Table of Contents

• You might want to know


• Main Topic


• Key Insights Table


• Afterwards...

You might want to know

1. What drove Cambricon’s A-share market capitalization to exceed RMB 1 trillion within the STAR Market?

2. How will the entry of internet giants and expanding domestic GPU suppliers reshape competition in the Chinese AI chip ecosystem?

Main Topic

On June 30, Cambricon (stock code SH688256) briefly pushed its A-share market capitalization above RMB 1 trillion, making it the first STAR Market-listed company to reach that milestone relying solely on its A-share valuation. This intra-day market cap landmark came amid a broader surge in investor interest in AI compute chips and followed a strong set of annual results that marked the company’s first full-year profitability since listing. Market data through the close of trading on June 30 also show that Cambricon’s A-share market value placed it among the top ten by market capitalization across all A-share listings, with a year-to-date share price increase exceeding 75%.

Cambricon’s 2025 annual report was a key proximate driver for investor enthusiasm. The firm reported full-year revenues of RMB 6.497 billion, representing a year-over-year increase of 453.21%, and disclosed net profit attributable to shareholders of RMB 2.059 billion — its first profitable full year since listing. These figures provided a fundamental rationale for re-rating, shifting market focus from doubts over when the company might reach sustained profitability to expectations about the addressable market for AI compute and the company’s role in it.

The macro and industry context helps explain why Cambricon has captured investor attention. Global AI investment has accelerated, and at the core of much of that investment sits compute — specialized chips that power model training and inference. Analysts and industry experts have pointed out that compute demand is being driven by two parallel trends: the rising scale and complexity of large language and generative models, and the broader commercialization of AI that moves models from research toward production workloads. As models grow larger and more demanding, underlying compute requirements escalate rapidly, creating increased demand for AI accelerators and related systems. Industry forecasts project substantial multi-year growth for intelligent compute capacity within China, with quoted compound annual growth rates in the range of 33.9% for the period 2022–2027 in some market studies, and projected EFLOPS-scale capacity expansions by 2027.

However, the evolution of competition in the domestic AI chip industry now extends beyond raw chip performance. Over the past two years, international and domestic competition focused on model parameter scale, training compute, and high-end GPU supplies. As AI adoption shifts from model breakthroughs to large-scale commercial deployment, the critical battleground increasingly includes inference efficiency, systems integration, and end-to-end solution capabilities. Firms that can lower the unit cost per token, per task, and per watt of energy consumption — through coordinated optimization across model architectures, chips, memory, networking, data-center design, and scheduling software — will have an advantage in bringing AI to a broader set of enterprise and consumer use cases.

Concurrently, market dynamics in China reflect a growing ecosystem of domestic capabilities. Major internet platform companies such as Baidu and Alibaba are expanding in-house chip development, seeking tightly integrated solutions for their large-scale services and specialized application scenarios. Domestic GPU and AI accelerator vendors are broadening their product lineups, while several chip developers are pursuing or preparing public listings. These moves increase supply-side capacity and intensify competitive pressures, while also enriching the domestic compute ecosystem — a shift that can expand market opportunities but also raises strategic dependence risks for individual vendors.

Cambricon’s management has emphasized the firm’s positioning as a provider of general-purpose AI accelerators and a platform that spans cloud, edge, and endpoint deployments. In public investor communications, the company highlights a product and software stack that targets diverse AI workloads — including vision, speech, natural language processing, and other traditional machine-learning tasks — claiming breadth of applicability across multiple industry scenarios. This breadth is important because customers increasingly evaluate chip vendors not only on peak throughput or benchmark results, but on ecosystem support, software maturity, toolchains, and integration capabilities that reduce time-to-production and operational cost.

Yet risk factors remain. Industry commentators and experts warn that Cambricon’s current performance is somewhat concentrated with respect to major customers; reliance on a few large buyers creates vulnerability if demand patterns shift, if customers decide to develop in-house silicon, or if the AI investment cycle cools. Some analysts describe the company’s valuation as elevated relative to potential downside scenarios. Moreover, as internet giants continue to develop bespoke chips optimized for their proprietary stacks or specific application classes, competition will be differentiated: bespoke chips may win in vertically integrated use cases, while general-purpose accelerators compete on flexibility and a broader addressable market.

The competitive landscape therefore appears to be moving into a new phase: from a narrow focus on device-level performance to an integrated competition among chip architecture, software ecosystems, systems engineering, and industry-specific solutions. Achieving sustained commercial success will likely require companies to demonstrate not only superior hardware performance metrics but also robust software tooling, partnerships across cloud and data-center providers, and the ability to support customers through comprehensive system-level deployments.

Finally, the appearance of a STAR Market company crossing the RMB 1 trillion threshold is also a market-signaling event. Observers interpret it as evidence of heightened investor appetite for semiconductor and AI-related technology stocks in China, and as validation of regulatory approaches that allowed emerging, capital-intensive, and initially unprofitable hard-tech enterprises to list under a registration-based IPO regime. For the sector, this may translate into easier access to capital for scaling product road maps, supply chains, and ecosystem programs — but it will also create pressure to deliver on the high expectations that accompany premium valuations.

In sum, Cambricon’s milestone is both a reflection of the company’s recent financial progress and a marker of wider industry trends: accelerating AI compute demand, a deepening domestic ecosystem, and a shift in competitive emphasis toward integrated systems and software-enabled differentiation. The coming years should test which players can convert technological potential into durable commercial positions across cloud, edge, and enterprise deployments.

Key Insights Table

Afterwards...

Looking forward, the domestic AI compute market is likely to see increased capital allocation, more entrants, and intensified competition across hardware and software layers. Firms that can integrate chip design with mature toolchains, ecosystem partnerships, and system-level cost advantages should be better positioned to capture the expanding addressable market. Investors and industry participants will closely watch customers’ procurement choices, internet giants’ internal chip programs, and how effectively vendors translate benchmark leadership into consistent production deployments. The next phase of competition will test whether emerging leaders can sustain growth, diversify customer bases, and evolve from technology pioneers into comprehensive solution providers.