We are living through the start of the largest private debt cycle in history. This is not a prediction anymore. The bonds are already printing.
Looking at the numbers, building the data centers, power infrastructure, and factories of the AI economy will cost between 3 and 5 trillion dollars over the next several years, and some estimates put the total at 7 trillion by 2030. The hyperscalers funded the first half a trillion from their own cash. That phase is over.
In 2025, the five largest technology companies issued 121 billion dollars in bonds, compared with an annual average of 28 billion over the five years before. Wall Street expects roughly 570 billion dollars of AI linked debt issuance globally in 2026 alone. Morgan Stanley frames the decade in three numbers: 2.9 trillion of data center investment through 2028, 1.4 trillion coverable by Big Tech cash flow, and a 1.5 trillion gap that debt must fill. Around 800 billion of that gap is expected to land in private credit.
Why debt and not equity? Because that is how this era works. Record valuations have coincided with a borrowing boom, while equity issuance has stayed flat. Dilution isn’t an option for management. Twenty years of buyback culture made it taboo, and confident leadership teams treat their own stock as the most expensive currency they own. Meanwhile, rising equity values quietly expand debt capacity: a larger equity cushion means tighter spreads, and tighter spreads mean the same cash flows can carry more debt. The market has even built instruments that make this explicit. One AI infrastructure company borrowed billions through a convertible bond at a coupon rate under 2%. Equity value is literally subsidizing the cost of debt. High valuations are the throttle on this supercycle, and it's wide open.
Most of the commentary misses that a supercycle of this size does not need to deny anyone a loan in order to change everyone's cost of capital. The Federal Reserve Bank of Dallas ran the arithmetic: this year's AI related investment grade issuance creates duration supply equal to roughly an eighth of what the US Treasury itself will ask markets to absorb. Duration at that scale pushes term premia higher, and term premia set the discount rate for every borrower in the economy. At the same time, the balance sheets that fund ordinary businesses are rotating. A quarter of bank lending to nonbank financial institutions now flows to private credit firms, up from one percent in 2013. Life insurers hold nearly a trillion dollars of private credit. All of that capital is tilting toward AI adjacent paper because it yields more. The early evidence is on the tape: hyperscaler bond order books that covered five times in February covered under two by July. Repricing has begun and is propagating.
So who pays for the AI buildout? Everyone who was not invited.
I mean specific companies:
The remittance firm prefunding corridors in 40 countries.
The payments company holding settlement reserves.
The earned wage access provider funds payroll advances, which are repaid two weeks later, like clockwork, by payroll itself.
The neobank funding its advance product.
These are some of the highest quality borrowers in finance: short duration books, self repaying receivables, default rates that would embarrass most investment grade portfolios. They are also exactly the credits that big allocators drop first in a supercycle because the tickets are too small and too short to move a portfolio chasing trillion dollar infrastructure. Their reward for being boring and reliable is a higher rate on every refinancing, driven entirely by someone else's buildout.
And on the other side of the same economy sit household deposits, still earning a fraction of one percent, while the institutions holding those deposits chase private credit yields.
That gap is the entire reason we wrote Liberation of Money, and the supercycle is now widening it from both directions at once.
Sentora's earn products connect those two sides directly. Households and businesses deposit stablecoins into risk curated vaults and earn yield generated by real borrowers. Those vaults fund exactly the borrowers being crowded out: treasury facilities for payment and remittance companies, earned wage access books repaid through payroll, working capital for fintechs that today pay banks SOFR plus a wide spread for the privilege of being ignored. Every position is short duration and self liquidating by design. We will not fund fifteen year infrastructure with demand deposits, and we will not chase AI paper alongside every allocator on earth. Our strategy is simpler: be the refinancing venue for the good credits the crowd repriced on its way past.
We built the infrastructure for this moment. More than 1,000 risk models across 40 protocols and 12 blockchains. Over 3 billion dollars deployed through our platform. An insurance layer that backstops covered events. The rails are live, the first facilities are being structured now, and the arithmetic gets better for us every time the term premium ticks higher.
I will make this falsifiable, because confident claims should be. If middle market borrowing costs outside AI do not rise relative to equivalently rated paper within the theme over the next 18 months, I am wrong about the crowding. I think that is unlikely. The largest debt cycle in history has a bill; the bill is being distributed to businesses that had nothing to do with it, and the firms that build a better funding channel for those businesses will define the next decade of credit.
That is the gap. We intend to fill it.
If you run treasury at a payments company, a remittance firm, or an earned wage access provider, and your facility is repricing this year, my team would like to show you the math.
