Demand for AI computing power in China rose 417% from a year earlier in the first quarter of 2026, while supply grew 128%, according to the state-backed China Academy of Information and Communications Technology, as reported in a Caixin cover story published on 31 August. By our arithmetic, demand growth was about 3.3 times supply growth.
Third-quarter filings from China’s chipmakers this month show fast sales growth. Moore Threads, a designer of graphics processors that competes with Nvidia, reported quarterly revenue up 1,773.92%, about 19 times a year earlier by our arithmetic. Cambricon’s revenue rose 104.47%, but Bloomberg reported that it fell short of estimates because of manufacturing constraints and memory chip shortages. CXMT’s filing cites short supply of DRAM.
Short explainer · China AI compute demand up 417%
A vertical news Short summarising this story. Figures come from Caixin (China Academy of Information and Communications Technology data) and the companiesFigures come from the Meta AI and UC San Diego paper on arXiv; our arithmetic is labelled.#8217; Shanghai filings; our arithmetic is labelled.
Business Pill · The Memory Wall
A short Business Engineer explainer on the memory wall. It teaches the general idea only and does not discuss these companies or figures.
The key insight: China’s first-quarter AI computing demand rose 417% against 128% for supply, per the state-backed academy Caixin cited, so demand growth was about 3.3 times supply growth by our arithmetic; Moore Threads then reported third-quarter revenue up 1,773.92%.
Caixin wrote that “the sector has transformed into a sellers’ market, where securing production capacity essentially guarantees a steady flow of customers”. The figures it cited cover the first quarter of 2026 and come from the China Academy of Information and Communications Technology, a state-backed research body.
According to Caixin, the shortage forced leading AI model developers, including Z.ai, Moonshot AI and Alibaba, “to pause or limit new subscriptions for individual users between April and July”. The rest of the Caixin article is behind a paywall, and we have not read it.
Put another way, first-quarter demand was 5.17 times the year-earlier quarter and supply 2.28 times, by our arithmetic. If both are measured from the same base, demand rose about 2.3 times as much as supply.
China’s first-quarter growth in AI computing demand and supply (China Academy of Information and Communications Technology data, via Caixin, 31 Aug 2026) beside third-quarter revenuegrowth at Moore Threads, CXMT and Cambricon, from their Shanghai filings. Chart by FourWeekMBA.
What Was Posted on X
โChinaโs homegrown AI chipmakers are experiencing an unprecedented boom as surging demand for computing power collides with U.S. export controls.
The sector has transformed into a sellersโ market, where securing production capacity essentially guarantees a steady flow of customers.
On 31 October, @rmantri posted the opening of Caixin’s August cover story on X, including the 417% and 128% figures, with a link to the article.
What the Third-Quarter Filings Show
Moore Threads reported third-quarter revenue of 1,552.31 million yuan in its filing on the Shanghai Stock Exchange’s STAR market, up 1,773.92% from a year earlier. Net profit attributable to shareholders was 49.26 million yuan, and research spending was 36.79% of revenue in the quarter, the filing says.
Cambricon reported third-quarter revenue of 3.53 billion yuan, up 104.47%, and net profit attributable to shareholders of 1.40 billion yuan, up 147.74%, in its Shanghai filing. Bloomberg reported estimates of about 6.26 billion yuan and wrote that the result reflected “persistent challenges with manufacturing constraints and memory chip shortages”. Revenue was about 56% of that estimate, by our arithmetic.
CXMT, the Chinese DRAM maker, reported third-quarter revenue of 118.71 billion yuan, up 613.18%, and a gross margin of about 89.53%. Its filing attributes the rise to fast-growing global demand for computing and capacity shifts by major manufacturers, which left global DRAM in short supply, so that its sales volumes and prices both rose sharply.
The Business Lever: Capacity Sets the Price
In Caixin’s account, customers follow capacity. CXMT’s filing says short supply lifted both its volumes and its prices, and Bloomberg tied Cambricon’s shortfall to manufacturing constraints and memory chip shortages.
As we read it, the August figures and this month’s filings describe the same market from two ends: demand counted by a state research body, and sales counted by the companies.
As we read it, the constraint in these sources is supply: CXMT’s filing cites short supply of DRAM, and Bloomberg tied Cambricon’s miss to manufacturing and memory limits.
As we read it, the open question is whether the first-quarter gap has narrowed; the sources we read give no newer reading of demand or supply.
Caixin Global, 31 Aug 2026
“The sector has transformed into a sellers’ market, where securing production capacity essentially guarantees a steady flow of customers.”
Three Implications
FOR CHIP BUYERS Caixin reported that Z.ai, Moonshot AI and Alibaba paused or limited new individual subscriptions between April and July because of the shortage.
FOR CHIPMAKERS Cambricon’s third-quarter revenue was about 56% of the estimate Bloomberg cited, by our arithmetic, and Bloomberg linked the miss to manufacturing constraints and memory chip shortages.
FOR MEMORY MAKERS CXMT reported a third-quarter gross margin of about 89.53%, and its filing says short supply lifted both its sales volumes and its prices.
The framework argues that physical constraints are “creating geopolitical leverage that operates independently of who has the best models.”
As we read it, China’s chip shortage is a case of that: Caixin describes a market where production capacity, not modelquality, decides who gets customers. The framework is the owner’s argument and does not discuss these companies or figures.
What Is Not Established
The 417% and 128% figures are first-quarter data, reported by Caixin in August; we have not seen the China Academy of Information and Communications Technology’s own release or a newer reading. The academy’s method for measuring computing demand and supply is not described in the part of the article we read.
The company figures are unaudited quarterly results. Moore Threads’ growth starts from a small base of about 82.8 million yuan, by our arithmetic. The filings do not say how much of each company’s revenue comes from AI data centres.
This Week In The AI Supercycle
Anthropic’s Reality Check, The Capital Squeeze, & The New World Order
This week’s issue: for two years the AI build-out was priced as a growthstory. This week it started being priced as a credit story.
A state-backed research body put China’s first-quarter AI computing demand growth at 417% against 128% for supply. Moore Threads’ third-quarter revenue rose 1,773.92%. CXMT’s filing cites short supply, with revenue up 613.18% at an 89.53% gross margin, and Bloomberg tied Cambricon’s miss to output constraints.
A note on sourcing. The demand and supply figures come from the China Academy of Information and Communications Technology as reported by Caixin Global on 31 August 2026; company figures come from the third-quarter reports Moore Threads, Cambricon and CXMT filed in Shanghai (unaudited) and from Bloomberg. We did not contact the companies or the academy.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
