The adjustment triggered at block 941,472, setting difficulty at 133.79 trillion. The immediate cause was mechanical: the average block time during the preceding 2,016-block epoch stretched to 12 minutes and 36 seconds, well above the protocol's 10-minute target. When blocks arrive slower than the target, the Bitcoin network automatically lowers difficulty to compensate. But the underlying cause of those slow blocks is structural. Miners are leaving.

Network hashrate, a measure of total computational power directed at Bitcoin, has dropped to roughly 942.83 exahashes per second (EH/s). That pulls it back below one zetahash per second, a threshold the network first crossed in 2025. Difficulty itself has now fallen nearly 10% below where it stood at the start of 2026, and sits about 14% below its peak of 155.9 trillion reached in November 2025.

A 14.7% rebound in February, after weather-related disruptions subsided, briefly masked the trend. That rebound has since been substantially unwound, with the year-to-date difficulty decline now reaching approximately 9.6%.

The Economics Behind the Exit

The math driving miners away from Bitcoin is straightforward. The April 2024 halving cut block rewards from 6.25 BTC to 3.125 BTC, compressing miner margins while energy costs stayed fixed. Today, hash price (revenue earned per petahash per second per day) has fallen below the $35 per petahash per second per day break-even level for many operators.

Depending on electricity rates and hardware generation, the all-in cost to produce one bitcoin ranges from roughly $38,000 to $92,000. Only miners with electricity costs below $0.08 per kilowatt-hour and the latest-generation ASICs, such as the Bitmain Antminer S21 XP or MicroBT Whatsminer M66S, remain comfortably profitable.

AI data center hosting, meanwhile, generates $200 to $500 per megawatt of energy deployed. Bitcoin mining on the same infrastructure yields $57 to $129 per megawatt. As BitGo Research put it plainly: "Profit per megawatt is currently higher for GPU hosting than Bitcoin hashing." AI contracts add a further structural advantage: they are typically multi-year agreements denominated in dollars, providing revenue stability that Bitcoin-denominated mining income cannot match.

Analysts project that for companies which have secured AI contracts, mining revenue will fall from approximately 85% of total revenue in early 2025 to less than 20% by the end of 2026, with AI hosting delivering operating margins in the range of 80 to 90%.

The scale of the corporate pivot has become difficult to ignore. By October 2025, mining companies had announced a combined $65 billion in AI infrastructure contracts. Core Scientific sold $175 million in bitcoin, drew down its treasury from a peak of 9,618 BTC to roughly 630 BTC, and converted all ten of its sites to AI colocation. Hut 8 signed a $7 billion, 15-year lease with Fluidstack, an AI infrastructure company, in a deal backed by Google.

Cipher Digital struck a $5.5 billion, 15-year agreement with AWS for 300 megawatts of capacity. MARA Holdings formed a joint venture with Starwood Capital targeting one gigawatt of AI capacity, expandable to 2.5 gigawatts.

Bitfarms' chief executive offered perhaps the most direct statement of the shift: "We are no longer a Bitcoin company."

Across the publicly listed miners tracked by CoinDesk, peak bitcoin holdings have been reduced by a combined 15,096 BTC.

What This Means Outside North America

For miners in lower-cost regions, the difficulty drop is tangible relief. Every downward adjustment increases the amount of bitcoin produced per unit of computing power. Operators who stayed the course through the squeeze now earn more per machine per day.

Ethiopia has emerged as the clearest beneficiary on the African continent, though the picture carries important caveats. The country ranks eighth globally by hashrate, contributing approximately 2.6% of total network computing power, roughly 27.5 EH/s. Electricity from the Grand Ethiopian Renaissance Dam runs as low as $0.032 per kilowatt-hour, among the cheapest in the world. A $250 million data center investment was announced in early 2026, and more than 21 mining firms, predominantly Chinese operators, work there under government agreements. Political instability and intermittent civil conflict introduce long-term operational risk that investors must weigh carefully alongside the favorable energy economics.

Kenya offers a different model. Gridless Compute partners with rural mini-grid operators using hydro, geothermal, and biomass energy. By acting as a buyer of last resort for surplus power, Gridless has helped reduce electricity costs for villages from $10 to $4 per month. The company is a founding member of the Green Africa Mining Alliance.

Nigeria is seeing activity in flared gas capture, solar-powered micro-mining through startups like Lagos-based SunCoin, and mobile-accessible mining services. The regulatory landscape across Africa is far from uniform, however. Algeria, Egypt, Morocco, and Tunisia have imposed outright bans on cryptocurrency mining and trading, limiting the continent's addressable mining territory considerably.

The picture in South Asia is more constrained. Bangladesh and Nepal have outright bans on crypto mining. India imposes a 30% flat tax on crypto income and a 1% transaction deduction at source, creating a hostile environment for any scaled operation. Pakistan faces persistent grid instability and high industrial power costs, further limiting its prospects. No South Asian country appears in the global top 20 by hashrate. The difficulty drop improves margins, but it does not overcome the regulatory and energy cost barriers these markets present.

A Concentration Risk Worth Watching

The hashrate decline has sharpened an existing concern. Foundry USA now controls 33.63% of global Bitcoin hashrate. AntPool holds 17.94%. Together, the two pools exceed 50% of total network computing power, crossing the theoretical threshold at which a coordinated 51% attack on the blockchain becomes conceivable. No attack has occurred, and executing one would be economically irrational, but the concentration is a relevant data point for developers and node operators building Bitcoin infrastructure in any region.

The next difficulty adjustment is estimated for April 3 to 4, 2026, with projections pointing to a further decline of approximately 0.7%. Whether that stabilizes or extends the current trend depends largely on how quickly the remaining bitcoin-focused miners can absorb the improved economics, and how many more of their larger peers complete their exits.