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

The Department of Energy marked the first anniversary of President Trump's executive orders on nuclear last week, and the list of milestones is quite interesting. TerraPower broke ground on the Natrium project in Kemmerer; DOME, the world's first microreactor test bed, opened at Idaho National Lab; Holtec began drawing on DOE financing to restart Palisades; and TVA filed a construction permit application for a GE-Hitachi BWRX-300 in Oak Ridge.
 

The stated goal is 300 GW of new nuclear capacity by 2050, which would take the US fleet from roughly 100 GW today to 400 GW. That certainly deserves the attention it's getting, but the timeline matters more than the headline here.
 

https://x.com/ENERGY/status/1930701215941202353

The big hyperscaler Power Purchase Agreements of the last 12 months have been with the existing fleet, not with reactors that haven't been built yet. Vistra has signed roughly 3.8 GW of long-dated nuclear PPAs — about 1.2 GW with AWS at Comanche Peak in Texas, and 2.6 GW with Meta across Perry, Davis-Besse, and Beaver Valley in PJM, with deliveries phased from 2026 through 2034. Constellation has a 20-year, 1,121 MW PPA with Meta for the entire Clinton output starting in June 2027, and is drawing on a $1 billion DOE loan to restart Crane behind a Microsoft offtake. These are durable, multi-decade commitments — and importantly, they all run on reactors that already exist.
 

However, large light-water reactors in OECD jurisdictions remain 7–10 year projects from siting through commercial operation. Gen IV designs and Gen III+ SMRs — Natrium, BWRX-300, NuScale, Kairos Hermes — are still pre-commercial; first electrons are penciled in for the early 2030s in optimistic scenarios. The DOE's UPRISE initiative, designed to chase the fastest-to-grid capacity from uprating existing plants, is targeting 2.5 GW by 2027 and 5 GW by 2029. When compared to the 300 GW goal of new nuclear capacity by 2050, that number doesn’t sound as big.
 

And with hyperscaler capex set to exceed $600 billion in 2026, these campuses need to go from groundbreaking to first watts in 12–36 months. These capex decisions are made today for power that needs to flow in the next few years. Meta's 1 GW Prometheus supercluster in Ohio should come online this year; its 5 GW Hyperion campus in Louisiana scales through 2030. Bloomberg, citing Sightline Climate, puts transformer lead times at up to five years against those 18-month build cycles.
 

When looking at the numbers, it becomes clear that nuclear cannot anchor the 2026–2030 hyperscaler buildout by itself.

https://x.com/jackprandelli/status/2050329015937089805

Even if every SMR program hits its current target, the next four years will be carried by gas peakers, existing-asset restarts, behind-the-meter generation, and flexible compute that can absorb stranded, off-peak, or curtailed power. That last category is the one most often left out of the conversation, and it's where Bitcoin miners stop being a side character (and instead become a precious ally for grids).

https://x.com/matthew_sigel/status/2026363359025639851

Nuclear will probably be America’s answer to the demand for the 2030s and 2040s. But for the back half of this decade, the answer is going to be everything else. The operators best positioned to absorb the bridge are the ones who spent the last ten years learning how to put compute behind the meter.

In the News

• NextEra Energy and Dominion Energy Outline Post-Merger Leadership and Structure

• Core Scientific appoints former Equinix CEO Steve Smith to board

• Applied Digital Inks $7.5B Lease for 300 MW at Polaris Forge 3, Expanding AI Factory Footprint

• The AI Hedge Fund Hunting for Bitcoin Miners' Hidden Megawatts

• Mining on Mars? OG Bitcoin Miner Chun Wang Books SpaceX Starship Flyby

• New West Data Eyes Alberta Oil Deal to Power AI and Bitcoin Mining Expansion

Network at a Glance

• BTC price (USD): ~$75,175

• Network hashrate: ~990 EH/s

• Difficulty: 136.61 T

• Hashprice: ~$34.79 / PH / day

Project Spotlight

After the Avalon4 and Avalon5 generations established 1 TH/s as the industrial baseline, Avalon6 pushed the ceiling to ~3.5 TH/s — roughly a 3× step in raw hashrate per box, paired with meaningful efficiency gains. By the time it shipped, mining had moved decisively beyond the hobbyist era and into multi-megawatt deployments.

• Chip: A3218 (the successor to the 28nm A3222)

• Hashrate: ~3.5 TH/s

• Power: ~980W

What made Avalon6 notable wasn't only the spec jump. By this generation, mining infrastructure had standardized — containers, megawatt-scale sites, dedicated substations. But Avalon6 also found a second life with smaller operators. Its reliability profile and relatively contained noise envelope made it the unit of choice for shop-floor and home-scale builders who wanted real hashrate without the operational overhead of running an industrial farm. That parallel community — operators with one to twenty boxes, not ten thousand — has stayed with the Avalon brand through every generation since, and it's the thread that connects to where home mining is heading today.

Follow and Contact Us

• X: @canaanio

• Youtube: Canaanmining

• LinkedIn: Canaan Inc.

• Website: canaan.io

• Support: Customer Care

Disclaimer: This newsletter shares industry commentary and third-party news for informational purposes only. The views and opinions expressed by third-party sources are those of their respective authors and do not necessarily reflect the views of Canaan Inc.  For official news, please refer to Canaan’s press releases and SEC filings at https://investor.canaan-creative.com/.