So, here’s the thing, I just realized something pretty interesting happening in the AI infrastructure sector. Bitcoin miners have actually built assets that are now worth more than their own computing equipment—energy infrastructure. Substations, power transmission, long-term power supply contracts, 24/7 technical teams. All of this costs billions of dollars and years of negotiation.

Now, the funny part is that the AI industry now needs exactly the same assets. They can’t build them fast enough. So bitcoin miners are starting to realize—they have a very valuable bottleneck in the next digital economy. Installed energy capacity, operational cooling systems, trained teams in high-density computing environments. Many are still mining, but many are also pivoting or running parallel operations.

The most telling part is in the debt markets. Over the past 12 months, companies in this sector have raised around $33 billion in senior debt instruments. And just look at the coupons they’re paying—an extremely clear story about how creditors value this model. CoreWeave 9.25%, Applied Digital 9.2%, TeraWulf 7.75%, Cipher Mining 7.125% down to 6.125%. All of these are companies transitioning from mining operations to AI infrastructure providers.

But this is interesting. Fixed income investors aren’t financing narratives—they’re financing cash flow. When creditors charge 300-500 basis points higher interest to AI infrastructure companies compared to traditional utilities, it’s a statement about cash flow predictability. Utilities have revenue backed by contracts reviewed by regulators, approved tariff structures, assets with decades-long lifespans. Meanwhile, ex-mining companies have offtake agreements from AI clients, but creditors still treat them differently.

And that’s not irrational. Contracts from AI clients depend on the solvency of the clients and ongoing demand for the models they run. If there’s a demand correction in the AI market or customer concentration risk with some tech giants, cash flow becomes less predictable. So the spreads you see are literally pricing in the risk differences.

For digital asset investors, these spreads have an additional meaning. It’s the cost of transition. Bitcoin miners need to accumulate enough cash flow history from long-term contracts before the market treats them the same as consolidated utilities. Until then, the credit market continues to treat them as growth bets. This puts pressure on operational margins because part of the cash flow goes directly to expensive debt service.

The scale of the risk is clear when looking at planned electrical capacity. Bitcoin miners are planning 30 gigawatts of new capacity for AI workloads—almost 3x what they’re currently operating. Of course, not all of this will be completed on schedule or within cost projections. Delays, transmission constraints, construction costs—all these historically narrow the returns promised to investors.

But the capital direction is clear. Nvidia’s results confirm demand is still strong—94% profit growth, 73% revenue growth, $68.1 billion in quarterly sales. So the compute demand driving these investment decisions definitely isn’t waning.

This emerging business model combines two previously separate logics. On one side: the logic of energy infrastructure operators—maximize uptime, minimize cost per megawatt-hour, negotiate power contracts that protect margins. On the other: the logic of compute service providers—attract intensive workload clients, sign long-term contracts that justify debt issuance, build recurring revenue bases that eventually convince creditors to lower coupons.

The success of this model depends on whether they can narrow the spreads before current debt matures. If they can refinance in 2-3 years at 5-6% instead of the current 9%, the business improves structurally. But if offtake contracts aren’t renewed, clients switch to proprietary infrastructure, or energy prices rise faster than compute service revenues, expensive debt becomes a burden with narrow returns, forcing equity dilution or restructuring.

For those evaluating exposure to this segment, the question isn’t whether bitcoin miners’ migration to AI makes sense as a long-term thesis—of course it does. The question is what capital structure makes sense to own. Debt at 9% offers a return with priority in liquidation but limited upside. Equity captures appreciation if the model works but absorbs losses first if contracts don’t hold. The spreads on these bonds aren’t just market data—they’re entry prices for a question that still has no answer.