Another implication of the Deepseek results is that inference compute becomes much more important than training compute moving forward. Colocating inference data centres with Bitcoin miners seems like a match made in heaven. Inference compute resources will be used heavily during the working day, less so at night, at which point the miners come online. Anyone thinking about this from an investment standpoint? nostr:npub1a2cww4kn9wqte4ry70vyfwqyqvpswksna27rtxd8vty6c74era8sdcw83a nostr:npub1s5yq6wadwrxde4lhfs56gn64hwzuhnfa6r9mj476r5s4hkunzgzqrs6q7z
Maybe this all changes when most inference calls are made by AI agents who don’t need to sleep…
nostr:note18e8q3rhku7s5lkjf3u8dxrja94a32v2syygq3l8m5d6zq59ymgwqth4rlf