Intel and AMD are negotiating long-term supply agreements, contracts spanning one to more than two years that lock in purchase volumes while letting prices float with the market, with major Chinese server customers, according to a Reuters report relayed by semiconductor analyst Jukan on X. The deals respond to a real squeeze. Some server CPU prices in China have risen more than 10% month over month and over 40% since the start of the year, while lead times for some Intel products have stretched to as long as six months. Investor Serenity flagged the same report and placed it inside a much bigger pattern. Long-term agreements, once rare outside slow-cycle industries, have spread this year from memory chips, where Micron, Samsung and SK hynix have signed multi-year DRAM and NAND deals, into photonics parts at Lite and Coherent, and now into CPUs and reportedly even passive components like MLCCs. Serenity's point is that a bottleneck this wide across so many different parts makes it harder to dismiss the AI buildout as a bubble, since suppliers sitting on years of contracted, pay-regardless-of-use demand do not behave like companies riding a fad that could vanish overnight.
Why it matters · A supply squeeze broad enough to reach ordinary CPUs, not just scarce AI chips, suggests the current hardware shortage is structural rather than a one-off memory story, which changes how investors should price both chipmakers' revenue visibility and the risk of an AI capex bubble.
Worth asking · If long-term, pay-regardless contracts are now spreading from memory to CPUs and even passive parts, is that proof the AI hardware boom has become structural demand, or just a bigger bubble buying itself more time?