Earlier this week, a sweeping US winter storm pushed Bitcoin miners to curtail, pulling a noticeable chunk of computing power off the network in a short window.
Data shows a 40% dip in hashrate between Jan. 23 and Jan. 25, with around 455 EH/s going offline, and block production slowing to around 12 minutes for a stretch.
The fact that the sharpest drop came from Foundry USA, the largest mining pool with the largest presence in the US, tells you that the drop was caused by curtailments.
Why can so many miners now shut off quickly? Why would they ever choose to do it, and what do those choices mean for Bitcoin’s security budget, transaction flow, and the politics of plugging a large industrial load into a grid that can get stressed in extreme cold?
Curtailment 101: miners as flexible load, not fragile infrastructure
While curtailment is simple in definition, it’s kind of messy in practice. At the simplest level, it’s miners reducing electricity consumption, either partially or fully, because power is scarce, expensive, or contractually more valuable to sell back to the grid than to burn through ASICs.
In the US, and especially in Texas, that choice has matured into a full-blown business model. ERCOT has explicitly created mechanisms meant for “large flexible customers” that can reduce load during peak demand, and it named Bitcoin mining facilities as a core example.
The idea is straightforward: if a load can drop quickly, reliably, and repeatedly, a grid operator can treat it as a pressure-release valve during tight conditions.
In real mining fleets, curtailment tends to fall into three buckets.
The first is purely economic. Miners watch a simple spread: revenue per unit of hash versus the all-in cost of producing that hash. When real-time power prices spike, the cheapest decision can be to just stop hashing.
This is no charity, and it’s certainly not some kind of corporate moral stance. It’s just your basic unit economics measured minute by minute, especially for miners exposed to wholesale pricing.
The second is contracted. Some miners sign demand-response arrangements where the “off switch” is effectively part of the product they sell.
Texas has offered multiple ways for flexible load to participate in reliability programs, and the last few years have produced plenty of cases of miners profiting during stress events by curtailing or selling contracted power back into the market.
Company disclosures show miners can earn money for not consuming power when the grid is tight. In Riot’s August 2023 update, the company split the haul into two buckets: $24.2 million of “power Credits,” which it describes as power curtailment credits earned by selling contracted power back into ERCOT at market spot prices, plus $7.4 million of something called “demand response credits,” tied to participation in ERCOT demand response programs.
The smaller, routine versions of this show up in almost all of Riot’s monthly reporting. In its November 2025 update, Riot listed $1 million in estimated power curtailment credits and $1.3 million in estimated demand response credits, noting that those demand response credits came from participation in ERCOT and MISO programs and that the combined credits are netted against its all-in power cost.
Iris Energy’s investor update from August 2023 said its Texas site generated about $2.3 million in “power sales,” described as power credits primarily driven by voluntary curtailment under hedge contracts tied to ERCOT real-time prices.
In that setup, a mining site is closer to a hybrid of data center and power trader than the old mental model of a warehouse that just runs ASICs until they break.
The third is emergency or rule-driven. Texas now expects the biggest new loads to be curtailment-ready as a condition of interconnection in grid-emergency scenarios, explicitly naming crypto miners and data centers among the targets.
That matters because it turns curtailment from something nice to have into something that’s now built into the operating plan.
What makes this week’s storm useful as a teaching moment is that the incentives line up.
Cold snaps lift heating demand, which tightens reserve margins and often triggers conservation alerts. The storm was seriously disruptive for the US energy system, with price spikes and operational strain reported across regions.
So, if you’re a miner sitting on a flexible load arrangement, curtailment is often the cleanest, most rational response to a grid that is suddenly valuing a megawatt more than a terahash.
This is also why the pool-level picture can move fast. When US-heavy operators curtail, their pools register it almost immediately. The week’s curtailment effect was most easily seen through the visible drop in Foundry’s hashrate and the knock-on slowdown in blocks.
While the network is global, the marginal hashrate swing can still be regional when enough capacity clusters behind a handful of operators and grid regimes.
Bitcoin’s difficulty timer: why slow blocks are usually a temporary tax
A hashrate shock scares people because they map it directly to security. That’s true, but in a very narrow sense, because fewer hashes per second means the brute-force cost of attacking the chain is lower than it would be at peak hashrate.
But the more important operational question is what Bitcoin does when hashes disappear quickly. The answer is that Bitcoin has a built-in recalibration mechanism with a built-in delay.
