Canon said on 27 October 2026 that since August, shortages of printed circuit boards and the electronic components used in them have emerged as a result of growing AI-related demand. In its third-quarter results presentation, the company said the shortages had a significant impact on production and led to missed sales opportunities, and it raised the estimated sales impact on its medical business to about 90 billion yen, from about 45 billion yen.
A day later, SK hynix said long-term agreements with customers now cover more than 50% of its production capacity for 2027 and 2028. As we read it, together with other company statements and reports from 26 to 28 October, the two show demand for AI hardware reaching from memory makers to finished products.
Short explainer · The AI supply squeeze
A vertical news Short narrated from the published text of this piece and related FourWeekMBA pieces.
Business Pill · The memory wall
An evergreen Business Engineer explainer on why memory, not only compute, limits AI performance. It explains the general idea only and does not discuss Canon or the companies in this piece.
The key insight: As we read it, Canon’s presentation and SK hynix’s release describe the same pressure from two ends: a memory maker with more than half of its 2027 and 2028 capacity under long-term agreements, and a medical equipment business that could not procure the circuit boards it planned for.
Canon listed three causes for the delays in its medical business: disruption following a system update, a fire at a supplier’s facility, and supply constraints for printed circuit boards and electronic components. It had expected production to return to normal in September; it now expects normalization from 2027 onwards.
The company also said higher memory prices reduced its third-quarter profit by 19.8 billion yen. “At Canon, we have already secured the memory components we need for this year, and fourth-quarter procurement prices remained at roughly the level we had anticipated,” the presentation says.
One sourced figure at each step from memory to finished devices, 26-28 Oct 2026: SK hynix long-term deals; CoWoS price rises being prepared (Commercial Times; TSMC has not confirmed); Seagate nearline allocation; Canon’s medical sales impact (board shortages are one of three causes); Samsung delivery trims (The Korea Herald). Steps are our grouping; figures measure different things and are not added.
The Chain, Step by Step
Memory. SK hynix reported third-quarter operating profit of 83.3338 trillion won, an 80% margin, on revenue above 100 trillion won for the first time, and said its long-term agreements cover more than 50% of production capacity for 2027 and 2028. In China, CXMT expects to start producing HBM3E memory within the next two months, according to an Investing.com summary of a Reuters report published on 28 October.
Reuters reported that CXMT’s manufacturing yields are estimated at just 20% to 30%, compared with about 70% at SK Hynix, the summary says.
Packaging. Foundry and outsourced packaging-and-test companies are preparing a new CoWoS price adjustment they expect to be 10% to 20%, Taiwan’s Commercial Times reported on 26 October, citing supply-chain sources. TSMC has not confirmed any change to its prices.
Storage. Seagate said on 27 October that the vast majority of its Data Center nearline supply is allocated through calendar 2028, with some cloud supply agreements extending through 2030. Bloomberg reported the same day that Solidigm, SK Hynix’s US subsidiary, is in talks with the US Commerce Department to invest $45 billion in a new US chip plant.
Devices. Samsung’s mobile division has asked suppliers to trim fourth-quarter component deliveries by 20 to 30 percent from earlier plans, The Korea Herald reported on 26 October, citing industry sources. Apple has reportedly cut October orders for certain iPhone 18 Pro and Pro Max parts by 15 to 20 percent, the Herald reported. The Herald tied the cuts to soaring memory prices.
Our Map of the Week
We grouped these statements into five steps, from memory to finished devices, with one sourced figure for each in the chart above. The grouping is ours. Each figure is the source’s own and measures something different, so they are not added or compared.
Update, 28 Oct: the substrate side. Samsung Electro-Mechanics said in a regulatory filing that third-quarter operating profit rose 165.5% from a year earlier to 691 billion won and revenue rose 30.3% to 3.76 trillion won, both quarterly records, Yonhap reports.
The company attributed the result to higher sales of multilayer ceramic capacitors (MLCCs) for AI data centers, including servers and network equipment, and for cars, and said sales of flip-chip ball grid array substrates for AI data centers and automotive applications expanded. It said it aims to respond promptly to rising customer demand for package substrates used in AI data centers, per Yonhap. As we read it, that is the same AI pull on boards and substrates that Canon describes above, seen from a parts maker that gains from it.
On 27 October we also reported that Teradyne revenue jumps 70% as AI demand lifts chip testing; its CEO said demand extended from memory and hard disk drive test to high-speed interconnect. What’s new: Canon, a camera, printer and medical equipment maker, names AI-related demand as a cause of board and component shortages in its own results presentation.
The Structural Read
As we read it, the five steps in the chart are links in one supply chain: memory, packaging, storage, circuit boards and finished devices. Each figure comes from a different source and measures a different thing.
As we read it, Canon’s note separates two effects: memory prices, which it says reduced third-quarter profit by 19.8 billion yen, and board and component shortages, which it says stalled production in its medical business.
As we read it, the CXMT report shows the same demand reaching China’s memory industry, where Reuters reported yields estimated at 20% to 30% against about 70% at SK Hynix.
Canon, 3Q 2026 results presentation, 27 October 2026
“Since August, shortages of printed circuit boards and the electronic components used in them have emerged as a result of growing AI-related demand.”
Three Implications
FOR DEVICE MAKERS Canon said shortages of circuit boards and components prevented it from procuring planned volumes, and that it is pursuing every available measure, including alternative sourcing.
FOR MEMORY BUYERS SK hynix said long-term agreements cover more than 50% of its 2027 and 2028 capacity; Seagate said the vast majority of its Data Center nearline supply is allocated through calendar 2028.
FOR PACKAGING The Commercial Times reported on 26 October that packaging companies expect a CoWoS price adjustment of 10% to 20%; TSMC has not confirmed any change.
The framework argues that the binding constraint on AI scaling has shifted to High Bandwidth Memory, and puts it plainly: “Memory is becoming the new compute.”
As we read it, this week’s statements extend that constraint past the memory chip itself, to packaging, storage and the circuit boards inside finished products. The framework was published in December 2025 and does not discuss Canon, Seagate or the reports in this piece.
What Is Not Established
Canon does not say how much of the 90 billion yen comes from board shortages rather than the system update or the supplier fire. We read Reuters’ CXMT report only through an Investing.com summary. We could not read a Nikkei Asia report on circuit board shortages beyond its headline, so it is not used. No source ties Apple’s discussed $2,899 MacBook Pro price to memory costs, so it is not in this piece.
This Week In The AI Supercycle
Execution Fault Lines, Concentrated Demand, and Geopolitical Firewalls
This week’s issue: the AI build is not short of demand for its parts. It is short of cheap money, and its end revenue rests on a narrow base of large, often venture-funded buyers.
Canon said on 27 October that AI-related demand has caused shortages of circuit boards and components since August, contributing to a medical sales impact it now puts at about 90 billion yen. SK hynix said long-term agreements cover more than half of its 2027 and 2028 capacity. Seagate, the Korea Herald’s sources and Taiwan’s Commercial Times describe tightening at the steps in between.
A note on sourcing. Canon facts and quotations are from its third-quarter 2026 results presentation (27 October 2026). CXMT details are from an Investing.com summary of a Reuters report (28 October). SK hynix, Seagate, Solidigm, The Korea Herald, Commercial Times and Teradyne details are from our earlier reports, each attributed. The five-step grouping is ours. This is not investment advice.
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.
