November 9, 20262 min read0 views

INTERVIEW SERIES: AI’s Power Crunch Echoes Bitcoin Mining’s Efficiency Push

Key Takeaways

Bitcoin mining firmware and AI power controls address a similar challenge: increasing computing output within an electricity limit.
Cadence’s Sherman Ikemoto says lower GPU power settings can create room for more processors and greater total output.
AI operators must also account for workload…

Reading Level🟢BeginnerNo prior crypto knowledge needed

Listen to this article

Key Takeaways

Bitcoin mining firmware and AI power controls address a similar challenge: increasing computing output within an electricity limit.
Cadence’s Sherman Ikemoto says lower GPU power settings can create room for more processors and greater total output.
AI operators must also account for workload performance, cooling constraints and the cost of additional hardware.

AI operators face a power-efficiency trade-off familiar to bitcoin miners, with GPU tuning and chip-to-chiller design shaping output per megawatt.

Bitcoin miners have used firmware to adjust chip settings and extract more (or sometimes less) hashrate from limited electricity supplies. AI infrastructure operators face a familiar calculation: how to produce more computing output when additional power is difficult to obtain.

For miners, the operating measure is hashrate, or the number of calculations their machines perform each second. For AI inference operators, one measure is token throughput—the rate at which models process or generate units of data.

In both cases, pushing individual machines toward their highest output can come at the expense of electrical efficiency. The operating point that produces the most from one machine may differ from the configuration that produces the most across an entire power-constrained facility.