BIDU · Baidu, Inc. ADS · Stock research

Current Thesis

• The date is no longer an estimate. Two sessions from this writing. Everything in the sum-of-parts case (Kunlunxin, Apollo Go, the dual-primary conversion) either gets a dated timeline on that call or stays a story.
• The ADR bleed flattened without reversing. Market cap was $37.10B on 2026-08-10 against $37.21B on 2026-07-24 — roughly unchanged over 17 days after the ~17% slide from the mid-$40B June range. The 2026-08-14 close of $103.67 sits -36.2% below the 52-week high of $162.52, with a 3-month return of -23.4% and RSI(14) at 45.7.
• What an investor is buying here is a discount-closing narrative, not an earnings-momentum one: a Cayman/VIE ADR trading at a fraction of the parts, where each 2026 event (dual-primary HK approval 2026-07-16, Kunlunxin's confidential HK filing ~2026-01-02, Apollo Go's London start 2026-07-28) is supposed to force the parts to be marked. The counter-evidence is three weeks old and simple — the July headlines moved the Hong Kong lines and left the US line where it was.
• The narrative is maturing. Fresh, dated headline flow continues (Apollo Go London 2026-07-28/29; Kunlunxin winning share in all three packages of China Mobile's AI general-computing procurement, reported August 2026), but the ADR has not translated any of it into price since 2026-07-16. Well known, still generating news, moderating flow.

Bullish and bearish views on Baidu, Inc. ADS

The model's bull view on Baidu, Inc. ADS (BIDU), in brief: Kunlunxin sum-of-parts, now with a revenue reference. The bear view: The July de-risking produced no durable ADR bid. Both cases follow in full.

Bull Case

• Kunlunxin sum-of-parts, now with a revenue reference. The chip unit filed confidentially in Hong Kong on ~2026-01-02 at a reported ~RMB21B (~$3.0B) valuation; by 2026-06-29 press reports put the IPO target near $50B and the ADR rose 7% that session. In August 2026 Kunlunxin was reported taking share in all three bid packages of China Mobile's centralized AI general-computing procurement at a "billion-yuan" order level, with China Southern Power Grid and BYD named as external customers. A $50B mark would exceed the entire consolidated $37.10B cap.
• Dual-primary HK listing, board-approved 2026-07-16, effective within 2026. Completion admits southbound Stock Connect flow and reduces the delisting tail that the Pentagon CMC designation of 2026-06-08/09 deepened.
• Apollo Go went international with Western dispatch partners. Road testing of the sixth-gen RT6 began in London on 2026-07-28 with Freenow (Lyft) and Uber, rides targeted from 2027 — on top of 22 cities, 17M+ cumulative orders and a 20-city completion target for Q4 2026.
• The last reported quarter grew where it matters. Q1 (2026-05-18): AI Cloud +33% YoY, accelerator-subscription revenue +128% YoY, adjusted EPADS $1.75 against $1.69 consensus.
• The Street's mark is far above the tape. Consensus price target $167.09 across 32 analysts (S&P Global data, page dated 2026-08-14) versus the $103.67 close; standing individual marks span Barclays $124 (2026-07-14), B of A $165 (2026-07-14) and JP Morgan $205 (2026-07-17).

Bear Case

• The July de-risking produced no durable ADR bid. Cap $37.21B (2026-07-24) → $37.10B (2026-08-10). Two of the strongest structural catalysts this name has had — HK dual-primary approval and Apple Intelligence China clearance, both 2026-07-15/16 — are three to four weeks old and the US line has not reclaimed the $110 shelf it lost in July.
• The Kunlunxin valuation gap is 16x wide and unresolved. ~RMB21B (~$3.0B) at the confidential filing versus the ~$50B figure reported 2026-06-29. Nothing public reconciles the two; a prospectus that prices nearer the filed mark removes most of the sum-of-parts arithmetic.
• Estimates came down while ratings stayed put. Three price-target cuts inside four days in July (Barclays, B of A, JP Morgan) with no rating changes attached. Barclays' $124 low mark sits ~20% above the current tape, meaning the market is trading through the most conservative published model.
• Apollo Go spends before it discloses. ~$1.2B of three-year fleet capex against a segment whose unit economics have not been broken out; London rides are guided to 2027, so the cash goes out well ahead of any revenue line.
• Geopolitics is a standing multiple cap. The 2026-06-08/09 CMC designation is a mandate constraint for some US institutions regardless of operating results, and escalation paths (Kunlun export restrictions, HFCAA action) sit outside management's control.

