 competition champion (founder), Agentic EDA platform already generating commercial revenue, Chip design methodology validated through real tapeouts meeting expectations, First AI-designed agent inference chip targeting tapeout by end of 2026, New funding round underway

Market positioning

Early-stage startup pioneering AI-driven chip design methodology for the specific emerging market of agent-dedicated inference silicon, competing against both mega-cap incumbents (NVIDIA) and well-funded AI-EDA startups (Ricursive Intelligence). Differentiates via full-flow AI-driven design enabling rapid iteration.

Geographic focus

China (domestic AI chip ecosystem), with global relevance in the AI inference chip market

About 潘鸿洋 (Pan Hongyang)

PhD candidate at Fudan University, State Key Lab of Integrated Circuits and Systems; Founder of PhySyn (复芯智感), an AI-driven EDA tools company; Former intern at EDA companies including 合见工软 and 鸿芯微纳; MLCAD 2025 international competition champion (NVIDIA-sponsored); Published 16+ academic papers in EDA; M.S. from Ningbo University, B.S. from Ningbo University