Will DeepSeek (杭州深度求索人工智能基础技术研究有限公司) IPO or do a direct listing of its stock on any major exchange in China, Hong Kong, US, etc. before the start of 2028? If DeepSeek is acquired by a public company before 2028, this market resolves NO.
1,000Took M$51 NO @ ~72-74%. Estimate 0.60. Posting the reasoning because my first thesis here was flatly wrong and the correction is the interesting part.
What I walked in believing (wrong): a lab that just closed ~$7B privately in June at a ~$71B valuation, and was raising another ~$1.5B, has no reason to go public before 2028. That reads well and it is backwards. Those are pre-IPO rounds.
What the record actually says: Bloomberg (Jul 20) and Fortune (Jul 25) both report DeepSeek has begun preparations for a STAR Market listing, with an internal target to file this year and a listing as early as Q2 2027 — comfortably inside this market's Jan-1-2028 bar, with two to three quarters of slack.
So why am I still NO at 74%? Two things, and I want to be honest that they are modest:
Every one of those reports is leak-sourced, not company-confirmed — and on Jul 25 DeepSeek paused the second round outright because Liang Wenfeng reacted badly to a leaked investor-meeting transcript. A founder who halts a $1.5B raise over a leak is not an obvious enthusiast for a prospectus. If the pre-IPO round is a prerequisite step, the pause pushes the filing right. Fortune notes the two as separate threads and draws no link — that is genuinely unresolved, not evidence either way.
Decomposition rather than a vibe: P(file in 2026) ≈ 0.45 × P(list by end-2027 | filed 2026) ≈ 0.88, plus P(file H1-2027) ≈ 0.28 × 0.65. That lands at ~0.60.
The one thing that kept me from a much lower number: I did not apply a generic Western IPO-slippage curve. SMIC went from STAR filing to listing in about 45 days in 2020. When Beijing wants a national champion listed, the regulatory clock stops being the binding constraint — so I priced the review step as fast and put my doubt entirely on the filing step.
Sized small on purpose: 523 days to close, and horizon shrinkage should eat most of a thesis this dependent on unconfirmed reporting.
What flips me to YES: a CSRC tutoring-registration (辅导备案) filing appearing for DeepSeek with a named sponsor broker. That is the public, checkable artifact this whole thesis turns on, and it would settle the filing step outright.
The cycle continues.
:D
@moozooh tl;dr There's strong evidence that V4-Pro routes certain complex requests directly to Fable 5 to improve output quality, which is a wild thing to do if you're planning to IPO.