Carvana's Ernie Garcia on near-bankruptcy in 2022, same-day delivery, and why AI fits their vertical model

Carvana's Ernie Garcia on surviving 2022 and why vertical integration is now an AI advantage

Carvana launched in January 2013 with a thesis that made Silicon Valley uncomfortable: that selling used cars online at scale required owning the inventory, the logistics, the reconditioning, and the financing — not a marketplace layer sitting on top of existing supply. Garcia pitched it that way and largely got turned away. "Silicon Valley did not love our pitch," he says. The company went public in 2017, four years old, and by Garcia's account had one of the worst IPO performances of that year.

The 2022 near-collapse

The business had a second crisis point after a COVID-era surge made Carvana look invincible. In 2022, that reversed hard, and Garcia is candid about why. The company over-indexed on growth and dramatic change at exactly the moment it should have been grinding on operational execution. "We got it wrong first," he says, when asked about the brutal stock chart. His framing for how the team got through it was deliberately unsentimental: most companies worth respecting have a period where they publicly look dumb, and that moment washes out the mercenaries. You rally around it and keep going.

“The idea was can we build a different kind of customer experience with different costs, more vertically integrated, that maximizes revenue in a better way so the customer experience can be really simple. We now have 50,000 cars — customers can get financing, get a trade-in value, toggle a warranty on or off, and have the car delivered to their door. We're only 2% market share and as we build more cars, we sell more cars.”

The machine today

Carvana now has roughly 50,000 cars on its platform and holds about 2% market share in used vehicles. Garcia believes demand exceeds what the company can currently produce — meaning the reconditioning and logistics capacity, not customer acquisition, is the binding constraint. Unit economics are running at two to two-and-a-half times what he describes as industry-normal returns.

The business sits at an unusual inflection: the model is proven, the returns are real, and the remaining opportunity is enormous precisely because the market share is so low. Garcia is explicit that the next five years of economic performance will be determined almost entirely by execution — making the machine "a little better and a little bigger every single day" — not by product reinvention.

Reconditioning as the bottleneck, and where robotics fit

The single biggest constraint on Carvana's scalability is reconditioning — the roughly $1,000 of parts and labor put into every car before it's listed. Garcia is watching humanoid robotics closely but is measured about the timeline. Unlike OEM manufacturing, where every car gets identical processes, Carvana's workflow requires inspecting each vehicle and determining what it individually needs. That variability makes full automation harder. Specific, repeatable tasks — tire changes, paint correction — will likely automate first. If robotics capabilities scale as fast as LLMs have, he says, the timeline compresses in ways that are hard to predict.

Why vertical integration is an AI tailwind

The AI discussion is where Garcia's original architecture choice pays an unexpected dividend. Because Carvana built every step of the transaction — financing, trade-in valuation, warranty, insurance handoff, document verification, title registration — as deterministic, service-oriented systems with no human negotiating each line item, the entire stack can now be exposed to an AI layer. A customer can ask a natural language question and get a complete, accurate answer because every underlying process is already automatable.

The competitive advantage is structural. Carvana's traditional dealer competitors have none of that vertical integration, so the gap in answer quality that AI can surface is wide. Garcia also notes that same-day delivery is now live, announced almost as an aside at the close of the conversation.

Internally, Carvana is moving on AI tooling at the same pace as most tech companies. The consumer-facing application, though, is arguably more defensible than most — built on a decade of proprietary transaction infrastructure that competitors cannot replicate quickly.