1. Cambricon profit for nine months rises 131%
Cambricon Technologies, the Shanghai-listed designer of artificial intelligence chips, reported in its quarterly filing on October 23 that net profit attributable to shareholders climbed 131.49% in the first nine months of 2026 to 3.715 billion yuan. Revenue for the period came to 9.526 billion yuan, up 106.76% from a year earlier, but third-quarter sales of 3.531 billion yuan landed well below what analysts had expected.
The result captures a tension that runs through the whole AI hardware industry this year. Demand for processors that train and run large language models keeps expanding, yet the supply of the components needed to build those processors has not kept pace. Cambricon, widely regarded as one of China's most prominent domestic alternatives to Nvidia, sits squarely in that gap: its income statement shows a business growing very quickly, while its balance sheet shows goods that could not yet be shipped.
Cambricon designs the chips but does not manufacture them, a model shared by most of the world's leading chip companies. That arrangement keeps the company focused on architecture and software, though it also leaves it dependent on outside suppliers for fabrication, packaging and, crucially, memory.
The nine-month profit of 3.715 billion yuan and revenue of 9.526 billion yuan show how quickly the company has scaled. A business that reports sales rising by more than 100% over several quarters is unusual at this size, and it explains why Cambricon is watched closely by investors who follow Chinese technology shares and global AI spending alike.
2. What do the third-quarter figures show?
The third-quarter figures show a company that roughly doubled its sales and more than doubled its profit, while still falling short of market expectations. Revenue for July through September was 3.531 billion yuan, an increase of 104.47% compared with the same quarter of 2025. Net profit for the quarter was 1.404 billion yuan, which is 147.74% higher than a year earlier.
Profit therefore grew faster than revenue, both over the quarter and across the nine months. Nine-month net profit rose 131.49%, against revenue growth of 106.76%. When earnings outpace sales in this way, it generally signals that the business is spreading its costs over a larger base of orders, or that the products it sells carry healthy pricing. For a chip designer, whose largest spending items are research staff and the licensing of design tools, higher volumes can lift profitability quickly once the fixed costs of developing a processor have been covered.
Even so, the headline growth rates sit alongside a sizeable disappointment. Analysts had pencilled in third-quarter revenue of roughly 6.26 billion yuan, so the 3.531 billion yuan actually reported amounted to only a little over half of that projection. The miss is large enough that it, rather than the growth rate, is the more telling feature of the quarter.
It is worth separating two measures when reading the quarter. Against the prior year, the company looks excellent: sales more than doubled and profit rose nearly 148%. Against expectations, it looks poor. Both statements are accurate, and the gap between them is exactly what happens when a forecast assumes a smooth supply chain and the real one is congested.
3. Why did Cambricon miss the revenue estimate?
Cambricon's shortfall is tied to supply rather than to a lack of buyers. Reporting on the results points to supply chain problems, and in particular a shortage of high-bandwidth memory, commonly abbreviated as HBM, which restricted how many finished products the company could ship during the period.
High-bandwidth memory is a specialized form of memory in which several layers of chips are stacked and placed beside the processor, allowing data to move between the two far faster than conventional memory permits. Modern AI accelerators depend on it because training and running large models involves constantly moving enormous quantities of data. A processor that cannot be paired with enough of this memory cannot be sold as a complete product, however strong the underlying design might be.
Only a small number of manufacturers produce HBM at scale, and orders from the biggest AI customers have strained their output. Smaller or newer buyers can find themselves waiting for allocation. For Cambricon, that bottleneck appears to have held back the conversion of orders into recognized revenue, which would explain why the quarter looks strong in year-over-year terms yet weak against forecasts that assumed smoother deliveries.
The episode illustrates why chip designers without their own factories or memory production can be exposed to events upstream. A company can win customers, finish designs and still be unable to deliver if one component is unavailable. For buyers, it is a reminder that purchasing decisions in AI hardware depend on the availability of complete systems, not only on the speed of a processor.
