Bitcoin Miners' AI Pivot Still Needs Power - But Wall Street's Wow Is Gone

Megawatt headlines still attract attention, but the market now wants revenue proof

A 500 MW headline can still grab Wall Street's attention before 2027, but only if it starts converting into contracted revenue. The market is no longer rewarding raw megawatt claims by themselves. It now wants evidence that power can turn into committed customers, visible deployment, and durable cash flow.

The sector's bar has moved higher

Across the miner set, there is roughly 2,000 MW in AI conversion projects reportedly in motion, with Core Scientific and Tera Wulf leading the shift. Against that backdrop, AMD took an option on more than 500 megawatts of Core Scientific U.S. capacity beginning in 2027, with scope to reach 2.5 gigawatts. That is large enough to matter, but the rerating case still depends on conversion into committed hosting revenue, not on infrastructure potential alone.

Bulls see scarcity; bears want the term sheet

The bull case is straightforward: power, land, and infrastructure can constrain AI buildouts as much as chip supply, so controlled megawatts matter. The bear case is just as clear: the AMD deal did not disclose the conditions, timetable, or financial terms for expansion, and Core Scientific did not provide an AMD-specific revenue estimate. For now, investors still need disclosed timing, pricing, and named demand before megawatt claims become a full valuation case.

Why miners still have leverage in the AI infrastructure race

AI buyers want speed and scale, not branding

AI buyers are trying to rapidly deploy high-density data centers at scale because their workloads need faster model training and low-latency performance. That makes the buying decision largely about time-to-rack, power density, and whether a site can scale without running into utility or construction bottlenecks.

That is where miners can matter. Core Scientific offers 1,300+ MW of contracted power and markets pre-secured power as a way to accelerate deployment. In practical terms, AI buyers want to shorten the slowest parts of the buildout process, and controlled power plus prepared sites sits directly in that choke point.

The advantage is infrastructure, not AI technology

The miner edge is not deeper AI expertise. It is that they have already chased the asset AI developers cannot instantly create: dense, scalable power. The sector has moved toward multi-year, gigawatt-scale hosting agreements as mining economics softened and AI power demand rose. That is the real bottleneck investors should track.

The core economic question is still whether leasing that capacity to AI customers can produce better margins and more scale than mining. The upside looks real, but large commitments should still be treated as potential rather than proven economics until pricing and timing are binding.

That is why the AMD headline matters, but only as a scale signal. AMD secured more than 500 megawatts with scope to reach up to 2.5 gigawatts. The rerating trigger is not the option itself. It is whether that capacity becomes committed hosting revenue with clear timing and terms.

What the market should watch now

• Named demand: more customers like AMD, not just broad AI interest.
• Binding terms: pricing, start dates, and expansion conditions moving from headlines to contracts.
• Revenue mix: higher-value hosting showing up in reported results, not only in projections.
• Deployment proof: installed gear and shipped capacity matter more than fresh MW claims.

Contracts, mix shift, and deployment are what can restore the rerating

The option story is no longer enough. What can restore a rerating is signed hosting demand, a visible shift in revenue mix, and evidence that capacity is being deployed and earning.

The scoreboard

The market will start paying up again when AI stops being an infrastructure claim and starts showing up as contracted, higher-value revenue. The sector already has multi-year, gigawatt-scale hosting agreements, and projections suggest AI-related services could reach as high as 70% of revenue by the end of 2026. Even so, investors need to see that shift in filings rather than in roadshow materials alone.

Into 2027, the key catalyst is simple: move from power access to committed demand. Core Scientific's AMD deal provides a template, with capacity beginning in 2027 and room to scale toward up to 2.5 gigawatts. But because the companies did not disclose the conditions, timetable, or financial terms, the rerating trigger still lies ahead.

The practical split in the sector

Bears still have a real argument. The AMD announcement did not disclose financial terms, and Core Scientific did not provide an AMD-specific revenue estimate, leaving room for delayed monetization, weaker-than-hoped pricing, or expansion that never materializes at scale.

So the clearest split is this: the fastest monetizers will be operators that already have pre-secured power and can turn it into faster deployment alongside committed customers. The weaker setup belongs to companies marketing capacity before they have customers, pricing, or revenue visibility. That distinction is more likely to separate the reratings from the write-downs before 2027.