PDD — PDD Holdings Inc.

Deep-value analysis: PDD Holdings Inc. (NASDAQ: PDD) — 2026-09-03

REGIME REFRESH. Resumption of the baseline analysis from 08/04/2026 (verdict then: MONITOR), triggered by the mechanical deep_delta.py triage: 2 new filings since the thesis date and 1 falsifier still unresolved. Delta pack: rapoarte/deep/delta-PDD-20260902.md.

Price at analysis: USD …/ADS (data pack 09/03/2026; tracker 09/02/2026: …) · Market cap: USD 117.07 bn · Shares: 5,693.6 mn ordinary shares = 1,423.4 mn ADS as of 12/31/2025; diluted average Q2 2026 5,894 mn ordinary shares = 1,473.5 mn ADS (1 ADS = 4 ordinary shares, confirmed in the results release in the 6-K from 08/25/2026: “Earnings per ADS (4 ordinary shares equals 1 ADS)”) · Exchange rate: 6.7094 RMB/USD (current rate, used in all models) · Foreign private issuer, 20-F, Cayman holding company with Chinese operations via VIE structure, executive headquarters in Dublin.

What is carried over without re-verification (no new filing touches these): the business and moat description, Colin Huang’s control structure (Class B, ~25% economic / >60% of votes), the zero-dividend/zero-buyback policy confirmed in Item 16E of all five 20-Fs (FY2021-FY2025, all “Not applicable”), the VIE architecture and HFCAA/PCAOB risks, the SAMR/tax investigation from December 2025 – January 2026, the December 19, 2025 leadership reshuffle (mirrored Co-Chairman), the elimination of the U.S. de minimis exemption (August 2025), the EU’s EUR 200 mn DSA fine (05/28/2026) and the EU’s flat EUR 3/item customs duty (from July 2026).

What was re-derived from scratch: the FCF bridge, all five valuation models, the Monte Carlo simulation, the Quality of Earnings chapter and the quarterly-variation chapter — because the new filing changes both the figures and two methodological assumptions from the prior report (detailed below).

Executive summary (1 page: thesis, estimated value, verdict)

The resumption was requested for a single verifiable reason, and that reason has been resolved. The falsifier left open on 08/04/2026 was “revenue keeps growing at a mid-to-high single digit rate, it doesn’t reverse” — undecidable at the time, missing a second consecutive quarter. The Q2 2026 release (6-K filed 08/25/2026, accession 0001104659-26-100534) closes it: Q1 2026 revenue … YoY, Q2 2026 … YoY, H1 2026 …. The falsifier holds. And, more important than that: the tracking signal that the prior report explicitly left as a thesis test — “if quarterly operating margin continues compressing below 18% for another 1-2 quarters, the thesis becomes hard to defend” — not only failed to trigger, it reversed. Operating margin went …% (Q2 2025) → 23.1% (Q3) → 21.1% (Q4) → 18.4% (Q1 2026) → 24.7% (Q2 2026). The four-quarter compression has stopped. All five falsifiers of the baseline thesis survive.

But the resumption also surfaced three things the prior report did not have, two of them the analyst’s own methodological errors. First: the Monte Carlo simulation in the delta pack was returning a median intrinsic value of USD 0.01/share and a MOS of … across all percentiles — an obviously absurd result, produced because mc_dcf.py had received oe = … and nd = -… in billions of USD, when the model requires millions in the reporting currency. A scale error of ~7,000x. Second: the prior report had not adjusted the share count for the ADS ratio. With sh in ordinary shares and px per ADS, every per-share value came out 4x lower than what’s comparable to the price. Third, and most substantial: the prior report used CFO − capex = RMB … bn as the valuation base and, separately, added the net cash of USD … bn — double counting, because of that RMB … bn of CFO, RMB 12.3 bn is interest received (the “Interest received” line in the cash flow statement, 20-F FY2025) on exactly the same pile of cash added separately. All three are corrected here.

The thesis, restated on the new figures. PDD trades at USD …/ADS, i.e. USD 121.2 bn diluted market cap. Of that, net cash is USD 67.3 bn (RMB 451.6 bn) — …%. So the market is paying ~USD 53.9 bn for a business that generated, over the trailing 12 months, RMB 450.8 bn revenue (… vs. calendar-year FY2025), RMB 98.6 bn operating income (21.9% margin) and RMB 111.9 bn operating cash flow — i.e. 4.6x after-tax operating income. At a trailing P/E of 8.89 and P/B of 1.74, the valuation remains compressed. Normalized owner earnings, cleaned of financial income (to avoid double-counting the cash pile) and with capex brought to the new regime: RMB 71.4 bn/year.

Estimated value. Five triangulated models give a margin-of-safety range between … (bear DCF, with 6%/year decline and a 25% haircut on cash for repatriation/PRC risk) and … (bull DCF), with the median at …. Monte Carlo on the central scenario (20,000 simulations, stated assumptions) gives a median of …, P10 …, P90 …, probability of undervaluation …% and probability of MOS above … of …. Terminal weight: 41.2% — below the fragility threshold.

Verdict: GRADUAL ACCUMULATION, small position — one notch above MONITOR, not two. The case for the upgrade is that the open question from August closed favorably, and all falsifiers hold. What stops me from going further are two things, both new and both from the current filing: (1) in Q2 2026 a line appears for “Other income/(loss), net” of −RMB 7,399 mn, versus +119 mn in Q2 2025, at a company that had reported +2,727, +3,120 and +2,953 mn in FY2025, FY2024 and FY2023 — an unmentioned loss in the release’s text, unexplained in the financial press and with no explanatory note (6-Ks have no notes), which by itself turns … operating-income growth into … net income; (2) property and equipment jumped from RMB 1,306 mn to RMB 4,752 mn in six months (×3.6), at a company that had spent RMB … mn on capex in all of FY2025 — a new investment regime, not disclosed as such, which is exactly the variable my owner-earnings base hangs on. Both resolve at the Q3 report on 11/18/2026.