Micron perps slide about 6.5% in eight minutes after DeepSeek Flash model news

Micron Technology perpetual futures dropped approximately 6.5% in eight minutes over the October 17-18 weekend. The trigger was DeepSeek and its Flash model, which is pitched on doing more with less memory. What moved the market The pressure traces back to DeepSeek’s V4.1 Flash model, released on September 10, 2026. DeepSeek claimed it cuts KV-cache high-bandwidth memory (HBM) usage by 75% compared with the previous version. A quick translation helps. The KV-cache is the model’s short-term memory. It is a running notepad of everything said so far, so the model does not reread the whole conversation before each word. That notepad usually sits in HBM, the expensive, ultra-fast memory stacked next to AI chips. Micron is one of the companies that makes it. DeepSeek also claimed an 87.5% reduction in SSD requirements versus the prior version. The V4.1 Flash model was announced around 06:10 UTC. It uses a 552B-parameter mixture-of-experts (MoE) design. Then came a second nudge. On October 17, 2026, DeepSeek pushed a minor update that improved coding benchmarks. Pricing and the context window stayed the same. A sector-wide headache Micron was not alone. SanDisk and Western Digital both ca...