Auto Debt Hit a Record $1.69 Trillion. Calling Your Whole Database Got Cheap at Exactly the Wrong Time.

The weekend's number, and why it lands on your call list

On August 1, the Federal Reserve Bank of New York's latest household debt figures made the rounds: Americans now owe a record $1.69 trillion on auto loans, roughly 9% of total household debt and the second-largest consumer category after mortgages. Balances grew by $18 billion in the first quarter of 2026 alone, on about $182 billion of new originations.

For most readers that is a macroeconomic story. For a dealership, it is a list-quality story.

Your customer database is not an abstraction. It is a few thousand people, each attached to a specific vehicle, a specific balance, and a specific payment. Every equity-mining campaign, every "you may be in a great position to upgrade" call, and every AI outbound sequence is a bet on what those balances look like. The 2026 data says that bet has gotten worse - and the tooling has gotten cheap enough that most stores will place it more often anyway.

The myth: More outreach at a lower cost per contact means more sold units. The data: 30.9% of new-vehicle trade-ins in the first quarter of 2026 were underwater, by an average of $7,183 - the highest share since early 2021. Subprime 60-day delinquencies are at their highest level since Fitch Ratings began tracking the metric in the early 1990s. Meanwhile AI voice outreach now runs roughly $0.05 to $0.50 per minute on published vendor pricing. Cheap dialing does not make a payment work. Sources: Edmunds, 2026; Fitch Ratings, 2026; Aircall and Retell AI pricing guides, 2026.

What actually changed in the owner base

The headline debt number is the least useful figure in the release. The useful ones are underneath it, and they describe the composition of the people in your CRM.

Two details in that table are easy to skim past and shouldn't be.

The first is that distress is concentrated, not general. Prime borrowers have largely kept paying. TransUnion projects serious delinquency will stay roughly flat through the end of 2026, ticking to about 1.54% from 1.51%. Fitch's subprime index, meanwhile, is at a multi-decade high. Both can be true, and the gap between them is the whole point: the aggregate market looks stable while a specific slice of it is under real strain. Your database contains both slices, mixed together, sorted by nothing more useful than purchase date.

The second is that negative equity has stopped being an early-trade problem. Edmunds found the average age of an underwater trade-in reached 4.3 years in Q1 2026, the highest on record. These are not impatient buyers flipping out of a car after eighteen months. They are people who bought at pandemic-era prices, waited, and still could not outrun the depreciation curve. Of the underwater deals, 90.2% carried terms of at least 72 months and 43% stretched to 84. We wrote about what that does to the appraisal conversation; the point here is upstream of the appraisal, at the moment you decide whether to call at all.

The three lists hiding inside your one list

Most stores run outreach against a single segment definition - months in service, mileage, or a payoff estimate from the DMS. In a market where a third of trade-ins are underwater and subprime delinquency is at a 32-year high, that one list is really three, and they want three different conversations.

The middle row is where most of the damage happens. A customer who owes $7,000 more than the car is worth does not experience "great news, you are in a strong equity position" as helpful. They experience it as evidence that you did not look before dialing. The approval-rate data we covered last month showed lenders saying yes more often in 2026 - but saying yes with longer terms and less money down, which is exactly the mechanism that produced the underwater book you are now calling.

The bottom row is where the risk lives, and it is not only a customer-experience risk. Sen. Elizabeth Warren opened a congressional inquiry into auto lending and repossession practices in February 2026, sending letters to major lenders and buy-here-pay-here servicers after the collapse of subprime lender Tricolor. That inquiry is aimed at lenders, not at franchised dealers. But scrutiny of how stressed borrowers are contacted has a way of widening, and it arrives on top of the consent and record-keeping questions we covered in TCPA and AI calling. An automated sales pitch to someone who is 90 days down is a bad look in every direction.

Why cheap outreach makes list quality matter more, not less

Here is the part that runs against intuition. When calls were expensive - a BDC rep, a salary, a finite number of dials in a day - list quality was self-correcting. Scarcity forced prioritization. Nobody dialed 4,000 records because nobody could.

