Bitcoin hashrate estimate touched 1.44 zettahashes per second

CoinWarz’s stored Bitcoin history marks an estimated network hashrate of approximately 1.44 zettahashes per second at block 915,533 on September 20, 2025. The reading was the service’s highest modeled value in the reviewed series and illustrated the scale reached by Bitcoin mining after the network’s April 2024 subsidy reduction.

The number requires careful qualification. Bitcoin does not report how many mining machines are operating or count their hashes directly. CoinWarz inferred the rate from the network difficulty and the pace at which blocks appeared. The result was therefore a statistical estimate—not a measurement of every miner and not proof that 1.44 zettahashes were performed during each second.

What the chain establishes

Contemporaneous block records show that Bitcoin was operating at difficulty 142,342,602,928,674.9 during the relevant sequence. Blockchain.com timestamps block 915,511 at 01:43:18 UTC on September 20, while mempool.space timestamps block 915,543 at 06:12:33 UTC. Both carry the same encoded difficulty as the height identified by CoinWarz.

Difficulty expresses how much expected work miners must perform to find a valid block relative to Bitcoin’s original reference target. It remains fixed within an adjustment period and changes every 2,016 blocks according to the elapsed timestamps of the preceding period. The protocol targets an average interval of approximately ten minutes, although individual blocks can arrive much sooner or later because proof-of-work discovery is probabilistic.

One zettahash equals 1,000 exahashes, or 10²¹ attempted hashes per second. CoinWarz’s approximately 1.44-ZH/s estimate consequently represented about 1,440 EH/s under its model.

Why the milestone mattered

Hashrate is one indicator of the computational competition protecting Bitcoin’s transaction history. Sustained growth generally raises the amount of equipment and energy an attacker would need to rival honest miners. It also intensifies competition among miners because an operator with unchanged computing power commands a smaller expected share of network rewards when the aggregate rate rises.

That distinction was important in September 2025. The block subsidy was 3.125 BTC, fixed by the April 2024 halving, so expanding computational competition did not increase the subsidy available in each block. Miners also received transaction fees, but the public block record alone could not establish whether any operator was profitable. Electricity contracts, machine efficiency, uptime, financing, pool performance and bitcoin’s exchange value differed across businesses.

The record also did not establish a causal market signal. No bitcoin price, return or trading-volume claim is used here because hashrate estimates and spot-market prices have different measurement systems, and the reviewed records do not demonstrate that the September 20 reading caused a contemporaneous price movement.

The measurement limitation

CoinWarz explains that its live and historical figures are inferred from difficulty and observed block timing, and that short observation windows are noisy. The accessible historical record identifies block 915,533 and the approximately 1.44-ZH/s peak but does not disclose enough underlying observations to reproduce the precise estimate independently.

Mining luck can make blocks arrive unusually quickly even if physical capacity is unchanged. A short-window record can therefore exceed a longer moving average by a wide margin. The defensible conclusion is that CoinWarz’s model registered a record estimate on September 20—not that Bitcoin’s unknowable instantaneous hashrate was exactly 1.44 ZH/s.

Later context

Fidelity Digital Assets subsequently reported, using Glassnode data through September 25, 2025, that Bitcoin’s smoother 30-day mean hashrate had crossed 1 ZH/s on September 19. That later analysis supports the broader conclusion that mining capacity was operating at unprecedented scale, while also showing why the sharper September 20 estimate should not be treated as a directly observed constant.