Recently, I noticed something interesting in the AI infrastructure market. Bitcoin miners are actually building assets that are worth far more than their own computing equipment. This is no longer about Bitcoin, but about the power they hold.

Here's the story. For years, Bitcoin miners have invested billions of dollars in building power infrastructure—substations, transmission connections, long-term supply contracts, and operational teams capable of keeping equipment running 24/7. This work takes an enormous amount of time and cost. But it turns out that’s exactly what the AI industry needs now, and they can’t build it quickly.

This is where the strategy comes in. Bitcoin miners realize they have a highly valuable economic bottleneck: installed electrical capacity at strategic locations, operational cooling infrastructure, and trained technical teams in high-density computing environments. Selling this capacity for AI workloads yields margins that are much more stable compared to Bitcoin mining, which is vulnerable to price volatility and halving events.

What’s most interesting is what the debt markets are telling us. Over the past twelve months, companies in this sector have raised around $33 billion in senior long-term bonds. But look at the coupons they’re paying—this reveals a lot.

CoreWeave issued at 9.25% and 9%. Applied Digital pays 9.2%. TeraWulf issued at 7.75%. Cipher Mining at 7.125% and 6.125%. All these companies are undergoing the same transition: from mining operators to AI computing infrastructure providers.

And here I see something crucial. Fixed-income investors aren’t financing the narrative—they’re financing cash flow. When lenders charge 300 to 500 basis points higher for AI infrastructure companies compared to regulated utilities, they’re actually saying something about the certainty of those cash flows.

Public utilities have guaranteed revenues through contracts, regulator-approved tariff structures, and assets with decades-long lifespans. But companies shifting from mining to AI? They have off-take agreements with customers, but lenders still don’t give them the same institutional status. The difference makes sense—contracts with AI customers are only as strong as their ability and ongoing demand for the models they run.

If AI demand adjusts or if customers are concentrated among a few big tech companies, cash flows from these operations become more unpredictable than those of electric distribution companies. Lenders price in this risk, and the coupons we see in the market reflect that very precise calculation.

The scale of what’s happening becomes clear when we look at the capacity being built. Bitcoin miners have about 30 gigawatts of new capacity being developed for AI workloads—almost three times what they operate today. Of course, not all of it will be built on schedule or within projected costs. Delays, grid constraints, and cost overruns could cause setbacks.

But the direction of capital is clear. Nvidia just reported 94% profit growth, 73% revenue growth, and quarterly sales of $68.1 billion. This confirms that the demand for compute power driving these investment decisions shows no signs of slowing.

The emerging business model combines two previously separate logics. On one side, the power infrastructure operator logic: maximize uptime, minimize cost per megawatt-hour, and negotiate supply contracts that protect margins from spot volatility. On the other side, the compute service provider logic: attract customers with intensive workloads, sign long-term contracts that enable debt issuance, and build a recurring revenue base that ultimately convinces lenders to lower coupons.

The key to success is whether these companies can narrow the gap before their current debt matures. If within two or three years they can refinance at 5% or 6% instead of 9% today, their business will change structurally. But if off-take contracts aren’t extended, if customers shift to proprietary infrastructure, or if electricity prices rise faster than compute revenues, then the heavy debt burden will compress returns and force restructuring.

For digital asset investors evaluating exposure to this sector, the question isn’t whether the Bitcoin miner transition to AI makes long-term theoretical sense—that’s clearly logical. The question is which part of the capital structure is right to hold. Debt at 9% offers yield with priority in liquidation, but limited upside. Equity captures appreciation if the model works, but absorbs losses first if contracts fail. The spread on these securities isn’t just credit market data—it’s the entry price for a question that still has no definitive answer.