Amit Agrawal

U.S. Data Center Interconnection Expert. 8+ years of experience in US Power Markets Interconnection and Transmission Planning. Opinions stated here are my own.

[Amit Agrawal @energy4ai]

This is why I like regulated utilties in WECC. They are fast in implementing solutions like this. Microgrid + Grid power in the simplest way is real speed to power. For context, A PJM or MISO based utility have to go via regional planner and a generation queue between the contract and firm delivery. Black Hills, in WECC, can still do bilateral large-load service, customer-paid upgrades, and on-site generation under state regulation and its own OATT. That is the structural reason the WECC regulated-utility path is usually faster for first megawatts as compared to those siting in RTO/ISOs.

[Amit Agrawal @energy4ai]

@duncancampbell 100% specially in interconnection and power markets world. Anyone interested in building Harvey for Interconnection with me ? Lets chat

[Amit Agrawal @energy4ai]

Flex data center is the right solution to increase the grid utilization, however the biggest problem is that the curtailments are not bounded. Flex data center will not know how much curtailment they can expect and at exactly what time. The flex datacenters (non-firm power) are the first one to get curtailed before any emergency actions are called for. That means the trigger of those curtailment could be due to generation shortage, transmission shortage or both. Hence, the biggest challenge right now with flex data center and thier investors is that unbounded risk of curtailment. Certainity of power is as important as speed to power but these are good signs and industry will evolve sooner.

[Amit Agrawal @energy4ai]

@ShanuMathew93 Can you send me the article link ?

[Amit Agrawal @energy4ai]

@ShanuMathew93 You based in london ? Lets catch up.

[Amit Agrawal @energy4ai]

@MTSlive @ts_fisher @ts_fisher how is reliability designed around this data center ? N+1 ? What about the cost as compared to grid power ? Are they still chasing grid power and the current setup is brige solution ? Will read more about this.

[Amit Agrawal @energy4ai]

@DaveShapi @grok explain how regulations kills competition. Give one example that happened before and benifited the incumbent companies ? Geniunly curious to know

[Amit Agrawal @energy4ai]

True and I think this where WLPUN falls short and complex to implement. There was lot of push by TO’s in MISO to assume ZGIA generators as “offline” while performing reliability studies of associated load but MISO pushed back by stating that ZGIA generator off should be treated as grid event (P1/P3 contigency) that assuming it as offline in base case (P0). This makes huge difference, not only it avoids lot of network upgrades but it truly gives speed to power as not many upgrades get triggered with ZGIA on base case. Also, since MISO consider ZGIA associated load as firm NITs, they are bounded to pay transmission and capacity charges for gross load (not netting allowed there) but ZGIA generators can be accrediated to offset those capacity charges. I think MISO did great job by making sure the colocated generator is accredited.

[Amit Agrawal @energy4ai]

@ArushiSF @nerdalert Weekend read sorted 🙌🏻. Looking forward to listen to this.

[Amit Agrawal @energy4ai]

@stretchcloud Loves reading this. Seems your in london as well, lets catch up ??

[Amit Agrawal @energy4ai]

I have seen that given the current shortage and ROI for short term compute, the Interconnection Agreement of firm power from utility is the single biggest document that solves this gap. In current state if any developer have site selected smartly where they get utility contract signed as Firm power IA with minimum Transmisison upgrades required (very rare though) both lenders and tenants are willing to take bet. The chicken-egg situations turns into nice dinner for everyone if developer have signed IA of firm power. Also, which ISO you are in does matter for example, in MISO you can even self build and expediated transmission upgrades, in PJM state can go through SAA route to self build transmisison/substations to get power early and to expediate the infra build which is what Softbank and OpenAI are olanning to do in Ohio.

[Amit Agrawal @energy4ai]

Few more dimensions to add even after getting behind the meter/backup onsite generation sorted is the community backlash. Hence, selecting states like WV that already have mature microgrid district framework is going to be game changer. + Every utility and ISOs are coming up (mostly by end if 2026) with non firm interconnection and then the questions flips from where i get the power fastest to where I get power fastest using non-firm interconnection (flexible interconnection) with minimum curtailment risk. I think next year data center that have best curtailment risk hedge strategies will get financed more than speed to power equtions. Interesting time to be a power market consultant

[Amit Agrawal @energy4ai]

Unfortunately power headroom metrics is way more complex than presented here but great job on putting together other aspect. It would be interesting to add tarrif tracker. Is it opensource code, I think I can contribute to add few layers like transmission upgrades, interconnection requirements etc. Great Stuff.

[Amit Agrawal @energy4ai]

Getting Conditional Base Load classification in ERCOT Batch Zero could make these data centers significantly more bankable. We saw a similar outcome with Hut 8.

I’ve written a short blog post on my timeline explaining what Base Load classification means, why it matters, and how it can impact the bankability of large data center projects.