AI removed the constraint. Published pricing for AI voice agents in 2026 spans roughly $0.05 to $0.15 a minute on assembled infrastructure and up to about $0.40 to $0.50 on managed platforms with CRM integration, according to vendor pricing guides from Aircall and Retell AI. Read those numbers with the source in mind - they are vendors describing their own category - but the direction is not in dispute. Outreach volume is now effectively unbounded for any store willing to pay a four-figure monthly bill.

That changes what the binding constraint is. It is no longer how many people you can reach. It is how many people you can reach correctly before you burn the relationship. Every mistimed equity call to an underwater owner is a small, permanent withdrawal from a database you spent years and a lot of ad money building. At scale, cheap outreach against an unsegmented list does not produce more units. It produces more opt-outs.

The practical implication is unglamorous: the value of an AI outbound tool is now almost entirely determined by the data it reads before it dials, not by how natural it sounds. A voice agent that pulls a current payoff, an accurate valuation, and payment status before choosing a script is doing something genuinely useful. One that reads a name and a purchase date is an expensive way to annoy people faster.

What to measure this month

If you run a BDC or own the store, four numbers will tell you whether your outreach is aimed or merely loud.

1. Share of your outbound list with a verified payoff. Not an estimate derived from original term and purchase date - an actual or lender-sourced figure. If that share is low, everything downstream is guesswork.
2. Contact-to-appointment rate split by equity position. Run it three ways: positive equity, underwater, and unknown. If the underwater bucket converts at a fraction of the others and still consumes most of your dials, you have found the leak.
3. Opt-out and do-not-call rate per thousand contacts, tracked monthly. This is the cost of a bad list, and it is the one metric that gets worse silently. A rising opt-out rate alongside rising volume is the signature of an unsegmented campaign.
4. Suppression coverage. What share of contacts flagged as delinquent, in repossession status, or in an active service complaint are actually excluded from sales outreach? For most stores the honest answer is that nobody has checked.

None of those require a purchase. All four are answerable from your CRM and DMS this week, and they are the questions worth asking before you evaluate a single vendor.

The measured read

The macro picture is not a crisis for franchised dealers. Prime credit is holding, sales volume is stable, and forecasters are split on whether delinquency worsens from here. Nobody should redesign a store around one weekend's headline.

But the composition of the owner base has shifted in a way that specifically degrades the oldest, cheapest play in retail automotive - calling your own customers. More of them are underwater, they are staying underwater longer, and a small but historically large slice of them is under real financial strain. At the same time, the cost of calling all of them has collapsed. Those two trends pull in opposite directions, and the stores that do well over the next few quarters will be the ones that notice.

The database is still the best-margin lead source in the building. It is just no longer a list. It is three lists, and the work is telling them apart before the phone rings.

If you are evaluating an AI phone or outreach tool and want to pressure-test what customer data it actually reads before it dials, that is the specific question the team at Carbuki works through with dealerships.

Sources

• Americans' Auto Debt Hits $1.69 Trillion As Subprime Delinquencies Reach 32-Year High - HNGN, August 1, 2026, summarizing Federal Reserve Bank of New York household debt data, Fitch Ratings delinquency data, Cox Automotive repossession figures, and TransUnion projections
• Car Debt Grows Deeper as Loan Terms Stretch Wider - Edmunds Q1 2026 insights report, April 20, 2026 (30.9% negative-equity share; $7,183 average; $932 payment; 4.3-year average trade-in age; 90.2% at 72+ months; 43% at 84 months; 26% over $10,000; 9.3% over $15,000)
• 60-plus-day subprime auto DQs hit 32-year high - Auto Finance News, January 16, 2026, reporting Fitch Ratings data
• AI Voice Agent Pricing in 2026: Cost Breakdown, Comparisons, and ROI - Aircall, 2026 (vendor pricing guide)
• Call Center Outsourcing Costs in 2026: Full Breakdown and a Better Alternative - Retell AI, 2026 (vendor pricing guide)
• Auto Loan Defaults Are Increasing, But We Are Not Heading Into A Repo Crisis - Cox Automotive (repossession volume context)