BlockBeats news, November 18: Arthur Hayes posted that recently many individual and enterprise-level FLOP miners have consulted him about hardware selection and performance optimization plans, but he is unable to provide specific configuration recommendations. The block rewards for FLOP miners depend on the amount of inference tasks they process, measured in floating point operations (FLOPs). For example, if a miner processes 100 GFLOPs of inference tasks within a block, while the total network processing volume is 200 GFLOPs, then the miner can receive 50% of the block reward.
Hayes reminded that hardware performance is not the only factor determining miner profits; actual market demand for AI models is equally important. If the hardware deployed by a miner supports a certain AI model, but that model lacks market demand, the equipment will be unable to obtain inference tasks and thus unable to generate revenue. He suggested that miners refer to popular model data on platforms such as OpenRouter to anticipate model demand for Agentic tasks, and continuously adjust hardware configurations through actual testing and revenue performance.