Setup & Price Structure

• Reference close $103.67 (2026-08-14). The $110 area that acted as a shelf through mid-July is now overhead; nothing in the tape since 2026-07-24 has tested it. Range behaviour over the last three weeks has been compression near $103–107 rather than a reclaim.
• RSI(14) 45.7 — mid-range. No oversold washout to mean-revert from, and no overbought extension to fade. A -23.4% three-month return with a neutral oscillator describes a drift, and drifts do not resolve themselves; the print does.
• Positioning observables, stated without a verdict: Appaloosa Management's 13F filed 2026-08-14 showed the Baidu holding up 87% to 1,295,000 shares — a snapshot as of 2026-06-30, already 45 days stale when disclosed. An options whale-alert screen flagged the name on 2026-07-22. Retail-facing coverage in the window skewed to low-information filler — the same "how much you would have made in 20 years" piece ran on 2026-07-24 and again on 2026-08-10 — which is the coverage mix a name gets between catalysts. No insider transactions appear in the recent filing record.
• The gradeable structural line is $100: it is the round number under the July–August compression and the level whose loss confirms no bid was created by the July headlines.

Catalyst Calendar (next 30 days)

No upcoming dated catalysts on file — the dated entries below have passed.

Elapsed catalysts

• 2026-08-18 (confirmed, pre-market; call 8:00 AM ET) — Q2 2026 results. First full quarter reported after the Apple Intelligence China clearance and the dual-primary approval. (passed 8d ago)
• 2026-08-18 (same call) — any dated effectiveness for the HK dual-primary conversion approved 2026-07-16. A named date is the mechanical step that changes the delisting discount; "within 2026" repeated is not. (passed 8d ago)
• 2026-08-18 (same call) — Apollo Go disclosure depth: whether the 20-city Q4 2026 target is reaffirmed and whether any segment-level economics accompany the ~$1.2B three-year fleet capex. (passed 8d ago)
• ~2026-09 (est.) — possible move from Kunlunxin's confidential Hong Kong filing (~2026-01-02) to a public prospectus or listing-hearing date. No date has been disclosed; the trigger is the filing itself appearing. (passed 236d ago)

What Would Change Our Mind

• The structure that has to hold is the $103–107 compression that formed after 2026-07-24. A weekly close below $100 forfeits it and grades the July catalyst pair — dual-primary approval and Apple Intelligence clearance — as having produced no lasting ADR bid.
• On the fundamentals, the specific number to check on 2026-08-18 is AI Cloud growth against the +33% YoY printed for Q1 on 2026-05-18. Deceleration there while online marketing keeps shrinking removes the offset that the whole pivot rests on.
• The catalyst-passes-and-nothing-happens path: the 08-18 call comes and goes with no named effectiveness date for the HK dual-primary listing and no Kunlunxin timeline. That converts a dated re-rating case into an open-ended one.
• Upside change-of-mind: a Kunlunxin public prospectus with a valuation range anywhere near the ~$50B reported 2026-06-29, or a reclaim and hold of the $110 shelf on the print, would argue the ADR is finally marking the parts.

Correlation Notes

• Trades as part of the China ADR complex — Alibaba and Tencent lead the beta, and the KWEB basket sets the discount regime. On 2026-07-15 Alibaba gained ~4% premarket on the Apple Intelligence/Qwen headline while Baidu's US line did not follow; that divergence is the specific relationship to watch on the next sector-wide headline.
• Kunlunxin's mark is set by the Hong Kong AI-silicon listing tape (Biren's HK$5.58B listing, reported alongside the Kunlunxin filing in January 2026) and by every incremental US export-control action pushing Chinese buyers to domestic accelerators — ByteDance was reported in talks with Baidu and Iluvatar CoreX on 2026-06-15.
• Apollo Go now carries read-across to Uber and Lyft, both named partners in the London program from 2026-07-28, and sits inside the same robotaxi narrative that prices Waymo and Tesla's autonomy claims.
• Reporting currency is RMB; USD ADS figures move with the CNY translation independent of operations.