Post
Tesla posted record deliveries and fell 14%. What broke was not demand, it was earnings.
A dated timeline of what actually happened to Tesla in July 2026, and three judgments about how to read it. Deliveries hit a record, operating margin collapsed to 1.4%, and the trailing P/E sits at 295x. Reading that combination as simply overvalued gets you halfway.
On July 22, 2026 Tesla reported record quarterly revenue. The next day the stock fell about 14.5%. Deliveries were also a record.
When revenue and deliveries both set records and the stock breaks anyway, the problem is not demand. It is earnings. That is what happened here. Below is a dated timeline of the news, followed by how I read it.
The numbers come from two places and I have kept them separate. Values Opula computed carry their as-of date. Values I verified on the web are labeled as web figures.
How the price actually moved
Look at one horizon and you get one conclusion. Stack them and the story changes.
| Horizon | Tesla | Consumer Discretionary (XLY) | Technology (XLK) |
|---|---|---|---|
| 1 month | -16.55% | -3.48% | -4.71% |
| 3 months | -16.81% | -7.82% | +9.77% |
| 6 months | -30.29% | -11.14% | +21.22% |
| 1 year | -0.96% | -2.69% | +34.26% |
Opula, as of 2026-07-25.
Three things read off that table.
A one-year return of -0.96% does not mean a quiet year. It means the stock went up and gave all of it back. The giving back is concentrated in the last six months, at -30.29%.
Essentially the entire three-month decline happened in the last month. -16.81% and -16.55% are almost the same number. April through June barely moved. The drawdown compressed into four weeks.
This is not tech selling off. The technology sector is up 34.26% over one year. Tesla is down 0.96% over the same window. That gap is company-specific, not sector-wide.
Since the reported quarter ended on 2026-06-30, the stock is down 25.58% (Opula, as of 2026-07-25). Most of the damage happened around the print.
The July news, in order
| Date | News | Direction |
|---|---|---|
| Jul 3 | Driverless robotaxi service goes live in Miami, the third market after Texas and California | Positive |
| Early Jul | Q2 deliveries of 480,126 vehicles, up roughly 25% year over year, well ahead of the ~406,000 estimate | Positive |
| Jul 21 | Robotaxi expands to Orlando and Tampa | Positive |
| Jul 22 | Q2 results: record revenue of $28.24B, EPS below consensus | Mixed |
| Jul 22 | Operating income $398M, down 57%. Operating margin 1.4% versus 4.1% a year ago. First negative free cash flow since early 2024, with capex nearly doubling to about $6B | Negative |
| Jul 23 | Stock falls roughly 14.5%, shedding about $140B in market value | Negative |
All web figures. Sources: Electrek on the Jul 22 results, Electrek on the Jul 23 selloff, CNBC, Jul 22, Investing.com on the Q2 slides, Motley Fool on deliveries and the Miami launch.
Read the table vertically and it looks strange. Four positive items, two negative ones, and the stock traded only on the negative ones. That asymmetry is the whole quarter.
The consensus number differs by source
The size of the miss depends on whose estimate you use, so here are both.
| Item | Opula | Web |
|---|---|---|
| Reported EPS | $0.33 | $0.33 |
| Consensus | $0.5353 | $0.49 |
| Miss | -38.35% | -32.7% |
Opula as of 2026-06-30. Web figures as reported on 2026-07-22.
The actual result agrees. The estimate does not. Which analyst panel, collected on which date, and adjusted versus GAAP will separate two consensus figures by that much. The right move is to print both rather than pick one and delete the other. The direction is identical either way and the magnitude differs by about five percentage points.
Judgment one: this was a spending miss, not a demand miss
Deliveries grew roughly 25% year over year and beat the estimate by more than 70,000 units (web). Operating income fell 57%. Units did not shrink. Units grew and profit shrank anyway.
In the same quarter capex nearly doubled to about $6B, which pushed free cash flow negative for the first time since early 2024 (both web). Margin compression and a spending spike landing in the same quarter has a simple reading: current profit is being routed into future capacity.
So summarizing this print as "demand cracked" does not match the table. More precisely: the selling worked, and the company is building the next business at the same time, so nothing was left over. Those two diagnoses point at different follow-up metrics. A demand problem means watching deliveries. A spending problem means watching operating margin and capex against revenue.
One part I cannot verify. How much revenue the robotaxi service is actually recognizing is not in the data Opula sees. So this piece does not prove that the spending is converting into revenue. The interpretation of direction is where it stops.
Judgment two: a P/E of 295 does not mean expensive
Opula puts Tesla's trailing P/E at 295.31x against a peer median of 27.93x across Apple, Amazon, Meta, Alphabet, and Nvidia. That is the 100th percentile of the group, as of 2026-07-25.