[Amit Agrawal @energy4ai]

Getting Conditional Base Load classification in ERCOT Batch Zero could make these data centers significantly more bankable. We saw a similar outcome with Hut 8.

I’ve written a short blog post on my timeline explaining what Base Load classification means, why it matters, and how it can impact the bankability of large data center projects.

[Amit Agrawal @energy4ai]

This is great news for @Hut8Corp being classified as Base Load. It’s also a useful way to understand how large-load interconnection in Texas is changing.

If you’re a data center in ERCOT >75 MW, your load is studied through the large-load cluster process. Until recently, the process was largely serial: a data center submitted a request to its utility, the utility studied the load for grid impact, and requests moved through the process one at a time. That created significant delays.

ERCOT transitioned to a batch-based process, studying groups of large-load requests together. The first transition batch was "Batch Zero" And Batch Zero is NOT simply “first come, first served.” It starts with a critical sorting exercise:

Base Load → Studied Load → Removed

That classification is the whole game. Everything after that is essentially the math applied to the loads that did not make Base Load. So what is Base Load?

Base Load represents data center demand that is sufficiently advanced that ERCOT treats it as already real and firm for purposes of the study. Prior studies, existing agreements, and other evidence of project maturity can support that classification.

That is why ERCOT’s confirmation that @Hut8Corp’s load qualifies as Base Load is so important.

ERCOT is effectively preserving the load already supported in the planning process rather than subjecting it to the same haircut applied to less-mature requests. That’s why Base Load is being described as a potential “golden ticket” to speed-to-power.

And this matters enormously for data center financing.

A 15-year triple-net AI/data center lease is ultimately underwritten on: Delivered MW + delivery date + PUE.

As data center Interconnection and power strategy consultant, working with lenders and data center tenants on ERCOT interconnection, one of the biggest and frequent questions we try to access: Is the power firm — or is it just a hope?

Base Load provides a much stronger answer: The load is already in the plan. That means the remaining risks are much more about construction, QSA/agreements, and final verification rather than: “Your 1 GW request just became 180 MW of firm power. because there is no transmission available”.

@Hut8Corp data center have crossed that bridge and will not have any curtailment risk due to transmission bottleneck or available transmission capacity. For AI data centers, that distinction can be worth billions.

#datacenter #speedtopower #powerstrategy #interconnectio #powermarkets

[Hut 8 @Hut8Corp]

Hut 8 has received a conditional Base Load classification in ERCOT’s Batch Zero Process for Beacon Point, a step forward in the energization process. We remain focused on working closely with our transmission and distribution service provider, ERCOT, regulators, and local stakeholders as the project advances.

[Amit Agrawal @energy4ai]

No asset class in human history has produced this much revenue per kilowatt, and we are still early in adoption. The compute is the constraint, the power behind the compute is the real constraint, and that is what will decide who wins.

[Kevin Stevens @kevindstevens]

Interesting framing on how resource intensive AI can be.

Power: 150GW by 2028, adding a Spain to the global grid

Water: 1 campus = 5M gal/day or a town of 50K

Copper: 700 Mt in next 25 years, same as all of history

Land: 6-10M sq ft per GW scale campus, bigger than the Pentagon

It’s crazy to think about it in these terms, and the scale of the opportunity is enormous.

[Amit Agrawal @energy4ai]

The Data Center Developers Who Win Will Be the Ones Who Accept Power Curtailment

Speed to power dominates every conversation about connecting data centers to the US grid. Until recently a data center had one way onto the grid, and it was called "Firm Interconnection". That interconnection process was sequential and physical. A developer applied to the local utility. The utility ran a system impact study, then a facilities study. Between them they identified the network upgrades the data center project triggered: new substations, transformer banks, transmission lines, reactive support required to connect data center to grid.

The studies could close in weeks but the transmission infrastructure takes three to seven years. As queue volumes climbed, that gap rather than the analysis is what set the schedule, and "firm transmission service " was the only widely available option for data center to connect. But Firm service required the transmission upgrades to be in place first and hence it takes years to get power.

While "Firm Transmission Service" is still default and bankable option but it is no longer the only option on the table.

Four major grid operators across U.S. now offer structured non-firm and flexible interconnection paths for data centers. The core design across the ISOs/RTO is consistent: the data center does not have to wait for the network upgrades before taking first power. In exchange, it accepts curtailment risk i.e. cut down power usage when grid is stressed (It happens only few hours a year) and connect to grid faster.

The below table presents comparison between already approved or ongoing discussion regarding non-firm/felxible interconnection across US grid Every option below is faster and cheaper than firm service, because the customer gives up the guarantee of always-on power and are flexible.

I will be writing details covering each of those options across US wide RTO/ISOs in following posts/article.

#energy #datacenter #energymarkets #AIinfra #speedtopower #interconnection