Allegation: OpenAI spiked memory prices
A viral claim alleges OpenAI placed aggressive—or 'phantom'—hardware orders that helped trigger a global memory price spike and a consumer hardware crunch. The report is circulating widely but is framed as an explanation by a popular social user rather than a confirmed industry finding. (timesofindia.indiatimes.com)
OpenAI, Samsung Electronics and SK hynix announced letters of intent in October 2025 under the “Stargate” initiative that set a target of as many as 900,000 DRAM wafer starts per month for OpenAI’s infrastructure needs. (openai.com) Those announcements were described publicly as letters of intent or preliminary agreements signed during Seoul meetings with South Korean officials, language that companies routinely use when terms are under negotiation rather than when firms issue firm, binding purchase orders. (news.samsung.com) Industry reporting and market analysis estimate the 900,000‑wafer figure could equal roughly 35–40% of certain segments of global DRAM output if fully executed, a scale that analysts warn can materially tighten consumer DRAM and HBM availability. (trendforce.com) Market effects tied to large AI allocations were visible late 2025 and into 2026: Bloomberg documented sharp spot‑price jumps and said AI buyers including major cloud players were taking large shares of production, and Micron publicly shifted away from consumer channels to prioritize AI/data‑center customers. (bloomberg.com) No major trade or financial outlet has produced evidence that OpenAI issued intentionally “phantom” or fraudulent purchase orders that were later cancelled; analysts and memory‑industry insiders attribute current shortages to capacity reallocation toward HBM/AI workloads and to large but preliminary supply agreements. (notebookcheck.net) Memory‑industry forecasts and vendor statements put meaningful relief out multiple years: SK hynix and several analyst reports say tight consumer DRAM supply could persist through 2027–2028 as new fabs and HBM yield ramps take time to deliver volume. (notebookcheck.net)