A multiple is a division. It has two ways to get big and they mean opposite things.
P/E = price / earnings per share
path 1: price rises -> it got expensive
path 2: earnings shrink -> profit collapsed
This is the second one. The stock is down 25.58% since the quarter ended and the multiple still sits at the top of its peer group. A multiple rising while the price falls means the denominator fell faster. That denominator is the 1.4% operating margin (web).
So reading 295.31x as "expensive" gets you halfway. Precisely, it means current earnings cannot explain this price. What holds the price up is not this quarter's profit but a business that has not shown up in revenue yet. It is a plot of land priced at the presale value of a building that has not broken ground. If the building goes up, the price was justified. If it does not, you own dirt.
The reason that frame matters: change the scoring rubric and the same number flips. Score it as a car company's earnings and 295x is unexplainable. Score it as an autonomy service company's build-out phase and negative free cash flow reads as evidence of progress. One set of numbers, two conclusions.
Judgment three: blaming the sector is a misread
I have seen this decline explained as tech selling off. The table says otherwise. Technology is up 34.26% over one year and 21.22% over six months. Tesla is down 30.29% over six months (Opula, as of 2026-07-25).
Consumer discretionary is genuinely weak at -11.14% over six months, but that is about a third of Tesla's drawdown. Some of the move is sector. Most of it is not, and what is left is this company's earnings problem.
How this judgment gets proven wrong
Pick a direction and you owe the counter-case.
If the roughly $6B of capex that currently reads as a problem starts landing in revenue and margin next quarter, the interpretation flips with no fact changing at all. A pile of cash being burned gets reclassified as an asset generating cash. Record deliveries and the robotaxi market expansion are live ingredients for exactly that.
If operating margin is still sitting near 1% next quarter, the "building, so nothing is left over" explanation ages by another three months. At that point the frame itself needs re-examining, not the stock.
So the watchlist narrows to three things.
- Whether operating margin turns from 1.4% (versus 4.1% a year ago, web)
- Whether free cash flow returns to positive, and whether capex keeps climbing
- Whether robotaxi starts being recognized as its own revenue line
The one column this analysis cannot fill
Everything above is public data. Anyone can get it, and asking an AI will produce something similar.
One column is missing. What percent of your net worth is this?
At 5% of your assets, a six-month move of -30.29% costs you 1.5 percentage points overall. At 38%, it costs 11.5 points. Same analysis, same numbers, different conclusion. No stock analysis article can fill that column for you.
Indirect exposure leaks in two more ways.
The slice inside your ETFs. You can own no shares directly and still hold the position through a fund. It takes a look-through to see the effective weight.
Holdings that look like diversification. If an international equity ETF you own moves with the stock at a correlation of 0.57, it is roughly half the same bet. The number of tickers went up. The actual diversification did not go up as much.
That is exactly what Opula's
get_insight_report does. It surfaces only the analysis frames whose numeric test actually fired on current data (two fired for Tesla here), attaches an as-of date and a source to every figure, and lands the report on the user's real book: direct weight, ETF look-through, and correlation-based indirect exposure. Every Opula number in the tables above came out of that one response.The limits, stated plainly. Opula returns no news. The timeline above came entirely from web search, which is the LLM's job in this split. There is no account aggregation either, so holdings get recorded through conversation. Quote coverage is US-first. The discriminant thresholds are not yet calibrated against live use, so they can over-fire. And price targets and buy or sell calls are absent by design.
Questions people ask
Isn't a high P/E just expensive?
No. High because the price rose means expensive. High because earnings fell means profit collapsed. Two opposite situations produce the same number. You separate them by reading the price trend and the earnings trend together.
What happens to P/E when earnings are negative?
It goes negative or gets suppressed entirely. That is why a company with earnings near zero cannot be judged on the multiple alone, and why a print like this one sends the ratio into three digits.
Can I hand stock analysis to an AI?
Gathering the material, computing it consistently, and laying out the inputs for a decision is something it does well. Making the decision is a different job. Opula stops at computation and diagnosis, and does not produce targets or buy and sell calls.
Where did these numbers come from?
Price returns, the P/E, the peer comparison, and the EPS surprise are Opula's computed values, printed with their as-of dates. Revenue, operating income, free cash flow, deliveries, the robotaxi expansion, and the single-day drop are web figures with sources linked. No figure from one source was merged into a figure from the other.
The short version
Record deliveries and collapsed earnings landed in the same quarter. The stock traded the earnings.
A trailing P/E of 295.31x is not a signal that this company is expensive. It is a signal that current earnings do not explain the price. That distinction changes what matters next quarter: not demand, but margin and capex.
This is not a buy or a sell opinion. It is the material and how I read it. The last column, what percent of your money this is, only you can fill in.