How Carvana’s insiders signaled a 99% crash— and what they’re doing now

For three consecutive weeks in late July and early August 2021, Carvana CEO Ernie Garcia III sold 60,000 shares of his own company every single trading day, at prices between $330 and $370. Across the nine months ending August 2021, Carvana insiders collectively sold $4.2 billion of stock — the largest portion coming from Garcia III’s father, Ernest Garcia II, the company’s founder-shareholder.¹ Sixteen months later, the same shares traded at $3.55.

This is the story of one of the most thoroughly telegraphed corporate collapses in recent memory — and how the SEC filings that gave it away are still publicly readable today on EDGAR. It is also a story without an ending, because the same family is selling again, in similar size, at similar prices.

The setup: $4.2 billion in insider sales before the peak

Between December 2020 and August 2021, Carvana insiders sold 15,183,340 shares amounting to $4.2 billion, at prices between $227 and $370.¹ The largest seller was Ernest Garcia II, with significant additional participation from directors, legal counsel, C-suite executives, and other key employees.

The selling was not subtle. Form 4 filings during the final three weeks before the peak show a steady daily cadence from Garcia III’s trusts: 60,000 shares per trading day from July 29 through mid-August 2021, executed in open-market transactions at prevailing prices.² The transactions were not executed under a Rule 10b5-1 trading plan — they were discretionary open-market sales by the chief executive at price highs.

Garcia II’s selling was at a different order of magnitude. Across the August 2020 to August 2021 window, his disposals totaled $3.6 billion³— making him one of the largest individual insider sellers in U.S. equity markets that year. Each sale was disclosed publicly on Form 4 within the SEC’s two-business-day window. None of it was hidden.

The market interpreted the activity as post-IPO liquidity. Carvana’s stock had risen from around $30 in March 2020 to an all-time high of $370.10 in August 2021. Used-car prices were elevated, the company was profitable for the first time, and the dominant narrative was that the pandemic had permanently restructured how Americans bought cars.

The insiders apparently disagreed.

What the Form 4 filings actually contained

A Form 4 is the SEC disclosure filed by an insider — an officer, director, or 10%+ owner — within two business days of a transaction in the company’s stock. The form discloses the date, price, share count, and resulting beneficial ownership. It also discloses two structural facts that matter enormously for interpretation: whether the trade was an open-market discretionary sale, and whether it was executed under a Rule 10b5-1 trading plan.

A 10b5-1 plan is a pre-scheduled trading arrangement that an insider sets up in advance, typically months before any actual trade occurs. Once filed, the insider has no further discretion over timing — the plan executes mechanically. This is the SEC-sanctioned mechanism for insiders to trade without facing accusations of acting on material non-public information.

Garcia III’s 2021 sales were not under a 10b5-1 plan. They were discretionary. He was choosing to sell, every day, at the prices the open market was offering, with full knowledge of the company’s internal financials.

For a single insider, that’s noise. For the chief executive of a company selling 60,000 shares per day for three straight weeks at price highs, while his father sells in $3.6 billion size in parallel, while every other named officer is also selling — that’s a cluster.

Cluster insider selling is the empirically strongest insider trading signal in the academic literature. Studies from Lakonishok-Lee and Cohen-Malloy-Pomorski have shown that consensus selling among multiple economic principals at price extremes correlates with subsequent negative returns at significantly higher rates than individual insider sales.

Carvana in summer 2021 was a textbook cluster.

How Tracenotes surfaces this kind of signal

Reading any individual Form 4 takes about 30 seconds. Reading hundreds of Form 4 filings from one company across a 16-month window, normalizing share counts across multiple trust structures, separating 10b5-1 plans from discretionary trades, and synthesizing the pattern into a thesis — that takes hours and requires working knowledge of SEC filing structure most retail investors don’t have.

Tracenotes runs that work continuously. Form 4 filings are ingested from EDGAR as they’re published, clustered by reporting entity, and separated from pre-scheduled 10b5-1 plans. The cluster signal sits alongside AI-narrated filing-language deltas in the signals feed.

The raw Form 4 data is public. What’s missing for most retail investors is the layer that synthesizes a multi-month, multi-entity filing pattern into a paragraph an investor can actually read. Tracenotes does not predict price moves; no model can. It collapses the work of reading the filings into the conclusion the filings support — with the supporting trades cited.

Current insider activity for CVNA is viewable at /insider/cvna.

The collapse

Carvana’s fundamentals began deteriorating almost immediately after the 2021 insider selling stopped. The company swung from net income of $45 million in Q2 2021 to a net loss of $68 million by Q3 2021³— within a single fiscal quarter of the CEO’s final daily sales.

The macroeconomic environment then turned against the business model. As the Federal Reserve hiked rates aggressively through 2022, monthly payments on financed used cars rose sharply, choking off Carvana’s primary customer cohort. The company’s inventory — much of it purchased at the peak of the 2021 used-car bubble — had to be written down. Consumer discretionary income compressed. Carvana’s debt-financed acquisition of ADESA’s U.S. physical auction business, announced in February 2022, brought total debt to roughly $9 billion just as cash flow was inverting.

