publish date 2026-08-13
Welcome back! This is Canaan's weekly newsletter on Bitcoin mining, energy, and compute infrastructure.
In June, FERC issued show-cause orders to all six regional grid operators under its jurisdiction, directing them to justify or rewrite the tariffs governing how data centers and other large loads connect to the transmission system. The operators have until mid-August to file revised tariffs or explain why their existing rules already work. Meanwhile, the Department of Energy has invoked emergency authority against PJM four times this year, allowing the grid operator to direct any customer drawing at least 50 MW at a single delivery point onto onsite backup generation within 15 minutes of a signal.
ERCOT is not on FERC's list. Texas sits outside that jurisdiction entirely. And Texas got there first, and harder. Now, Governor Abbott is directing the PUCT and ERCOT to audit every data center advancing through the interconnection queue before any additional projects are approved. ERCOT responded with a market notice suspending Batch Zero, the first cohort under the new large load classification process, and will seek a good cause exception at the commission's August 20 open meeting. So there are projects representing over 474 GW of requests, roughly 90% of it data centers, and more than five times the grid's record peak demand.
Two separate regulatory tracks are converging on the same issue within the same eight weeks. What matters is what these processes actually test. The Texas audit is a credibility and readiness screen. Projects with existing operations or already approved capacity appear less exposed. The risk concentrates in planned deployments still waiting on a new interconnection or incremental approval, and no timeline has been set for completing the review.
Flexible load and inflexible load are being sorted into different categories, and the sorting has consequences. A facility that can drop to zero in seconds and come back on demand is a grid management resource. A facility that cannot is another line of demand on a system that has run out of headroom.
The evidence that will matter is operational, and it is specific: ramp-down time to zero load, the granularity with which consumption can be adjusted, minimum-run constraints, and documented response to curtailment instructions. The strongest version is historical curtailment data from real scarcity events, with dates, duration, and megawatts curtailed. Fleet-level telemetry that was previously an internal performance metric is becoming a regulatory filing.
There is a second reason to build that record now. In their July response to the Governor, the PUCT and ERCOT recommended that large computational loads register with both agencies and disclose technical information about their facilities, a concept Abbott has folded into his 2027 legislative recommendations. If it advances, it would reach existing operators, not only projects in the queue.
The through line across Austin and Washington is the same. Grids are moving toward a framework where a large load must demonstrate not just that it can consume power, but how it behaves once connected. Canaan's founder, NG Zhang, made a related observation, arguing the AI buildout still resembles the early years of Bitcoin mining more than most participants recognize. One of the least appreciated inheritances from those years is that mining spent a decade learning to operate as a controllable load because it had no other option.
In the News
Network at a Glance
Project Spotlight
The Avalon Immersion Cooling Miner was never just a miner. It was a rack-scale system, built as a mechatronic assembly designed to run 90 mining modules submerged in a single coordinated deployment. The modules themselves were stripped down for liquid: two hash boards delivering the compute of three air-cooled boards, because once the coolant did the work, heat stopped being the limit on density.
One specification reads differently today than it did at launch. Canaan rated the immersion configuration for overclocking of up to 38%. In 2022 that was a cost-per-terahash argument. Run the same silicon harder, get more out of the same capital.
Set it against this week's feature, and it says something else. A machine with 38% of headroom above nominal has an unusually wide operating band, and a fleet with a wide band can place load precisely rather than in on and off steps. Response granularity is exactly what grid operators are beginning to ask large loads to document. Immersion delivers it as a byproduct of solving a thermal problem.
Two other properties came free. Submerged modules do not ingest dust, removing a primary wear mechanism. And liquid carries heat somewhere useful, which is the basis of the hash-to-heat work Canaan is doing in the Nordics today with hydro-cooled A1566HA units.
Follow and Contact Us
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/.
In June, FERC issued show-cause orders to all six regional grid operators under its jurisdiction, directing them to justify or rewrite the tariffs governing how data centers and other large loads connect to the transmission system. The operators have until mid-August to file revised tariffs or explain why their existing rules already work. Meanwhile, the Department of Energy has invoked emergency authority against PJM four times this year, allowing the grid operator to direct any customer drawing at least 50 MW at a single delivery point onto onsite backup generation within 15 minutes of a signal.