Bitcoin targets one block roughly every 10 minutes, but it doesn’t adjust difficulty continuously. It adjusts difficulty every 2,016 blocks based on how long the last 2,016 blocks took to mine.
That structure creates the short-term “storm tax.” If a lot of miners shut off today, blocks will be slow today. However, the difficulty doesn’t instantly drop to compensate; the network just produces blocks more slowly until enough of them pass for the next adjustment to reprice the work.
You could see that taking place in real time this week. CoinWarz’s difficulty dashboard showed the network running slower than the 10-minute target, with average block time above target during the window it tracked.
When block production stretched to roughly 12 minutes, it was the lived experience of that lag: fewer blocks per hour, slower confirmations on average, and a mempool that can thicken if transaction demand holds steady.
But slow blocks aren’t Bitcoin “breaking,” they’re Bitcoin charging users and miners a time cost for abrupt changes in hash supply.
If the shock fades quickly and miners come back online as prices normalize and grid stress eases, the network may never need a difficulty adjustment. If the shock persists, the next adjustment will lower difficulty and pull block timing back toward the target.
The fee market can also behave in ways that confuse casual observers. A short spell of slow blocks can lift fee pressure if demand is steady, but it can also pass quietly if the mempool wasn’t tight to begin with and demand is soft.
The bigger point here is that Bitcoin’s design assumes mining power is opportunistic and sometimes transient. Difficulty adjustment is the protocol’s way of accepting that reality without turning every local infrastructure event into a systemic failure.
Winter storms as repeat stress tests: Uri, Elliott, and what 2026 adds
This isn’t the first time winter weather has affected Bitcoin. What changed is the scale of the US footprint and how integrated miners have become in grid programs.
Start with Winter Storm Uri in February 2021, the modern reference point for Texas grid trauma. Uri drove a historic demand surge while generation failed across fuel types, triggering widespread outages and a political reckoning.
Back then, large-scale bitcoin mining was far less intertwined with Texas reliability planning. The industry was smaller in-state, and the “miners as flexible load” concept was almost completely theoretical. That’s significantly different than today’s setup, where curtailment is easier to coordinate and far more common.
Uri matters for this story because it sets the political backdrop. After a crisis like that, any large new electricity user gets measured against a simple question: Will you make the next emergency better or worse?
Now jump to Winter Storm Elliott in December 2022, the episode that more directly resembles this week’s hashrate pattern. Galaxy’s 2022 mining report described Elliott as a moment when miners curtailed as much as 100 EH of hashrate, framing it as roughly 40% of network hashrate at the time, done to help stabilize the grid.
Separate academic and policy discussions have also cited the same order of magnitude, reinforcing that Elliott was a major curtailment event rather than a blip in hashrate.
Elliott is the clean comparison because it showed two things at once. First, large miners can shut off at scale on short notice during extreme cold. Second, once miners build curtailment into their commercial relationships, those shutoffs become legible and, in some cases, expected.
What does 2026 add? It adds the reality that “flexible load” is no longer mostly about miners, but about a broader class of giant compute loads.
The US Energy Information Administration has described Texas as a center of fast electricity demand growth, explicitly calling out data centers and cryptocurrency mining as major contributors and pointing to ERCOT’s task-force style oversight around large loads.
That matters because the grid politics change when flexible load stops being a niche. Once AI data centers and other compute-heavy facilities compete for the same interconnection capacity and the same public patience, miners lose the ability to argue that they’re a special case.
They become one category inside a broader debate about who gets power first during stress, and who pays for the grid upgrades needed to serve everyone.
Bloomberg’s reporting on the storm pointed in the same direction, discussing how large industrial loads, including crypto mines and data centers, reduced power use during the event and how ERCOT’s demand expectations moved as conditions evolved.
That sort of framing from the mainstream media is a reminder that the next decade of mining in the US will be narrated through grid governance as much as through Bitcoin price cycles.
So the hashrate drop this week is best read as a preview. As the US share of mining remains large and as compute loads keep scaling, weather events will keep producing these short-lived network slowdowns. The protocol can handle them. The political environment is less forgiving.
Bitcoin’s difficulty timer makes curtailment survivable for the chain, and flexible-load economics can make curtailment profitable for miners. The open question is whether regulators and residents accept the bargain: a large new load that promises to leave when asked, in exchange for the right to plug in the rest of the time.