UBER — Uber Technologies, Inc.

Deep-value analysis — Uber Technologies, Inc. (NYSE: UBER) — RERUN (refresh)

Analysis date: September 11, 2026 · Reference price: … (close … on 09/10/2026, tracker) · Market cap: … bil. · Shares outstanding: 2,042,560,121 (10-Q Q2 2026 cover, as of 07/31/2026)

Regime: REFRESH. The reference thesis is from 08/18/2026 (rapoarte/deep/2026-08-18-deep-UBER.md). The mechanical triage deep_delta.py from 09/08/2026 flagged a single falsifier — shares — and requested a rerun. This analysis’s first task was to check whether the falsifier was telling the truth. It wasn’t, but it wasn’t entirely wrong either — see the “What changed” chapter, point 0.

What I explicitly inherit, since nothing new was filed with the SEC between 08/18/2026 and today (the last filing is still the Q2 2026 10-Q from 08/05/2026, already in the file at the reference analysis): the business structure and moat evaluation; the capital-allocation analysis; the balance sheet line by line Q2’25→Q2’26; the entire O’Glove earnings-quality analysis, including the insurance-float series and the debt maturity schedule; the CEO’s Outsider profile; the nine accounting red flags; the free-cash-flow bridge and the owner earnings base of …, which is built on TTM Q3’25–Q2’26 and can’t move without a new report.

What I re-derive: (1) the audit of the broken falsifier, verified against the primary filing; (2) the entire valuation triangulation at the new price of … including a new 20,000-scenario Monte Carlo simulation; (3) four events that appeared after 08/18/2026 that change pre-mortem probabilities, not the figures; (4) the tracker-comparison chapter, which now has a calculated GBL score — it didn’t at the reference analysis.

Primary sources: FY2025 10-K (02/13/2026), Q2 2026 10-Q (08/05/2026), Q1 2026 10-Q (05/06/2026), FY2024 10-K (02/14/2025), 2026 DEF 14A (03/23/2026), SEC XBRL companyfacts (CIK 0001543151), data pack data-pack-UBER-20260911.md. The 09/11/2026 research brief is treated as testimony: its figures are checked against filings and a material error in it is explicitly corrected (“What changed” chapter, point 4).

Executive summary

The thesis, unchanged in substance. Uber is a rapidly expanding cash machine whose reported free-cash-flow figure contains three temporary subsidies nobody strips out. TTM operating cash flow (Q3’25–Q2’26) is …, capex …, so the “free cash flow” both the company and brokers publish is … — a 6.83% yield at today’s market cap. Stripped out: … of stock-based compensation, which Uber adds back as “non-cash” but neutralizes by spending … in cash on buybacks in six months; … of tax subsidy (effective cash tax paid of … TTM versus … at a normalized 22% rate on operating profit of …); and … of finance-lease principal. The result is normalized owner earnings of …, i.e. … per share and a 4.68% yield at … — not 6.83%. The figure is unchanged from 08/18/2026 because the TTM period is unchanged: nothing new was filed.

What actually moved. The price fell 3.2% versus August’s reference of … (and 7.5% versus … the quote at the time of the triage). Nothing from filings. The falsifier that triggered the rerun — “share count rises above 2,100 mil.” — was triggered by a unit mismatch: teza_deep.py reads sharesOutstanding from yfinance in raw units (2,042,560,121) and compares it to a threshold written in millions (2,100). Any company with more than 2,100 shares issued automatically breaks this falsifier. The real figure — 2,042.56 mil. — is below the threshold, comes from the 10-Q cover from August 5, and was already known at the reference analysis. The rerun was triggered by a false positive.

But the underlying signal is real and worth stating. Between the 06/30/2026 balance sheet (2,039.99 mil. shares) and the 07/31/2026 10-Q cover (2,042.56 mil.), the share count grew by 2.57 mil. in a single month — the first monthly increase after five consecutive quarters of decline. The cause is visible in the cash flow statement: buybacks fell from … in Q1 2026 to … in Q2 2026, at a practically identical price (~…), because … bil. went into the Delivery Hero stake. RSU settlement continued unchanged. In other words: the spirit of the falsifier — “a rise in share count would signal the program freezing and a return to net dilution” — is closer to the truth than the mechanical test shows, just at a scale of 2.6 mil. shares, not 60.

Estimated value. Five independent models, re-priced at …: EPV Greenwald … (…), DCF bear without insurance float … (…), DCF base on normalized owner earnings … (…), historical P/FCF multiples 18x at … (…), DCF bull on fully reported FCFE … (…). Median: … …. The Monte Carlo simulation over 20,000 scenarios, with the same central assumptions (OE … · g1 12% · r 9.5% · gt 2.5% · net debt 0 · 2,042.56 mil. shares), gives a median intrinsic value of … and a …% probability of undervaluation, with P10 at … and P90 at ….

What I changed from August: only the price. This is a deliberate decision and it’s the most important sentence of this refresh. The assumptions (oe … · g1 12% · r 9.5% · gt 2.5% · nd 0) are identical. In a rerun triggered by a false positive, with no new filing, rewriting the assumptions would have been an opinion move disguised as a facts move. The consequence is that the entire margin improvement (from … to … median, from …% to …% undervaluation probability) is exclusively the arithmetic of the price decline, not a business improvement. Anyone reading only the numbers needs to know that.

Verdict: MONITOR, optional small position below … Unchanged. Uber isn’t cheap in the Graham sense: P/B 5.42x, zero dividend, reported P/E × P/B = 86 against the 22.5 threshold (135 on normalized P/E), and the no-growth earning-power value covers 42.5% of the price — the remaining …% is prepayment for growth that depends on the robotaxi race. The tracker’s GBL score, calculated in the meantime (09/06/2026), is …% — SPECULATIVE, within three-tenths of the “~48%” manual estimate made in August, when the column was empty. Convergence is complete and detailed in the final chapter.

Three new events move probabilities, not figures: Delivery Hero’s boards recommended accepting the offer (09/03/2026) and a pre-negotiated antitrust remedy — selling the business in 14 overlapping markets to SSW Partners for … bil. — makes closing more likely, so pre-mortem scenario 3 (leverage + frozen buybacks) rises in probability; Nevada approved robotaxi fleets for Tesla (up to 5,000 vehicles in 12 months), Uber and Waymo on 08/20/2026, i.e. exactly the multiple-fleets-competing-for-distribution case; and the first bellwether trial out of over 3,900 sexual-assault case filings enters jury selection on September 14, 2026, three days away, with an aggregate legal reserve of … bil. that fell from … bil. in six months.

Position sizing. …% of scenarios give a margin above 30%, …% are underwater. The right tail (P90 …, P95 …) remains longer than the left (P10 …, P5 …). It’s the profile of an option, not a value position. Below half of a normal allocation, adding only below … At a minimum GBL threshold of 50% in the current plan, UBER isn’t even eligible today (…%).