PowerCompute raises Columbus contracted demand to 11 MW under data-center tariff
Source report: 2026-10-06 · Editorial analysis published: 2026-10-07
The Mississippi contract adds 2.5 MW. PowerCompute says no capacity commitment charge was assessed and a $300,000 deposit is being replaced by a utility bond.
Analysis and practical implications
This section is our analysis and illustrative calculations, separate from the source report.
A larger power contract with an unchanged operating practice
PowerCompute announced on October 6 that its Mississippi subsidiary signed a new contract with Columbus Light and Water on September 30. Contracted demand at the Columbus site increased from 8.5 MW to 11 MW, with the new agreement taking effect on October 1. The company says the site continued receiving power without interruption. The agreement moved the site to the local Large Data Service schedule implementing TVA’s data-center rate, rather than leaving it in the previous general industrial classification.
The increase is 2.5 MW, or approximately 29.4% of the former 8.5 MW contract. That calculation measures additional contracted demand, not an increase in Bitcoin output or energized mining hardware. Our interpretation is that the contract improves the site’s available development envelope, while fleet installation and commissioning still determine how it is used. A reader should track the contract quantity, actual demand and operating hashrate independently. The statement that service continued is evidence of continuity, not a report that every watt of the enlarged allocation was already consumed by miners.
A capacity charge is different from a per-kWh electricity rate
PowerCompute says none of its 11 MW was assessed TVA’s capacity commitment charge. The release describes a charge of approximately $1.5 million per MW for qualifying new or expanded data-center demand above the first 5 MW, and attributes its treatment to the timing and classification of the contracted load. These are the company’s statements about its agreement. They should not be generalized into a legal conclusion that every existing mining site or every September-signed expansion qualifies for the same treatment.
Using the release’s simplified example, a completely new 11 MW load would have 6 MW above the stated threshold; multiplying 6 by $1.5 million gives approximately $9 million. This is an illustrative comparison with the described charge, not PowerCompute’s electricity bill, annual profit or independently verified savings. A demand-related capital charge and an energy price per kWh belong in different parts of a site budget. The tariff’s complete applicability and payment terms must be checked with the utility before using that comparison in a real project estimate.
Returning a deposit releases liquidity without creating revenue
The utility is returning a $300,000 power deposit, which PowerCompute is replacing with a utility bond. The company intends to use the returned cash for capacity buildout at Columbus. This changes where working capital is held; it does not represent Bitcoin production or customer revenue. The release also says the additional contracted demand required no capital contribution to the utility. The bond’s premium, collateral requirements and ongoing conditions are not quantified in the announcement, so they should not be assumed to be zero.
Our reading is that the liquidity benefit and the cost of the replacement instrument should be evaluated separately. A deposit return can make cash available for installation work while a bond introduces its own obligations. A project cash-flow model would record the returned deposit, the buildout payments and any bond costs on their actual dates. Treating the full returned amount as recurring income would overstate operating results. This distinction matters especially for miners that compare site upgrades using a daily profit calculator: a one-off working-capital movement is not a substitute for a sustainable change in revenue or energy expense.
Curtailment changes utilization as well as electrical cost
According to PowerCompute, the Columbus schedule continues to distinguish on-peak and off-peak power, and the site curtails during on-peak periods. The company says the new schedule did not change its on-peak contract demand or this operating practice. It can still take on-peak power at the applicable demand and excess-demand rates. The announcement does not publish a complete hourly tariff or the number of hours that will be curtailed, so it cannot support a precise future energy bill or an assumed 100% mining uptime.
A continuous 11 MW load would use 264 MWh in a 24-hour day, calculated as power multiplied by time. That is an upper-envelope illustration for continuous operation, not a measured Columbus consumption result. Curtailment reduces both electrical consumption and the opportunity to earn mining rewards. Our analysis is that an operating comparison must apply the same uptime assumption to the energy and revenue sides. Demand charges, however, can depend on how and when peaks occur, so they need separate treatment rather than being hidden inside a single averaged energy price.
