I don't think what's being proposed here by PJM is as black and white as you're indicating. I would not suggest that these rules are a condition of interconnection. They're 1 option. It's far more likely that data centers are going to go the BYOG route to assure firm interconnections
“If your solution to some problem relies on ‘If everyone would just...’ then you do not have a solution. Everyone is not going to just. At no time in the history of the universe has everyone just, and they’re not going to start now.“
The final rule text isn't yet available, but there's still *a lot* we already know about the Trump EPA's coming repeal of the endangerment finding and greenhouse standards for vehicles.
Here's an explainer (with a gift link) to get you up to speed @bloomberg.com: www.bloomberg.com/news/article...
Live look at the Cavs
At #prf26 with AHG and Latitude Media - just learned from Sunny Wescott, Chief US Meteorologist 2 fun new nightmares
1. Coal seams can autocombust at 105 F
2. Climate induced drought-flood cycles are driving more earthquakes
Insane take by Heatmap readers
Some of these resources, may be colocated with DCs, but we shouldn't expect to see flexibility coming from the core compute operations (despite it being technically possible)
It's clear that as DC loads drive tightness in power markets and grid capacity, there will be a premium on flexibility. That being said, we should expect that need to be served by purpose-built DER technologies (storage, DR, VPPs, etc).
5. In 2025 a narrative took hold that there is a big distinction between training and inference loads. This seems to have been an overly simplistic mental model for non-AI folks (like myself). There just don't appear to be these nice clean huge training loads that can be easily curtailed.
4. Any capacity-style payment for flexibility falls far short of the incentive required for a DC to opt-in. This was highlighted by the overwhelming resistance to PJM’s proposal to remove data centers from the capacity market in exchange for curtailment rights. Value of Compute >> Cost of Power
3. Many have posited that "speed to power" is the carrot that will drive flexibility, but this theory is finding its limits. DCs are currently rejecting offers for non-firm power more than a year sooner than firm power. This is largely because of the inability to accept the downtime uncertainty.
2. Flexibility is only acceptable to DCs if it comes with tightly contained uncertainty - but uncertainty is exactly what grid operators want to unload! DCs only want to provide curtailment within tightly defined limits, restricting their value to the grid and complicating dispatch decision-making
1. Data center operators would much rather pay a third party for flexibility than mess with their own operations. We are seeing a shift toward outsourcing flexibility to DER ecosystems. This enables enables maintaining high uptime on the most costly components - the chips.
At the end of 2025, @stephenlacey.bsky.social named the "flexible data center" as the most overhyped story of 2025, and I tend to agree.
My two cents is that data centers will be exactly as flexible as regulators and policymakers make them be, and no more.
A 🧵 on why:
#energysky 🔌💡
Electricity rates are a blunt instrument for funding social goods like wildfire mitigation and are ultimately regressive. In 2026, expect to see state level policy fights about using more targeted taxpayer funded approaches to drive climate resilience
lnkd.in/epTnzrNc
#energysky 🔌💡
Hate to see ARPA-E catching some strays here
It’s gone full “Eye of Sauron” here in DC
More like see-rate, amirite?
Chaos in the streets here in DC
Looks like the new ChatGPT was trained by this guy
"[It's] as if we’ve collectively agreed that fiscal responsibility is best taught through the ritual humiliation of second graders."
This is an excellent piece, not just about school lunch debt, but how we manage the relationship between working for systemic change and mitigating immediate harm.
This is why it’s more of an indictment of the user than the machine when someone criticizes AI for being factually wrong. If you’re using it for facts, that’s actually on you
These are possible names considered for a smaller organization to replace FEMA, from a March memo:
- Office of Crisis Management (OCM)
- Office of Crisis Response (OCR)
- National Crisis Response Agency (NCRA)
- National Office of Emergency Management (NOEM) 👀
www.bloomberg.com/news/article...
Good faith actors in the space who consistently note that CDR is not a replacement for mitigation would love for this to *not* be the case. The O&G incentive to back the tech is an albatross around the neck of the CDR industry
I think even (especially?) DAC’s fiercest critics would say that DAC has pretty obvious utility for the oil and gas industry whether or not it “works”
It's absolutely disappointing that the facilities are underperforming, and reasonable folks can argue that DAC is unlikely to succeed, but there is nothing even scam-adjacent here
The fact that their still scaling facilities have not yet covered all of their company-wide emissions is not at all scandalous
i was also disappointed to see the amount of piling on and uncritical sharing of the original story by people who absolutely should have known better. The story just reported the very public numbers that Climeworks shares about its emissions!
If you read a hit piece about a climate tech that features a particular litigious Stanford professor, you should question the seriousness with which the story was reported
This great piece by @jackandreasen.bsky.social provides a more valuable and reality based take on what is going on in DAC world