The stock fell 18% in May 2022, after which the company laid off more than 2,500 workers. It continued to plummet through November 2022, when Carvana laid off another 1,500 employees.⁴ By the end of 2022, the company had reported seven consecutive quarters of losses totaling $3.5 billion,³ and was facing serious bankruptcy concerns. Carvana plummeted to an all-time low of $3.55 on December 7, 2022 — a 99% plunge from the record high of $370.10.⁵

The company restructured its unsecured debt rather than filing for Chapter 11. It survived, but barely. Equity holders who bought at the 2021 peak were close to wiped out.

In April 2022, with the stock down 70% year-to-date and trading around $80, Garcia II returned to buy 5.1 million shares for $408 million.⁶ He bought his own stock back for less than a quarter of what he had been selling it for nine months earlier. The stock then fell another 95% from his buy price before bottoming. Even his timing on the way back in was, in retrospect, early.

The receipts

The Carvana selling pattern is one of the most-studied insider trading cases in recent academic and investigative work.

“What I’m saying is the Garcias knew it was short-lived… The Garcias knew the music would eventually end.”

The United Association National Pension Fund has filed a securities class action against the company and Garcia III. Plaintiff attorneys are seeking discovery on “all communications to and from Garcia III concerning his knowledge of or benefit from his father’s sales in the class period,” which runs from May 2020 to July 2022.⁴ The case is ongoing.

Hindenburg Research published an in-depth report in 2025 documenting the pattern and its apparent repetition, calling out the structural parallel between the 2020-2021 selling and the company’s current behavior.³ The report compounded the analytical case with additional concerns about accounting and related-party dealings.

None of this is inference from price action. It is documented in real-time SEC filings any retail investor could have read at the time, with academic and investigative validation that arrived later.

The pattern returns — with one important difference

Carvana’s stock has been on one of the most extraordinary recoveries in recent market history. After crashing 99% by the end of 2022, the stock is now up roughly 10,961% in two years.⁷ Wall Street analysts have largely embraced the turnaround narrative.

Despite that recovery, Ernest Garcia II has aggressively sold stock again. He dumped $1.4 billion in Carvana stock over the last year, at times selling on a near-daily basis, making him by far the largest insider seller of the company’s shares.³

Garcia III has also begun systematic selling. In one filing alone, Garcia III reported sales totaling 923,155 shares of Class A common stock executed on August 7, 2025, at prices ranging from $352.72 to $367.88 per share⁸ — a price range strikingly close to his August 2021 sales.

But there is one important structural difference between the 2021 and 2024-2025 cycles. The current sales are executed under a Rule 10b5-1 trading plan adopted on December 13, 2024.⁸ The transactions are pre-scheduled — Garcia has no further discretion over timing once the plan was filed.

This matters analytically. A 10b5-1 plan is specifically designed to insulate insiders from accusations of trading on material non-public information. By choosing this mechanism this time, the Garcias have made it structurally harder to mount the same legal argument the class action is making against the 2021 activity.

The form is cleaner. The direction is the same. The same family, with the same track record on this stock, selling in similar dollar size, at price levels strikingly close to the prior peak — under a mechanism designed to be unimpeachable.

The honest counter-case

A reasonable analyst could read the 2024-2025 selling differently. Three counter-arguments are worth taking seriously.

First, wealth concentration. Garcia II’s net worth is heavily concentrated in a single stock that has just rebounded 10,000%+. Any prudent financial advisor would recommend systematic diversification at this point. A 10b5-1 plan is the textbook mechanism for that diversification. The selling does not require an information-asymmetry explanation.

Second, macro environment. The 2021 macro setup — peak used-car prices, an imminent rate-hiking cycle, an over-leveraged balance sheet — is not the current setup. Rates are stabilizing, debt is restructured, and Carvana has demonstrated operational discipline through the 2023-2025 turnaround. The same insider behavior in a different macro context may not produce the same outcome.

Third, base rates. Insider selling alone has never been a perfect predictor of subsequent returns. Cluster selling at price highs has historical statistical edge, but the false positive rate is meaningful. Many companies whose insiders cluster-sell at highs do not subsequently collapse. Survivorship in the case-study literature creates selection bias.

None of these arguments dismisses the signal. They contextualize it. The signal still says: the same insiders are taking the same action they took before a 99% drawdown. What the signal does not say is what happens next.

What this means now

The job of an insider intelligence tool is not to call the top. It is to surface the pattern, narrate the context, and let the investor decide what to do.

Carvana’s CEO and his father sold $4.2 billion of stock at the prior peak, then watched their shareholders lose 99%, then bought back at the bottom. Today they are selling again, at prices close to the prior peak, in similar dollar size, under a 10b5-1 plan that makes the action harder to legally challenge but easier to repeat.

That is the signal. The decision is yours. Make it from the filings, not from the narrative.