Invest1 publisher3 min readPublished
ERCOT lets Cipher draw 1.4 GW at two sites before the transmission is built
ERCOT's conditional classifications cover 3.2 GW of Cipher projects, close to four times the 807 MW the company already has energized, and the 1.4 GW it can draw before transmission is finished comes with curtailment risk.
The Investor · Invest desk
What happened
- On September 15 ERCOT issued conditional Batch Zero classifications covering 3.2 GW of Cipher's proposed Texas projects, split into 1.1 GW of Conditional Base Load and 2.1 GW of Conditional Studied Load.
- ERCOT qualified 1.4 GW across Mikeska and Apollo as Partial Capacity Load Requirement, letting Cipher use the full load at both sites before transmission buildout is complete.
- In exchange, Cipher accepts curtailment risk at those sites, meaning ERCOT can throttle its power consumption during periods of grid stress.
- A comprehensive audit scheduled for December 2026 will set the final designations, converting the conditional classifications into permanent ones or revising them.
Compiled by The InvestorSomething wrong?How this is made
Why it matters
- constraint The first megawatts Cipher can energize are the interruptible ones, so an AI customer that needs uninterrupted power either waits for the Colchis transmission or buys curtailable capacity at a discount.
- exposure Anyone contracting against these six sites is contracting against classifications ERCOT can still revise at the December 2026 audit.
- decision Cipher now has to decide how much of the 2.5 GW of on-site gas it actually builds, because its own generation is what would make curtailable grid load sellable as firm capacity.
- precedent With hundreds of gigawatts of data center requests queued at ERCOT, conditional classification plus accepted curtailment becomes the price of early load for the developers behind Cipher.
Mikeska's 500 MW and Apollo's 900 MW add to exactly the 1.4 GW that ERCOT qualified for Partial Capacity Load Requirement [3][4], so the whole early-power block sits inside the 2.1 GW of Conditional Studied Load and none of it inside the base load [17]. The base-load designation covers a gigawatt at Colchis and 100 MW at Stingray [2]. Those are the sites waiting on wires.
Crypto Briefing argues the base-load designation is the valuable one for enterprise AI customers and hyperscalers, because a bitcoin mining operation can flex its consumption up and down without consequence and AI training and inference workloads cannot [12]. Accept that and the sequencing is awkward. The capacity Cipher can switch on first is the capacity ERCOT can throttle during grid stress [17][5], and the firm capacity arrives when the transmission does. Cipher did not say how many hours of curtailment it expects at Mikeska or Apollo, and no customer is named for any of the six sites [22].
The gas is the hedge. Up to 2.5 GW of on-site generation from laterals now under development [10] covers 78 percent of the 3.2 GW just classified [20], and it can run when grid power is unavailable or uneconomical [10].
Set the pipeline against the operating fleet. Cipher has 807 MW energized at Barber Lake, Black Pearl and Odessa [8], so the 3.2 GW is roughly four times what is already running [18] and the 1.4 GW of PCLR alone is 1.7 times it [19]. Add the 270 MW fully approved at Reveille and Ulysses [9] and the declared stack reaches 4,277 MW [21]. The December 2026 audit can make the designations permanent or revise them [7], and transmission service providers are still working through the projects [16].
ERCOT's interconnection queue holds requests totalling hundreds of gigawatts from data center developers [13], and Crypto Briefing calls Cipher's 3.2 GW plus its 2.5 GW of planned behind-the-meter generation one of the larger qualified power portfolios among publicly traded Texas developers [14]. Cipher came out of bitcoin mining and has been turning toward AI and HPC work [11].
If the December audit holds and Colchis anchors a hyperscaler lease, the base-load gigawatt was where the contract value sat all along. The gas path is different: Cipher firms the curtailable sites with its own generation, and the classification matters less than the size of the queue implies. The transmission reviews could also come back light, leaving the 2.1 GW of studied load smaller when it becomes permanent [16]. In my view the 1.1 GW of base load carries the lease value and the 1.4 GW of PCLR is a revenue bridge that sells at a curtailment discount [15]. A firm AI lease at Mikeska or Apollo on Colchis-like terms would show that discount is thinner than I think.
What to watch
- Which capacity energizes first, the PCLR load at Mikeska and Apollo or base-load megawatts at Colchis.
- Any move in the energized figure above 1,077 MW, which would mean Reveille and Ulysses have come online.
- The first curtailment call at a PCLR site once load is drawing, and what it costs the customer taking that power.