Used-car retail stopped being a lot business and became a pricing-engine business: instant appraisal, algorithmic inventory routing, national inventory visible from a phone. CarMax (KMX) reports fiscal Q2 with consensus at $6.97 billion in revenue and $0.72 in EPS, and the line that decides whether that software investment earns its keep is gross profit per vehicle.
The Development
The setup going in is a split one. Retail demand is soft, gross profit per vehicle is declining, wholesale volume is rising, and cost cuts are doing visible work on the expense line. Those four facts describe a specific machine state.
Wholesale is the throughput channel — cars sourced and auctioned onward to other dealers, thin margin, fast turns. Retail is the monetization channel, where the appraisal algorithm's accuracy converts into spread. When wholesale volume climbs while retail softens, the sourcing and logistics layer is running at full tilt and the layer that actually prices the asset is taking less per transaction.
The Revenue Bridge
Every platform business has a revenue-per-unit metric. In software it's ARPU; here it's gross profit per retail unit, and it behaves the same way. A $6.97 billion revenue figure can be cleared on volume alone, because volume is the easiest variable in a business that can buy inventory aggressively. EPS cannot be cleared the same way if per-unit profit is compressing.
That leaves cost reduction carrying the earnings line, and cost reduction is a one- or two-quarter lever. It flatters the print without repricing the unit. The investment case for rebuilding a used-car retailer around software was always that better appraisal accuracy would widen spread per car, not simply move more cars through the same funnel. This quarter is the first clean look at which one the technology actually delivered.
Across the Stack
The obvious read-through is silicon. Texas Instruments (TXN) trades at $278.05 after clearing $275.21, with demand cited across AI, data centers, industrial and automotive, a recently increased dividend, and capital expenditure coming down off its build-out peak. Automotive analog content is the consensus reason to like the name here, and it is largely embedded at this price.
It is also the wrong lever for this story. Analog content per vehicle is a function of cars built, not cars resold — used-vehicle churn does not put a single additional power-management die into the supply chain. What is carrying TXN at these levels is the data-center and industrial mix, with easing capital intensity turning that revenue into free cash flow faster than it did during the fab build.
TXN is holding a cushion of just $2.84 above the level it broke through — a little over 1%. That is a thin floor for a stock being asked to price two separate demand cycles at once.
The less-crowded beneficiary sits at the connectivity layer. Connected-vehicle service attaches to the installed fleet rather than to new production, so a high-churn resale market structurally recirculates subscriptions rather than shrinking the base. Verizon (VZ) closed at $47.08 on September 25, 8.49% below its mid-September peak and still up 15.6% for 2026, with CEO Dan Schulman having framed Q3 as the proving quarter. The pullback came alongside the Federal Reserve's first rate hike since 2023 and competitive noise in wireless subscriber guidance — not from anything that changes the fleet math.
The Leading Indicator
A calibration note first: my recent macro-anchored calls have missed, including a bet that Brent clearing $88 would become an explicit cloud-cost variable. It never got there. So the levels below sit on prices these names have actually traded through, not on round numbers I find aesthetically pleasing.
TXN is where this argument is most directly tested. A close above $282.00 within 10 sessions says the AI and industrial mix is carrying the stock on its own merits. A close back below $275.21 — the level just reclaimed — says the breakout was thin and the automotive leg of the story is being marked down. What would change my mind on the CarMax side is straightforward: gross profit per retail unit rising alongside wholesale volume would mean the pricing engine is extracting spread rather than subsidizing it, and the software thesis works.
Watch gross profit per retail used vehicle in Tuesday's print. One number, and it grades everything above it.