Installed miners do not fully describe the enlarged allocation
The release reports approximately 2,373 installed Antminer machines and about 205 PH/s at Columbus as of September 30. Dividing the aggregate hashrate by that approximate count gives about 86.4 TH/s per installed machine. This arithmetic is only a fleet average from the disclosed totals. It does not establish that every device is the same model or runs the same profile, and it does not identify how many machines are active at a particular hour. We do not assign an individual miner configuration from that average.
Similarly, dividing the full 11 MW contract by the miner count would not reveal measured per-device consumption. The contract can cover unused capacity, electrical overhead and other site loads, while the fleet count describes installed hardware. A more informative operating update would provide active hashrate, metered demand, auxiliary consumption and the configuration mix over a defined interval. Our interpretation is that these measurements are the bridge between the contract expansion and mining economics. They would show whether additional electrical access translated into more productive capacity without conflating a commercial allocation with wall-power telemetry.
The two-site portfolio has more than one computing use
PowerCompute describes 15 MW at Calumet, Oklahoma and 11 MW at Columbus, for a 26 MW portfolio. The Oklahoma site is outside TVA’s service territory and is not affected by the Mississippi contract change. The company also describes approximately 22.5 MW primarily supporting Bitcoin mining, with a portion assigned to an enterprise HPC pilot. At Calumet it rents GPU compute through the Vast.ai marketplace. These disclosures establish different uses and jurisdictions; they should not be combined into a claim that the Columbus tariff applies across the portfolio.
For a buyer comparing mining and GPU hosting locations, the relevant operating boundary is the particular site and service being offered. Electrical access can support several business models, but their customer obligations and utilization patterns differ. Our analysis does not treat the mining allocation as a fully leased AI campus or the existence of a pilot as proof of a profitable large-scale HPC conversion. The new Mississippi classification is commercially relevant, yet each workload still needs appropriate infrastructure and a customer or revenue model. Portfolio megawatts alone do not establish the performance of those services.
A tariff classification does not complete an HPC conversion
The company presents the Columbus classification as compatible with the type of service an HPC facility on TVA power would require. That statement concerns the electrical tariff. It does not itself provide GPU servers, the network connectivity, cooling configuration or customer contracts needed for a sustained HPC service. Our interpretation is that tariff readiness can remove one development obstacle while leaving the rest of the implementation path open. The release does not announce a completed new AI deployment at Columbus or quantify a contracted HPC revenue stream there.
A practical conversion review would first define the workload and the service-level obligation. Interrupting a self-mining ASIC fleet can have a different economic effect from interrupting a reserved computing job for a customer. Cooling design, redundancy and data connectivity should be sized for the actual installation rather than copied from the mining fleet. These are our review criteria for using the enlarged allocation, not additional company commitments. They help explain why valuable grid access and a suitable tariff can improve a project’s options without proving that every alternative computing use is already operational.
The next useful evidence is metered utilization and buildout
The announced contract runs through September 16, 2030. Follow-up disclosures can show how the returned working capital is applied, what equipment is installed and how much of the larger allocation is actually used. Those milestones would connect the commercial agreement to physical output. We retain the company’s statement of no assessed capacity charge while leaving detailed tariff interpretation with the utility and the contract. The release does not establish a universal hosting price, an individual miner payback period or a forecast of the site’s future Bitcoin production.
For ASIC operators, the practical lesson is to compare complete electrical terms: available demand, energy prices, peak treatment, curtailment obligations and any capital or security requirements. A favorable result on one component can still leave another component decisive for profitability. The Columbus update adds a specific example of contractual capacity expanding without an assessed capacity charge, with the company’s existing operating pattern retained. Actual utilization, buildout spending and metered mining performance will determine how much operating value follows from that access. Those are the measurements to watch after the contract headline.
Source: PowerCompute ↗
Mining calculator ↗
