Energy-AI Category Creator, Investor. In the business of electron arbitrage across space & time. Classical dancer, liberal thinker. CSIS, C3 Solution, IHS.
Earth
Joined January 2024
- Tweets3.5K
- Following1.2K
- Followers1.9K
- Likes2.4K
Pinned Tweet
@AuroraER_Oxford discussing a #PCLR #loadflex sensitivity analysis on live public webinar. happening now Awesome work and delighted to join them today to discuss how quickly and smartly we're writing the rules for bankable curtailment risk and financeable firming solutions in @ERCOT_ISO and what risks we can solve with NPRR1348. Anyone serious about shaping, and using, non-firm transmission service access should know our work, and start understanding its transferability to US RTO and non-RTO grids. Front of meter resources are incredibly effective at mitigating volumetric curtailment - the solution for #PCLR invites more steel on the grid for people. Learn more from the study, and thanks to @Google for funding this work. Location-specific risks drive speed to power, and the faster people understand this, the faster we solve AI loads on grids. Local load shed is the driver of the conditional interconnection, and it is the driver of speed to power for #datacenters
"PJM is part of the Eastern Interconnection. All the generators east of the Rockies, except most of Texas and Quebec, are connected. Several other RTOs depend on PJM exports to keep their systems out of trouble. If one RTO goes down, it will pull most of the Eastern Interconnection with it. On Wednesday evening it ran the other way, and MISO was sending PJM about 2.8 GW. So, taking care of your neighbor is also taking care of yourself."
- Kilovar on SS, Jul 19, 2026 - Equipment Outages: The Hidden Drag on Reliability
It's still all three of us but in a different way
@JigarShahDC @ericwg @_LatitudeMedia #flexsummit26
So this issue- that people hate compute facilities- is one that hyperscalers ecosystem reps say now in private and public rooms, is one that"money can't solve ". Lots of implications.
In Las Vegas last week, data center operators, investors and bankers discussed concerns that market conditions and local backlash to data centers could cause the funding market to tighten.
The topic of one panel: “Are We Screwed?"
wsj.com/finance/firmus-ipo-a…
w/ @cdriebusch and @RWhelanWSJ
Flexible Interconnection is about the Physics of the Point.
Grid reliability is about the Capacity of the Network.
How, is flex interconnection at a point, tied to the reliability of the whole to deliver to that point?
Here is the place I work every day, and I quite love it.
Chips for everyone
Nvidia - making chips
Google - making chips
Meta - making chips
Microsoft - making chips
Amazon - making chips
OpenAI - making chips
Anthropic - making chips
Alibaba - making chips
Everyone’s pushing to vertically integrate chip design to reduce reliance on Nvidia. Customers are working to become competitors. Most aren’t immediate threats yet, but they’re all moving quick
Only makes sense for Jensen to compete back
Arushi Sharma Frank retweeted
Interesting thread and break down of this deal
>Hut 8 is building the data center and signed long 15-year leases with Nvidia as the tenant so it could finance construction cheaply against Nvidia’s credit; Nvidia then hands the space to Lambda, which installs the GPUs and rents the compute to Anthropic.
>Nvidia does this to sell a large number of chips at full margin, take a cut of Lambda’s revenue, and keep long-term control of the powered site after Anthropic’s shorter commitment ends.
Why does Nvidia do it? A full-margin GPU sale on ~115K Rubins, an estimated ~10% rev share on Lambda's take, and years 7–15 of a 704 MW powered campus while interconnect queues run 4–7 years.
A 15-year lease vs 6-year offtake isn't a mismatch. It's optionality. (4/5)
newsletter.semianalysis.com/…
Arushi Sharma Frank retweeted
But the physics of grids are not that different. And every year spent rediscovering what dispatchability can do is another year of stranded infrastructure opportunity, unnecessary costs, and private capital waiting for both permission and a runway to solve a problem it already knows how to solve. This is the staggering opportunity cost of institutional inertia.
Arushi Sharma Frank retweeted
@grok is Erik Goff an @ERCOT_ISO Expert or Newbie? Does he like flexible Load and DC²?
Arushi Sharma Frank retweeted
A faster interconnection queue is not necessarily a better interconnection product that can be commercially developed, underwritten, and turned into reality. The longer we wait to recognize dispatchability attributes of flex loads, the worse off we’ll be: creating “paper compliance” with FERC Orders without commercially viable products.
Signing on in October 2026 to advocate for many of the same solutions I’ve already had a hand in writing into reality in ERCOT, is as validating to my apparent omniscience as it is triggering of a very particular kind of regulatory despair.
While I've had my head buried in supporting enhancements to market designs in Texas which credit load bringing dispatachability in planning, and attribution that is recognized by the grid in operations, a strong team of @pjminterconnect players wrote a letter that I agreed with and signed onto, delivered to Jason Connell last week. I'm seriously in the market for a clone so I can do both simultaneously and well.
But the physics of grids are not that different. And every year spent rediscovering what dispatchability can do is another year of stranded infrastructure opportunity, unnecessary costs, and private capital waiting for both permission and a runway to solve a problem it already knows how to solve. This is the staggering opportunity cost of institutional inertia.
It is profoundly dispiriting to watch one power market begin debating the same engineering truths while another has already spent years working to establish in actual grid code. @pjminterconnect is not @ERCOT_ISO so legal or market architecture cannot simply be copy-pasted from Texas.
Since I've already spent 15 months creating, and now will spend 15 months enhancing, commercially viable market design for flexible loads in a 100GW+ scenario (@ERCOT_ISO ), I feel confident saying that @pjm could nominally comply with @FERC 's orders while creating a process that remains commercially dysfunctional.
While I've had my head buried in supporting enhancements to market designs in Texas which credit load bringing dispatachability in planning, and attribution that is recognized by the grid in operations, a strong team of @pjminterconnect players wrote a letter that I agreed with and signed onto, delivered to Jason Connell last week. I'm seriously in the market for a clone so I can do both simultaneously and well.
My concern is that PJM could accelerate the administrative process for studying enormous new electricity customers without adequately recognizing what those customers, and the resources they finance, can actually do for the transmission system. In shorthand, I asked PJM to provide parameters valuing these loads’ dispatachability, because vibes are not good enough to finance a commercial solution.
A faster interconnection queue is not necessarily a better interconnection product that can be commercially developed, underwritten, and turned into reality. The longer we wait to recognize dispatchability attributes of flex loads, the worse off we’ll be: creating “paper compliance” with FERC Orders without commercially viable products.
Arushi Sharma Frank retweeted
I needed this entire interaction so badly. Thanks to all, and have a great weekend!
#ClimateCapATX 2023
#deploy
#tshirt
I bought two of these. One went to @ericwg and this one...my husband stole from me and wore in #ATX. The 👕 drank while he drank, and his torso became famous. @JigarShahDC was super embarrassed when I posted this from my old account, but he must be over it by now? 😂
This quoted post is unavailable.