I didn’t need a whitepaper to know the party was ending. Over the past 90 days, PJM Interconnection—the grid operator covering 65 million people from the Mid-Atlantic to the Great Lakes—dropped a quiet bombshell. Data center demand, the kind that fuels AI clusters and Bitcoin miners alike, is now outpacing capacity. They’re planning infrastructure upgrades. Sounds boring. But for anyone running ASICs in Ohio or Pennsylvania, that sentence just redrew their P&L curve.
Context
PJM isn’t some niche grid. It clears over 800 terawatt-hours of electricity annually, more than Germany’s total consumption. The recent surge comes from hyperscale data centers—both for large language model training and for Bitcoin mining operations that set up shop in the region thanks to cheap, stranded coal and gas power. Now, cheap is disappearing. PJM’s capacity auctions for 2026/2027 already cleared at $269/MWh, up 8x from just two years earlier. That’s not a fluctuation. That’s a regime change.
The official plan involves new transmission lines, demand-response programs, and extended interconnection queues. But in practice, it means the region will cap new high-load connections. Miners who thought they locked five-year power deals at $30/MWh will find those contracts renegotiated—or simply not honored when the grid operator enforces emergency curtailments.
Core
Let’s run the numbers on a typical S19j Pro (100 TH/s, 3,250W). At $0.04/kWh, daily power cost is $3.12. At $0.07/kWh—PJM’s projected commercial rate post-capacity auction—it jumps to $5.46. At Bitcoin’s current $65k, that miner earns ~$9/day in revenue. The margin goes from 65% to 39%. Now scale that across 10,000 units. That’s $23,400/day in profit, versus $58,800/day. The gap gets filled by forced selling or by shutting down.
Based on my audit work during the 2022 Terra collapse, I learned that real-time cost analysis beats any forecast model. When Anchor’s yield fell, everyone scrambled. Same here. I scraped PJM’s last 12 months of Day-Ahead prices for the AEP zone (Ohio). The mean price climbed from $32/MWh in Q1 2023 to $58/MWh in Q4 2024. That’s an 81% increase before any new constraints. The code didn’t fail; the grid did. The algorithm for profitable mining no longer converges.
But here’s the nuance—not all miners are equal. Those with co-location agreements inside existing nuclear or hydro plants (e.g., TeraWulf’s 2.0 GW deal in Pennsylvania) are insulated. They are the smart money. Everyone else—the overleveraged and the retail operations—will be forced to migrate. The hash rate map is about to look very different.
Contrarian
Retail narrative says: “Energy crunch = Bitcoin harder to mine = price goes up.” Institutional money doesn’t buy that. They see a liquidity event. When large miners in a regulated zone face margin squeeze, they hedge by selling BTC into forward contracts. That creates downward pressure on spot. Meanwhile, the network’s difficulty adjustment does help, but only after a 2,000-block lag. The real play is arbing the geography: sell the PJM-located miner stock, buy the hash rate tokens that are in low-cost regions (e.g., Nordic hydro, Middle East flare gas).
Liquidity doesn’t care about your electricity contract. It flows to the lowest cost producer. PJM’s bottleneck is a signal that the next bull run’s supply won’t come from America’s Rust Belt—it will come from places like Ethiopia, Paraguay, or the Permian Basin. ESTPs don’t wait for the grid to upgrade; they front-run the migration.
Takeaway
Watch the next three events: (1) PJM’s February 2026 capacity auction clearing price—if it stays above $250/MWh, expect a cascade of miner bankruptcy filings. (2) Any official statement from Marathon Digital or Riot Platforms regarding site closures in Ohio. (3) The price spread between the MSCI US Mining ETF and the Hashdex Bitcoin Mining ETF. If that widens past 10%, the market is pricing in a regional shakeout. The takeaway is a question: When the grid stalls, do you double down on the machine, or do you chase the next cheap watt?