Welcome back! This is Canaan's weekly update on bitcoin mining, energy, and compute infrastructure.

District heating is one of those pieces of energy infrastructure that most of the world never thinks about that much. In Iceland, district heating supplies hot water to 95% of the population. Almost all of it comes from geothermal energy, and almost none of it touches a domestic boiler. Across the Nordics, the model is similar, with cities such as Helsinki, Stockholm, and Copenhagen running some of the most sophisticated heat grids on the planet. The IEA estimates district heating now meets around 10% of global building heat demand, and the share is rising.

With the pipes already in the ground, the heat exchangers in the basements, and municipal utilities already able to run the network, the only thing missing is a better source of heat. Currently, 90% of district heat is still produced using fossil fuels, mostly coal and gas. The EU's new rules are increasingly requiring decarbonization of district heating, and waste heat from data centers is being recognized as a qualifying input.

Any data center (Bitcoin mining, AI inference, or traditional cloud) already converts almost all of its electricity into heat. That heat has to go somewhere. In the conventional design, the input (electricity) and the output (heat) are treated as separate problems: one is a cost line on the operating budget, the other is something to be moved away from the equipment as quickly as possible. District heating reframes those two as a single system.

Bitcoin mining hardware is unusually well-suited because modern liquid-cooled miners can deliver hot water above 70°C, mining containers can be sited inside or adjacent to a town, near the heat offtake, and the mining loads can be modulated, paused, or relocated more flexibly than most other compute workloads.

Yesterday, Canaan announced an 8 MW hash-to-heat deployment with a Nordic district heating operator, after being selected through a competitive bid process. About 2 MW (228 hydro-cooled units) was already operating and feeding hot water into the local network. Now, a follow-on order placed in March will add another 6 MW (692 units), bringing the project to its full 8 MW configuration, and heating roughly 2,800 homes. "Heat reuse is no longer an ancillary byproduct of compute," as Canaan CEO Nangeng Zhang put it in the announcement. "It is central to building a more efficient, sustainable energy future."

The project sits alongside Canaan's earlier 3 MW heat-recovery pilot in Manitoba, which sends hot water above 75°C into a commercial greenhouse. Greenhouses, district networks, aquaculture, and industrial preheating all share the same shopping list: a steady, high-temperature, dispatchable heat source. The technology to supply it is already in production.

Last week, we walked through the A3222, Canaan's 28nm chip, and the first major process shrink in the post-40nm era. This week, we look at the unit that put it to work.

The Avalon4 was the first Avalon product to break the 1 TH/s barrier on a single unit — a milestone that, in 2015, felt like the inflection point between hobbyist and industrial mining. Drawing roughly 580–680 W from a fleet of A3222 chips, it offered a step-change in efficiency over the Avalon3 generation and arrived just as small mining farms were starting to scale beyond garage setups.

What made the Avalon4 notable wasn't just the spec sheet. Canaan continued the open-hardware ethos that defined the A3255 era — A3222 chips were available individually on EHash.com, and community engineers built their own controller boards around them. BitcoinTalk contributors like Valkir and Dogie wrote setup and tuning guides that became the de facto deployment manuals for small operators standing up their first multi-rack rooms. The Avalon4 wasn't the most efficient miner of its time, but it was arguably the most legible one — a machine you could understand, modify, and repair.

It is also a useful reference point for today's heat-integration conversation. When the Avalon4 was designed, the thermal output of a 650 W miner was simply a constraint to engineer around. A decade later, the same energy flow is being designed as a deliverable — the same compute, the same heat, but now treated as one system instead of two.

Each node shrink translated directly into more hashrate per watt, and 28nm was the point at which mining began to look less like a curiosity and more like a real industrial process. But the part of the A3222 era that's easier to overlook is the distribution model. Canaan made the chip available for individual purchase, continuing the open-hardware tradition that had defined Avalon from the start. You could buy raw chips and build your own miner.

That mattered because the industry was bifurcating fast. By 2015, mining was clearly becoming an industrial business - the era of soldering boards in a garage was ending. But Canaan kept selling chips to anyone who wanted them. The community that built around that decision became part of the broader Bitcoin technical culture in ways that closed-ecosystem competitors never quite replicated.