Cambricon reported that its third-quarter profit doubled and that it earned 3.7 billion yuan in the first three quarters, though its market value has cooled and a trillion-yuan valuation is now a thing of the past.

Cambricon's third-quarter report looks strong on profit alone. For the first three quarters, the company reported revenue of 9.53 billion yuan and net profit attributable to shareholders of 3.72 billion yuan. In Q3 alone, revenue of 3.53 billion yuan doubled year on year, and net profit attributable to shareholders of 1.4 billion yuan rose 147.7%, with profit growing noticeably faster than revenue.

But the market's current focus is how the revenue targets in the equity incentive plan are tracking.

In late July, Cambricon unveiled an equity incentive plan covering nearly 1,000 employees. It proposes granting 5 million shares, or 0.80% of total share capital, at a grant price of 750 yuan per share. The company-level performance threshold is revenue: 13.5 billion yuan in 2026, a cumulative 40.5 billion yuan over 2026 and 2027, and a cumulative 100 billion yuan over 2026 through 2028.

Cambricon has reached 70.6% of its 2026 target through the first three quarters. To meet it, the company needs to generate nearly 4 billion yuan in Q4, which is 440 million yuan more than the record-high Q3 and a 12.54% increase over Q3. Looking only at this year, the target does not appear out of reach.

The real test lies in the two longer-term targets. If 13.5 billion yuan is achieved in 2026, the company would need to deliver another 27 billion yuan in 2027 and reach 59.5 billion yuan in 2028.

On the last trading day before the report was released, Cambricon closed at 932.2 yuan, giving it a total market capitalization of 586.2 billion yuan. Just over three months ago, its market value briefly exceeded 1 trillion yuan. From that peak, the market value has fallen by more than 40%, and the market has begun to reprice the stock. Two years ago, investors were focused on when the company would turn a profit. Now, breakeven is no longer in question. Cambricon needs to convince the market that it can deliver on Q4 and meet its revenue targets for the next three years.

Capital consumption is rising, and the company's asset model is becoming heavier. Cambricon increasingly looks less like an asset-light design company, and how to value it is a new question for investors.

**01. Q4's 4 billion yuan depends on delivery**

For a company whose revenue doubled year on year over the first three quarters, reaching nearly 4 billion yuan in Q4 is not especially difficult. The challenge is that Cambricon's revenue recognition cycle is longer than that of most chip companies.

After a customer has computing demand, Cambricon must first order wafers from foundries, lock in HBM capacity, and then complete packaging and testing, shipping, deployment, and tuning. Only after the customer accepts the system is revenue recognized. This process can span one or even several quarters. How much revenue Cambricon can book in Q4 depends on whether the products it has prepared can complete deployment and acceptance before year-end.

In short, Q4 is a test of delivery execution.

Looking back, delivery has clearly accelerated. Revenue was close to 2.9 billion yuan in Q1, 3.11 billion yuan in Q2, and rose further to 3.53 billion yuan in Q3. To add nearly 4 billion yuan in Q4, delivery must speed up further.

*[Image caption: Key financial data. Source: Cambricon third-quarter report]*

The quality of profit has improved compared with the half-year report. Net profit attributable to shareholders for the first three quarters was 3.72 billion yuan, while net profit after deducting non-recurring items was 3.52 billion yuan. The gap between the two is less than 200 million yuan, implying non-recurring gains and losses of about 190 million yuan. Core operations accounted for the vast majority of profit.

Profitability is also improving. Gross margin held above 56.5%, up 1.2 percentage points from the same period last year. At least on the gross margin measure, this growth has not come from price cuts.

Expenses are also being managed more efficiently. Selling, administrative, and financial expenses combined totaled about 200 million yuan over the first three quarters, equal to just 2.1% of revenue. R&D spending reached 1.23 billion yuan, but its share of revenue fell to 12.9%. Revenue growth is diluting expense ratios.

Pressure is more concentrated on the balance sheet and in cash conversion. Net operating cash flow for the first three quarters was only 230 million yuan, and net outflow in Q3 alone exceeded 80 million yuan. Over the same period, cash received from sales of goods and provision of services was 10.67 billion yuan, but cash paid for purchases of goods and services reached 8.99 billion yuan. Revenue is growing, but more capital is being directed toward inventory and capacity.

Combined inventory and prepayments increased by 9 billion yuan from the beginning of the year, and together they account for roughly two-thirds of total assets. One investor following Cambricon said that a typical Fabless model should be asset-light with relatively ample cash flow, and that Cambricon increasingly does not resemble an asset-light design company.

The third-quarter report also shows that Q3 revenue rose by 420 million yuan from Q2, while accounts receivable increased by 720 million yuan. Once revenue is recognized, collections must keep pace.

The half-year report showed that at the end of June, most inventory was in raw materials and outsourced processing, with finished goods under 100 million yuan. The third-quarter report did not disclose the inventory breakdown. The key question now is whether this inventory can be delivered, accepted, and collected before year-end.

