The lender that never asked what you earned

I co-signed a loan once.

It was for my younger brother. A used car, four years, nothing dramatic. I read the page twice and my hand still shook when I signed it.

That is what it feels like to put your name on somebody else's future.

On July 1 this year, the federal government stopped lending graduate students whatever their school said the degree cost.

The program was called Grad PLUS. It ran for twenty years. It had almost no underwriting — no income test, no debt-to-income ratio, only a light check for adverse credit — and it would lend up to the full published cost of attendance.

About 440,000 students a year used it. At the end of 2024 there was $112 billion of it outstanding, spread across 1.8 million people.

For twenty years the price of a graduate degree was set by a lender that never asked what you earned.

Now there are caps. A master's student can borrow $20,500 a year from the government and $100,000 in total. A medical or law student gets $50,000 a year and $200,000 in total. There is a $257,500 lifetime limit on top.

Medical school costs more than that.

So the difference has to come from somewhere else. It comes from private lenders now, and private lenders ask questions.

The part I keep coming back to is who ends up owning the paper.

The largest private student lender in the country wrote $7.4 billion of loans last year and holds roughly two-thirds of the market. It has guided to 12% to 14% origination growth this year, and it puts the graduate opening at about $5 billion once it fully arrives.

In the last quarter of 2025 it signed a funding agreement with KKR. Under that model, newly written loans get sold on — in some cases before the money has fully reached the school.

So the loan changes owner before the student sits down in the lecture hall.

I have no idea how these loans will perform. Nobody does… there is no cohort of privately underwritten doctors and lawyers to look back on. The federal program was the only lender most of them ever had.

A market that priced nothing for twenty years has started pricing. That is the story.

This connects directly to how graduate school gets paid for now that the government has capped what it will lend.

Sallie Mae signed an agreement with KKR in the fourth quarter of 2025. Its chief financial officer described it to investors this spring as a strategic partnership funding model.

In plain terms: the bank writes the loan, then sells it — in some cases before the school has received the full disbursement. It keeps a fee. The credit risk goes to a fund.

The company told investors it expects its own loan book to be flat to slightly down in the near term even as originations grow 12% to 14%. Both things can be true at once when you sell what you write.

It also guided operating expenses up about 16% at the midpoint, for graduate product development, new credit models and marketing. And it warned that the first year of graduate marketing will be inefficient. I appreciate the honesty.

The bank keeps the fee. Somebody else keeps the doctor.

And here is where it spreads. Somebody has to underwrite a 24-year-old with no income and eight years of tuition ahead of her.

Protect Borrowers looked at 34 private student lenders this year. The median minimum credit score was 670. The median share of loans with a co-signer was 77%. Underwriting looks at income and debt-to-income too.

None of these loans carry income-driven repayment. None carry Public Service Loan Forgiveness. None are required to offer you anything if you lose your job.

Now the odd part. Banks were leaving this business right before the demand for it appeared. Wells Fargo left years ago. Discover agreed in July 2024 to sell its $10.1 billion book of prime student loans to Carlyle and KKR for as much as $10.8 billion, with Nelnet's Firstmark unit taking over the servicing.

Carlyle had already bought a $415 million student loan portfolio earlier that year and invested in a platform that helps banks offer these loans without holding them.

So the deposit-funded lenders sold, and the funds bought. Total private education debt outstanding was $138.8 billion in March. The money to fill a new gap has to come from the buyers, not the sellers.

Meanwhile, a much older set of lenders has started filling the same gap, and it funds itself in the municipal bond market.

Connecticut's higher education loan authority launched a new graduate program this year and said plainly why: the federal graduate program had ended. Rhode Island's authority expanded its graduate lending with income-based repayment features and incentives aimed at nurses. Massachusetts broadened its graduate lending too, and it remains the only large state program that lends nationwide.

These agencies raise money through tax-exempt bonds or state-backed capital. Several of them predate the modern federal student loan programs entirely.

So a municipal bond fund in a retirement account can now hold a claim on somebody's second year of medical school. That is a new kind of collateral for that market.

The federal rate for graduate borrowing this year is 9.07% fixed. That is the number every one of these lenders is pricing against. Interesting, given none of them can print money.

Three more things worth keeping track of this week.

Used electric cars turned in the auction lanes. Cox Automotive's Manheim index came in at 207.4 in the first half of August, down 1.2% from July and flat against August 2025. Electric vehicle values fell notably after months of gains, and wholesale days' supply sat at 27.7, about two and a half days higher than a year ago. Those are the 2023 leases coming home.

There were 4,250 federally insured credit unions in the first quarter of this year, down from 4,411 a year earlier. That is 161 institutions gone in twelve months, about three a week. And it happened in a good year: system net income ran at $20.4 billion annualised, up 30.5%, with delinquency at 85 basis points against 80 a year before. Consolidation is not usually what strength looks like.

Moody's sampled 121 companies that pay suppliers through supply chain finance programs. Forty-seven of them take more than 90 days on average to repay the bank that fronted the money. Moody's calls 90 days the reasonable upper limit for treating that as a trade payable rather than debt. Above it, the obligation still sits in payables, where the debt ratios cannot see it. We'll see.

In 1971 Yale sold its students a piece of their own future.

The plan had a dull name. The Tuition Postponement Option. James Tobin helped design it, and he won the Nobel prize in economics ten years later.

A student could postpone tuition in units of $1,000. For every $1,000 postponed, he agreed to pay 0.4% of his income every year after graduation. The minimum payment was $29 a month. Repayment could run as long as 35 years.

The idea came from a 1955 paper by Milton Friedman, who argued a lender could take a share of a graduate's earnings instead of a fixed debt.

But Yale did not want to carry the losses. So it made the students carry each other.

Everyone who signed up in a given year went into a cohort. The cohort kept paying until the whole cohort's balance was cleared. If your classmate stopped paying, you paid for him.

The ones who did best got out first. Everyone else stayed.

Graduates who earned well could buy their way out early, by paying 150% of what they had postponed plus interest. Plenty of them did. That left the lower earners splitting the remaining balance among fewer people.

Then inflation arrived, tax rules changed, and some borrowers stopped paying altogether. The balance kept growing.

Yale ran the plan from 1971 to 1978 and then shut it. After years of alumni anger and threats of lawsuits, it changed the terms in 1999 and covered what was left. The final payments came in during 2001 — thirty years after the first student signed up.

Yale's president called it an experiment with good intentions and several design flaws.

Every generation gets offered its own version of this. Borrow against the future, pay from what you earn, and let the group absorb whatever goes wrong.

The group is never the one that set the price. We'll see.