ERCOT is not on FERC's list. Texas sits outside that jurisdiction entirely. And Texas got there first, and harder. Now, Governor Abbott is directing the PUCT and ERCOT to audit every data center advancing through the interconnection queue before any additional projects are approved. ERCOT responded with a market notice suspending Batch Zero, the first cohort under the new large load classification process, and will seek a good cause exception at the commission's August 20 open meeting. So there are projects representing over 474 GW of requests, roughly 90% of it data centers, and more than five times the grid's record peak demand.
https://x.com/ShanuMathew93/status/2087126229606387937
Two separate regulatory tracks are converging on the same issue within the same eight weeks. What matters is what these processes actually test. The Texas audit is a credibility and readiness screen. Projects with existing operations or already approved capacity appear less exposed. The risk concentrates in planned deployments still waiting on a new interconnection or incremental approval, and no timeline has been set for completing the review.
Flexible load and inflexible load are being sorted into different categories, and the sorting has consequences. A facility that can drop to zero in seconds and come back on demand is a grid management resource. A facility that cannot is another line of demand on a system that has run out of headroom.
The evidence that will matter is operational, and it is specific: ramp-down time to zero load, the granularity with which consumption can be adjusted, minimum-run constraints, and documented response to curtailment instructions. The strongest version is historical curtailment data from real scarcity events, with dates, duration, and megawatts curtailed. Fleet-level telemetry that was previously an internal performance metric is becoming a regulatory filing.
There is a second reason to build that record now. In their July response to the Governor, the PUCT and ERCOT recommended that large computational loads register with both agencies and disclose technical information about their facilities, a concept Abbott has folded into his 2027 legislative recommendations. If it advances, it would reach existing operators, not only projects in the queue.
The through line across Austin and Washington is the same. Grids are moving toward a framework where a large load must demonstrate not just that it can consume power, but how it behaves once connected. Canaan's founder, NG Zhang, made a related observation, arguing the AI buildout still resembles the early years of Bitcoin mining more than most participants recognize. One of the least appreciated inheritances from those years is that mining spent a decade learning to operate as a controllable load because it had no other option.
In the News
- Riot Platforms Details Q2 2026 Results, Providing Updates on 191 MW AI Lease and AMD Deployments
- Keel Falls 10% as Revenue Halves, Shuts Bitcoin Mines for AI Data Centers
- MARA Pledges 18,750 BTC for $600 Million New Bitcoin-Backed Loans
- Facing an estimated 474 GW of interconnection requests, Texas hits pause on data centers
- Bitcoin difficulty falls year over year for only second time
- Bitcoin miners slide as Bitdeer and Keel report quarterly results
Network at a Glance
- BTC price (USD): ~$63,495
- Network hashrate: 893 EH/s
- Difficulty: 127.4T
- Hashprice: ~$31.60 / PH / day
Project Spotlight
The Avalon Immersion Cooling Miner was never just a miner. It was a rack-scale system, built as a mechatronic assembly designed to run 90 mining modules submerged in a single coordinated deployment. The modules themselves were stripped down for liquid: two hash boards delivering the compute of three air-cooled boards, because once the coolant did the work, heat stopped being the limit on density.
One specification reads differently today than it did at launch. Canaan rated the immersion configuration for overclocking of up to 38%. In 2022 that was a cost-per-terahash argument. Run the same silicon harder, get more out of the same capital.
Set it against this week's feature, and it says something else. A machine with 38% of headroom above nominal has an unusually wide operating band, and a fleet with a wide band can place load precisely rather than in on and off steps. Response granularity is exactly what grid operators are beginning to ask large loads to document. Immersion delivers it as a byproduct of solving a thermal problem.
Two other properties came free. Submerged modules do not ingest dust, removing a primary wear mechanism. And liquid carries heat somewhere useful, which is the basis of the hash-to-heat work Canaan is doing in the Nordics today with hydro-cooled A1566HA units.
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/.