Stockpiling also carries costs. In the first three quarters, Cambricon recorded asset impairment losses of 500 million yuan, compared with 87 million yuan in the same period last year, an increase of more than 400 million yuan. The chip industry iterates quickly. If demand or product transitions move more slowly than inventory build-up, impairments could further weigh on profit.

In a report dated October 8, Morgan Stanley highlighted a key detail: Cambricon's flagship MLU692 began test deliveries to major cloud service provider customers in Q3, and Morgan Stanley expects most of the product's revenue to be recognized in Q4. If that holds, Q4 deliveries will include both the conversion of earlier inventory and the ramp-up of new products.

**02. After the trillion-yuan peak, the market is recalculating**

Profit is moving higher, so why has market value fallen from its trillion-yuan peak to around 580 billion yuan?

The investor quoted by Dingjiao One said the market's current concern is whether profit can be sustained.

First, consider how Cambricon makes money. In the first half of this year, 99.98% of its revenue came from its cloud product line. That line includes cloud AI chips and accelerator cards, as well as intelligent systems that incorporate its in-house chips. Cambricon's revenue depends heavily on cloud computing products.

Its customer base is also highly concentrated. In 2025, the top five customers accounted for 88.7% of Cambricon's sales. These customers are mainly large internet companies, telecom operators, and intelligent computing centers. This structure has the advantage of fast order ramp-ups and strong momentum, but the drawback is that performance is highly sensitive to the purchasing schedules of a small number of large customers.

Given limited supply of high-end GPUs, a large customer's first purchase may partly be a substitute. Repeat orders in the second and third rounds show that a product has truly entered production systems. For Cambricon, repeat purchases matter more than first orders, and customer stickiness depends largely on the software ecosystem.

Large models are iterated every month. If customers had to rework their underlying software stacks each time they switched models, the cost in manpower and time would be high. The smoother the model migration and the more usable the development tools, the more likely customers are to keep buying.

From this perspective, the adaptation progress Cambricon disclosed in its half-year report is meaningful. Its inference software platform has adapted mainstream open-source models including GLM, DeepSeek, Qwen, Kimi, and MiniMax. Its training software platform is also working on adaptation for DeepSeek, Qwen, and Hunyuan model series. This is effectively building the scaffolding for customers in advance, which lowers deployment barriers once they receive the chips.

But an ecosystem moat alone is not enough, and the competitive landscape is clearly tightening.

In terms of shipment volume, Cambricon remains a challenger. According to IDC, China's total AI accelerator card shipments in 2025 were about 4 million units. Nvidia shipped about 2.2 million units, or 55%, and remains the dominant player. Domestic vendors together shipped about 1.65 million units, or 41%. Among domestic vendors, Huawei led with about 812,000 units. T-Head shipped about 265,000 units. Kunlunxin and Cambricon each shipped about 116,000 units. Cambricon accounts for about 3% of the overall market and about 7% of domestic vendors.

So why was the market willing to assign it a trillion-yuan valuation? Because in the capital markets, Cambricon is something of an outlier. It is one of the few companies whose core business is cloud AI chips and that has already achieved large-scale revenue and consecutive quarters of profitability. Its Siyuan series has also entered large-scale commercial deployment in core industries, including internet companies.

The investor said that this first-mover advantage in commercialization is harder to replicate than simply leading on chip specifications. Once a major customer migrates its business systems, a newcomer must bear higher adaptation and validation costs to displace them.

But this window will not stay open for long. If challengers offer stronger performance, better software compatibility, and more attractive pricing, large internet companies have no reason to tie themselves to a single supplier.

Morgan Stanley forecasts that China's AI chip self-sufficiency rate will rise from 35% in 2025 to 68% in 2027 and 95% in 2030. The forecast is aggressive, but it points to one clear implication: many more domestic chip companies will enter the market in the coming years. The pie is growing, but more companies will be competing for it. As Moore Threads, MetaX, Enflame, and others go public, Cambricon will inevitably be compared with them more often, and its scarcity premium is likely to be diluted.

Therefore, market value does not necessarily move in step with profit. The key is whether current profit growth can deliver on the extremely optimistic expectations the market previously held.

When Cambricon's market value surged past 1 trillion yuan, investors were focused on the growth potential for 2027 and 2028. On October 8, Morgan Stanley raised its target price for Cambricon from 1,408 yuan to 1,700 yuan, and shifted its research focus to new product ramp-ups after 2027.

Now that the market has cooled down, and its focus has returned from 2027 to 2026, one question matters most: profit has already shown up, but can cash, repeat purchases, and market share keep up?

Cambricon is now in the middle of a test. In the first half, it used continuously doubling revenue and profit to push its market value above 1 trillion yuan. In the second half, the market will look at both growth and the quality of profit. Only once repeat purchases by large customers, cash collections, and market share have been validated will the market be willing to pay for the next trillion-yuan story.