💰钛媒体•Mar 11, 2026Stalecollected in 37m

Verisilicon AI Boom Hides Profit Struggles

Post LinkedIn

💰Read original on 钛媒体

#ai-chips #funding #ipo #profitabilityverisiliconverisilicon

💡AI chip IP leader's 1.8B raise & IPO amid profit woes—vital for hardware scaling.

⚡ 30-Second TL;DR

What Changed

Surge in AI orders driving revenue

Why It Matters

Reveals profitability pressures in AI chip IP amid demand surge, signaling potential supply chain cost optimizations for practitioners.

What To Do Next

Assess Verisilicon IP cores for cost-effective AI accelerator integration.

Who should care:Developers & AI Engineers

Key Points

•Surge in AI orders driving revenue
•Recent 1.8B funding round
•Pursuing Hong Kong stock listing
•Low-margin growth via discounts, no profit targets

🧠 Deep Insight

Background and context from public sources — not the original article. 8 sources cited.

🔑 Enhanced Key Takeaways

•VeriSilicon's 2025 full-year revenue is projected at 3.152 billion yuan, with net losses narrowing to 528 million yuan, driven by nearly 6 billion yuan in new AI-related orders accounting for 73% of total orders[1].
•Q3 2025 revenue hit a record 1.28 billion yuan, up 120% quarter-over-quarter and 78.77% year-over-year, with AI computing comprising 65% of new orders and order volume surging 145.8% YoY[2].
•In 2024, VeriSilicon reported a 605 million yuan net loss due to 53.35% R&D spending on AI and chiplet tech, with AI IP licensing contributing 40% of revenue and a 2.4 billion yuan order backlog[4].
•Company launched AI-based image processing IPs AINR1000/2000 (AI Noise Reduction) and AISR1000/2000 (AI Super Resolution) in February 2025 for surveillance, automotive, and video analytics[3].

🛠️ Technical Deep Dive

•Core IP portfolio includes six processing units: Neural Network Processing Unit (NPU), Graphics Processing Unit (GPU), Video Processing Unit (VPU), Digital Signal Processing (DSP), Image Signal Processing (ISP), and Display Processing IP[3].
•Specialized AI IPs AINR1000/2000 use proprietary AI pixel processing for noise reduction in surveillance, automotive vision, cloud gaming, and real-time video analytics[3].
•AISR1000/2000 provide AI Super Resolution capabilities, enhancing applications in the same sectors via advanced pixel algorithms[3].
•Developing UCIe-compatible chiplet solutions alongside AI IP licensing to support modular semiconductor designs[4].

🔮 Future ImplicationsAI analysis grounded in cited sources

VeriSilicon achieves breakeven in 2026

Over 5 billion yuan in hand orders from AI demand, combined with economies of scale in mass production and narrowing losses, position the company for profitability as forecasted[1][4].

R&D investments yield returns via AI IP growth

High R&D spending on chiplets, AIGC, and autonomous driving, with AI IP already at 40% of 2024 revenue, supports projected demand in cloud training, inference, and edge AI devices[4][7].