TeraWulf Cuts Bitcoin ASIC Energy per Hash 6% as Mining Fleet Shrinks

TeraWulf’s latest SEC filing reveals a measurable change in its Bitcoin mining hardware: reported fleet efficiency improved even as operational SHA-256 hashrate contracted. The November 6, 2026 Form 10-Q lists miner efficiency of 16.4 joules per terahash (J/TH) for September 30, 2026, compared with 17.5 J/TH one year earlier. Operational hashrate declined from 8.5 to 5.7 exahashes per second (EH/s) over the same period.

What 1.1 J/TH means

The 1.1 J/TH improvement is a 6.29% reduction in energy per unit of hashing work, calculated as (17.5 − 16.4) ÷ 17.5. At a constant 1 EH/s, the reported difference corresponds to 1.1 megawatts of mining-machine power, or 26.4 megawatt-hours over 24 hours. At $0.05 per kilowatt-hour, that would equal $1,320 in daily electricity expense, assuming uninterrupted operation and no change in supporting infrastructure.

A smaller fleet, not necessarily a faster chip

TeraWulf’s reported operational hashrate fell approximately 32.9%, from 8.5 EH/s to 5.7 EH/s. Multiplying those reported hashrates by the respective miner-efficiency figures gives illustrative mining-machine loads of 148.75 MW and 93.48 MW. The 55.27 MW difference combines changes in active hashrate and fleet efficiency; it must not be described as savings attributable solely to improved ASIC silicon. The company also reports miners at its Lake Mariner facilities spanning 15.0–23.0 J/TH, with a separate 16.2 J/TH average in the filing. These values use different reporting scopes and should not be conflated.

Why operators should distinguish wall power from facility power

J/TH describes energy consumed per unit of SHA-256 throughput. A machine-level or fleet miner-efficiency statistic does not automatically include pumps, cooling towers, transformers, networking and other site loads. Facility-inclusive efficiency depends on those loads and operating conditions. The 10-Q supplies company-reported fleet figures, not an independently reproduced bench test of one specific ASIC model. It does not establish a new chip-design record or disclose a new sub-10 J/TH mining machine.

What it means for Power Efficiency Index research

The comparison illustrates the difference between the efficiency frontier and real deployed fleets. A manufacturer-rated 8.9 J/TH machine may represent a leading product specification, while a publicly reported operating fleet can remain materially less efficient because of older installed hardware, replacement timing and economics. TeraWulf’s 16.4 J/TH should therefore be tracked as a fleet-level observation, not substituted for the best announced miner specification. A stronger next step would be to obtain fleet model mix, metered power boundaries, uptime and facility power-usage effectiveness before assigning causes to the improvement.

Source and methodology

Primary source: TeraWulf, Form 10-Q, filed November 6, 2026. Figures refer to the September 30, 2026 and September 30, 2025 reporting dates. Calculations are editorial arithmetic based on disclosed rounded figures; power and cost examples are illustrative, not measured facility totals.

Read TeraWulf’s November 6, 2026 Form 10-Q.

Leave a Reply