$SNDK launch of HBF memory This weekend, FMS 2026 (Future of Memory and Storage) unofficially started in Santa Clara, California, where SanDisk and $SKHY held presentations for industry experts, engineers, and analysts. There they presented their new HBF memory. What is HBF? HBF is the solution before the memory wall within AI data centers. HBM (DRAM): Extremely fast, but has very low storage capacity and is gigantically expensive. Thoughts? eSSD (NAND): Gigantic storage capacity at a low price, but too slow to be fed directly to GPUs during ongoing AI computations. HBF (New category): Combines the capacity from NAND with the parallelism from HBM. It allows AI servers to keep enormous language models, KV-cache, and RAG databases active in memory at a fraction of the power consumption and cost of DRAM. What it does: 50–70 % lower TCO (Total Cost of Ownership): Which includes reduced production cost per generated token by up to 50 % to 75 % compared to a pure HBM/DRAM infrastructure. 60–70 % lower power consumption with significantly lower cooling requirements. Conversation context becomes longer: Unlimited KV-cache: Enables enormous context windows which includes remembering entire books, codebases, and long conversation histories without performance degradation or memory crashes. 10–20x higher concurrency: A single server can serve significantly more parallel users and AI agents simultaneously. New higher price target: This has been developed in collaboration with SK Hynix but SanDisk will be 6 months ahead with this technology. HBF will probably be presented at their next Q report which will adjust future free cash flows and give a significantly higher price target. I believe we will see an explosive rise of SNDK when HBF is presented.