Alibaba, Tencent and ByteDance have asked Samsung Electronics and SK hynix for long-term supply agreements on server memory, South Korea’s Maeil Business Newspaper reported on 27 October 2026, citing the industry. Maeil reported that some customers asked for contracts running ten years, to 2035, against the usual three to five.
The report says the Chinese cloud providers are moving after US big tech companies had already locked up memory volumes through long-term agreements. Neither Samsung nor SK hynix is quoted on the Chinese requests in the report we read.
Video explainer · The AI memory squeeze
A 16:9 video explainer narrated from the published headlines of this piece and four related FourWeekMBA pieces.
The memory contracts in three numbers. Source: Maeil Business Newspaper, 27 Oct 2026.
The key insight: As we read it, the request is about time, not only volume: Maeil reports that some Chinese customers asked for contracts to 2035, against the usual three to five years, and gives no volumes or prices.
According to Maeil, the requested products are server DDR memory and LPDDR, the low-power DRAM mostly used in smartphones. The paper reported that server makers are seeking LPDDR because it can raise the power efficiency of AI servers, at a time of memory shortage.
Maeil reported that the Chinese requests are for volumes separate from the long-term contracts the two Korean companies have already signed. The report gives no volumes, prices or contract values.
China has its own memory makers, CXMT and YMTC. Maeil wrote that their product competitiveness and production scale in high-performance server memory are assessed as falling short of the Korean companies, and that the difficulty of sourcing the needed volumes at home is one reason the Chinese providers are pursuing Korean contracts.
Share of Samsung and SK hynix’s planned memory output for next year under five-year long-term agreements with global big tech: 70% to 80%, an industry estimate reported by Maeil Business Newspaper on 27 October 2026. The ~30% and ~20% remainders are our arithmetic.
Who Already Holds the Supply
Maeil reported that Samsung has said it finished contracts with its five largest global data-center customers and is negotiating with five more large customers. The paper wrote that the industry sees those customers as Amazon Web Services, Microsoft, Google Cloud, Meta and Oracle.
SK hynix has said it concluded long-term agreement talks with about ten customers, including its key ones, according to Maeil. The paper added an industry estimate that 70% to 80% of the two companies’ planned memory output for next year is volume under five-year long-term agreements with global big tech.
The Price Backdrop
TrendForce wrote on 30 September 2026 that “Server DRAM is therefore expected to remain undersupplied in 4Q26.” It projected conventional DRAM contract prices to rise 10% to 15% quarter on quarter in the fourth quarter, and NAND Flash contract prices 15% to 20%.
The same TrendForce release noted that for some suppliers, server DRAM price increases are expected to lag the market average because of pricing mechanisms set in long-term agreements.
As we read it, the Chinese buyers are arriving after the queue formed: Maeil reports that Samsung has said it finished contracts with its five largest data-center customers, and that SK hynix has said it concluded talks with about ten customers.
As we read it, the products asked for are server DDR and LPDDR, not HBM: Maeil reports that server makers are seeking LPDDR for the power efficiency of AI servers.
As we read it, the domestic alternative is the gap: Maeil wrote that CXMT and YMTC are assessed as falling short of the Korean companies in high-performance server memory.
TrendForce, press release of 30 September 2026
“Server DRAM is therefore expected to remain undersupplied in 4Q26”
Three Implications
FOR MEMORY BUYERS Maeil reports an industry estimate that 70% to 80% of Samsung and SK hynix’s planned output for next year is under five-year agreements with global big tech.
FOR MEMORY MAKERS TrendForce wrote that for some suppliers, server DRAM price increases are expected to lag the market average because of pricing mechanisms set in long-term agreements.
FOR CHINESE CLOUD PROVIDERS Maeil reports that the Chinese requests are for volumes separate from the long-term contracts the two Korean companies have already signed.
The framework argues: “Memory is becoming the new compute.”
As we read it, a request for contracts to 2035 treats memory supply as long-dated capacity rather than a quarterly purchase. The framework was not written about these requests.
What Is Not Established
The report is based on industry sources. It does not say whether any contract has been signed, which customer asked for ten years, or what volumes and prices are involved. Samsung and SK hynix are not quoted on the Chinese requests in the report we read.
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.
Alibaba, Tencent and ByteDance have asked Samsung and SK hynix for long-term server memory supply, with some seeking contracts to 2035, Maeil Business Newspaper reported on 27 October 2026.
A note on sourcing. The contract requests are as reported by Maeil Business Newspaper on 27 October 2026, citing the industry; we read the Korean original and translated it. Maeil’s statements about what Samsung and SK hynix have said are attributed to the paper. Price projections are from TrendForce’s release of 30 September 2026. This is not investment advice. We did not contact Samsung Electronics, SK hynix, Alibaba, Tencent or ByteDance.
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.
Scroll to Top
Discover more from FourWeekMBA
Subscribe now to keep reading and get access to the full archive.