4. Inventory climbs above 11.6 billion yuan
The most visible side effect of the supply squeeze is on the balance sheet. By the end of September, Cambricon's inventory had risen to more than 11.6 billion yuan, a substantial increase from earlier in the year.
Inventory can mean very different things depending on its cause. In a demand slump, a rising stockpile signals unsold goods and often leads to discounts or write-downs. In a supply-constrained business, it can instead reflect partly completed products, purchased components and materials waiting for a missing part before they can be finished and delivered. The second reading fits the explanation offered for Cambricon, since the shortfall has been attributed to memory availability rather than weak orders.
That distinction is the heart of the investment debate. If the memory constraint eases, goods held in inventory could be completed and shipped, pushing revenue into future quarters. If it persists, or if customers switch to other suppliers while they wait, the inventory could become a burden. Because the company's own explanation points to supply, the stockpile is best read as potential revenue that has not yet been recognized, with the timing outside Cambricon's direct control.
5. What is driving demand for Cambricon chips?
Cambricon's own explanation for its growth is rising demand for artificial intelligence applications, with cloud-related workloads singled out. Operators of large data centers are buying accelerators in volume to serve chatbots, image tools and enterprise software that rely on large models, and those buyers want options beyond a single dominant supplier.
In China, that preference is reinforced by policy. Washington has restricted the export of the most advanced AI processors to Chinese customers for several years, which has pushed domestic companies to look for locally designed hardware. Cambricon is one of the businesses positioned to benefit, because its products are built for the training and inference of large models and are developed within the country's own technology ecosystem.
The company also continues to spend heavily on research and development, with its engineering focus on chips and software tuned for large models. Software matters as much as silicon in this market, since developers choose hardware partly on how easily their existing code runs on it. Years of investment in a software stack is one reason a domestic challenger can win orders when imported alternatives are constrained.
The practical result is a market in which several domestic suppliers are racing to meet demand that, for the moment, exceeds what the supply chain can deliver. That backdrop helps explain why Cambricon's revenue could double even in a quarter that disappointed analysts: the opportunity is large, and the limiting factor is how fast products can be built and shipped.
6. How investors may read the results
Cambricon's shares trade in Shanghai under the ticker 688256. For investors, the nine-month figures offer several encouraging signals at once: revenue has more than doubled, profit is growing faster than sales, and the company's own account of its slowdown is about supply rather than customer interest.
The caution is that forecasts matter to share prices as much as growth does, and a revenue figure of barely half of what analysts expected is a large gap by any standard. A company can post excellent annual growth and still disappoint if the market had priced in something even better. The distance between 3.531 billion yuan and 6.26 billion yuan shows how aggressively expectations had been set for the quarter.
Cambricon's results also say something about the broader AI hardware cycle. The binding limit on growth has shifted, at least for now, from how many designs a company can sell to how many complete systems the supply chain can deliver. Companies with secure memory allocations are in a stronger position than rivals who lack them, and that advantage can show up quickly in quarterly numbers.
Investors should also keep the size of the swing in mind. A single quarter in which sales come in near 3.5 billion yuan instead of the 6.26 billion yuan that analysts anticipated can move forecasts for the whole year, since the shortfall must either be recovered later or accepted as lost. Which of those outcomes occurs will depend largely on memory supply.
7. What to watch after the third quarter
The most important indicator is the inventory line. If high-bandwidth memory supply improves, the more than 11.6 billion yuan of stock could turn into shipments and sales, which would make the third-quarter miss look like a timing issue. If inventory keeps rising while revenue stays flat, concern about the company's ability to convert orders will grow.
A second item to follow is the fourth-quarter result, when the company's next filing will show whether delayed deliveries have started to flow through. Analysts will also look at whether profit growth continues to exceed revenue growth, and at any further commentary on how Cambricon is securing memory supplies.
For now, the picture is of a fast-growing Chinese chip designer whose reported results reflect strong appetite for AI computing and a supply chain that has not yet caught up. Whether the company can turn a growing warehouse of products into revenue is the question the next reporting period is likely to answer.