Bitcoin mining economics are weak, but the stocks are still rallying
The paradox is blunt: the CoinShares Bitcoin Mining ETF is up by more than 90% this year even as mining economics have weakened materially. That is why many investors are treating this as a rerating story rather than a simple short-term bounce.
By late last year, the average cash cost to produce one Bitcoin for public miners had climbed to nearly $80,000. With Bitcoin still trading around $81,000, margins were thin, and the picture worsened when price fell toward $65,000. If these stocks were trading purely as BTC beta, that pressure should be obvious in the share prices. It is not.
Bulls see a business-model reclassification into digital infrastructure; bears see investors getting ahead of a broken post-halving cycle. The market appears to be making the same call AI buyers are: scarce infrastructure can be worth more than the legacy activity it was built for.
The near-term watchpoint is straightforward: watch whether colocation revenue keeps scaling and whether companies keep turning secured power into billable AI capacity. If that starts showing up in reported numbers, the rerating has real support. If not, the market will remember that mining economics are still weak.
The AI infrastructure bid is about power, density, and speed
The rerating mechanism is simple: AI demand is concentrating into a narrow set of constraints, and miners already control some of the scarce inputs. Global semiconductor sales are projected to reach USD 975 billion in 2026, while industry growth is projected to accelerate to 26% in 2026. More important, AI chips now generate roughly half of total industry revenue even though they are less than 0.2% of units shipped. That helps explain why investors are paying for the infrastructure that can house, power, and cool high-value compute fast, not for chip units themselves.
Why power matters more than coin price
In this new framework, the valuable asset is not hash rate. It is secured power, high-density rack capacity, cooling, and speed to deployment. That is why Core Scientific now highlights 1,300+ MW of contracted power as a core selling point for AI workloads. When customers need compute now, they do not buy a promise of future power. They buy a site that can accept equipment, handle dense loads, and start billing quickly.
That is also why pipeline quality matters more than the headline capacity number. Core Scientific expanded its gross power capacity pipeline to 4.5 GW, but the more investable detail is what comes with it: new land and power in Texas expected to support roughly 430 MW of gross power capacity, with 243 MW of capacity already tied to approximately $350 million in average annualized colocation GAAP revenue. Bulls see power backed by revenue visibility. Bears will note that pipeline power is not cash flow yet.
The pivot is already showing up in results
This is no longer just a future-tense story. Core Scientific's colocation revenue was $77.5 million in the first quarter of 2026, up from $8.6 million in the first quarter of 2025. That shift matters because the new bid is coming from revenue tied to AI rack deployments, not bitcoin difficulty or short-term coin price.
What investors need to underwrite is not "crypto company tries AI." It is whether a company can turn power entitlements into billable AI infrastructure before peers catch up. The execution loop is straightforward:
- secure long-lead power
- build for high-density workloads
- compress time from acquisition to revenue
- reuse those builds to add more contracted capacity
The upside case is simple: if AI demand keeps concentrating into scarce power and fast deployment, miners with pre-secured infrastructure may get valued more like infrastructure operators and less like mining companies. The downside case is just as clear: if power expansions slip or colocation revenue fails to scale from the pipeline, the market may stop treating these assets as AI infrastructure and start treating them again as distressed miners.
What would validate the rally
The rally only holds if execution starts to confirm the story, not just capacity announcements.
The main validation points
The near-term bull case is about speed to cash, not story depth. Management has compressed Ready-for-Service timelines to 12-14 months and expects more than 450 billable megawatts by the end of summer 2026. If those dates hold, investors get evidence that pre-seeded infrastructure can turn into billable AI capacity on schedule.
That matters because the market is no longer paying for ambition alone. It is paying for delivered, billable power that can support higher-value compute now. A second validation point is the full 590 megawatt CoreWeave commitment by early 2027. If that build stays on track, it would support the case that the capital engine is real and can fund the next layer of customer traction.
The two risks that can break it
First, delivery risk. If Ready-for-Service dates slip, the rerating can wobble because the market would stop seeing a fast-moving infrastructure builder and start seeing another pipeline full of promises. That would revive the old bear case: too much capex, not enough billed capacity yet.
Second, demand integrity. The AI build-out is still vulnerable to a sharper demand correction, especially because HBM and advanced memory constraints are constraining consumer DRAM/NAND capacity. If AI capex starts pulling from other parts of the stack, miners may still have power and buildings, but the customer queue could thin faster than expected.
If those signposts hold, the rally can keep going. If they do not, the market is likely to push this trade back toward a distressed-miner valuation.