Contrarian Corpus
short seller research note initial thesis
2020-05-18 · 25 pages

GSX Techedu Inc. GSX

GSX Techedu is a near-total fraud: at least 73% — likely 80%+ — of paid K-12 users are bots, and Chairman Chen's $319M pledged stock heightens crash risk.

N 4 Narrative
V 2 Visual
C 2 Craft
Source URL unavailable

Thesis

Muddy Waters argues GSX Techedu is a 'near-total fraud' built on industrial-scale user fabrication. Using GSX's own attendance and sign-on records — 463,217 logins from 54,065 unique users across more than 200 paid K-12 classes — analysts identified three bot patterns ('Precise Joiners', 'GSX IP Joiners', 'Burst Joiners') showing that at least 73.2% of users are bots, rising to ~80.8% when 'Early Joiners' are included and ~90% under less company-favorable assumptions. A former GSX manager corroborated the analysis, describing in-house 'group robot' rooms that control 10,000+ devices and outside bot-network vendors paid 2-5% commissions to round-trip fake purchases. Because revenue depends on these bots, GSX is 'an almost completely empty box'. Chairman Larry Xiangdong Chen has pledged 6 million Class B shares (~$319 million) at Credit Suisse, creating margin-call risk that could accelerate the collapse.

SCQA

Situation

GSX Techedu is a NYSE-listed Chinese online K-12 tutoring platform with an $8.5bn market cap, a 24.9% float, and a post-IPO record of explosive reported user growth and unusual profitability for the sector.

Complication

Forensic analysis of 463,217 sign-on records and a former manager's testimony show that 73-90% of paid users are bots controlled by GSX's own 'group robot' rooms and outside vendors round-tripping cash through ad placements.

Resolution

Muddy Waters discloses a short position and publishes the dataset, methodology and a former-manager interview, urging the market to treat GSX as a near-total fraud and discount its reported revenue accordingly.

Reward

If revenue tied to fake users is fraudulent, GSX is 'an almost completely empty box' implying near-total loss for long holders, with Chairman Chen's $319M margin-pledged stock amplifying forced-selling downside.

The three reasons

  1. 1

    At least 73% of GSX's paid K-12 users are bots, likely 80%+

  2. 2

    Former GSX manager confirms 'group robot' rooms controlling 10,000+ devices

  3. 3

    Chairman Chen's $319M pledged stock at Credit Suisse adds margin-call crash risk

Primary demands

  • Treat GSX as a near-total fraud and exit long positions
  • Recognize that reported revenue is largely fabricated via bot users

KPIs cited

Bot prevalence (high confidence)
73.2% of 54,065 unique users in 200+ paid K-12 classes flagged as bots
Bot prevalence (likely)
~80.8% including 'Early Joiners'; ~90% under less company-favorable assumptions
Precise Joiner share
52.8% of users are Precise Joiners or linked to them; 5,742 unique PJs (10.6%)
Sample size
463,217 sign-on records across 54,065 unique users in over 200 paid K-12 classes (Jan-Mar 2020)
User type distribution
29 teachers (Type 1), 371 tutors (Type 2), 53,694 students (Type 0)
Chairman Chen share pledge
6,000,000 Class B shares (~9,000,000 ADSs) pledged 30 March 2020 at Credit Suisse, ~$319M market value
Bot-vendor commission
Outside bot-network operators paid ~2-5% of would-be revenue to round-trip fake purchases
Stock metrics at publication
Price $35.43, market cap $8.5bn, float 24.9%, 30-day ADV $194.6M

Pattern membership

Where this document fits across the library's 12 rhetorical / structural patterns.

Notable slides (5)

Notes

Classic Muddy Waters Word-document-style short report (not a slide deck). Cover page (p.2) opens with 'Rise of the Machines' headline and a bot-prevalence pie chart, followed by SCQA-style bullet thesis on pp.2-3. Distinctive evidentiary moves: (1) forensic analysis of GSX's own data files (not scraped), (2) bilingual transcript of a former-manager interview describing 'group robot' rooms (pp.11-12), and (3) a BVI registry filing screenshot evidencing Chen's $319M share pledge to Credit Suisse (p.15). Quotes Chen's pre-publication taunt ('there is a high probability that I think Muddy Waters will not be so stupid') and explicitly 'calls the bluff'. No price target — fraud thesis implies ~100% downside. Appendices 1 and 2 detail the data-extraction methodology (Chrome DevTools, HTTP proxy on iPhone) and teacher/tutor/student type identification.