IPO and Capital Markets Path
A-shares, Hong Kong and US listings are not three options — they are three different sets of expectations. Choosing the wrong venue usually surfaces in the first round of regulator questions.
Choosing a listing venue means choosing which standard judges you
The same AI company is assessed very differently in Shanghai, Hong Kong and New York. A-shares examine continuing operations and regulatory discipline, with acute sensitivity to revenue recognition, related-party transactions and competing businesses. Hong Kong’s Chapter 18C opens a revenue exemption for specialist technology companies but demands a full case on R&D spend and commercialisation. US markets weigh growth narrative and disclosure adequacy, tolerate more variation in share structure, but scrutinise data and privacy disclosure in detail.
AI companies also face an additional line of review: data and algorithm compliance. Whether training data was lawfully sourced, whether the algorithm has been filed, whether generated content carries the required label, whether data export has been assessed — questions that did not exist in traditional IPOs are now fixed chapters in the filing package.
What we do
- Venue assessment and timetable: board selection, market comparison, financial and compliance gap analysis
- Joint-stock conversion and corporate history: conversion from limited liability, nominee shareholding clean-up, capital contribution defects, historical state or foreign ownership issues
- Compliance workstreams: data compliance, algorithm and LLM filing, AIGC labelling, IP ownership, employment and social insurance
- Legal sections of the filing: lawyers’ work report, legal opinion, prospectus chapters
- Regulator questions: reply letters, special explanations, on-site inspection support
- Offering and post-listing: strategic placement, ESOP normalisation, disclosure and equity incentives after listing
Typical situations
Situation 1: Nominee shareholding in the history, cleaned up at the last minute
For registration convenience, part of the founding team’s equity was held by nominees without agreements or capital evidence. At filing, each holding must be explained and restored, triggering tax payments and equity changes that add more than six months.
Our approach: start the clean-up 18 months before filing — restore each nominee arrangement, complete capital evidence, settle tax, obtain non-dispute confirmations, and build an unbroken chain of title.
Situation 2: Algorithm filing missing, treated as a compliance defect
The company operates a generative AI application in production without completing algorithm filing or LLM registration. Two consecutive rounds of questions ask whether the filing obligation was met and whether the omission affects continuing operations.
Our approach: complete filing and registration before submission and obtain registration numbers; explain the historical gap, assess penalty exposure, and where possible obtain a compliance confirmation from the competent authority.
Situation 3: Data provenance cannot be evidenced
Part of the training data was scraped with no authorisation records and no source register. The regulator asks for lawful sourcing evidence; the company cannot produce a complete chain and is required to commission an independent data compliance audit.
Our approach: build a source register and authorisation chain (procurement contracts, licences, scraping policy documentation), replace or retrain on high-risk sources, and assemble auditable evidence.
Process and deliverables
| Stage | Work | Deliverables |
|---|---|---|
| 1 | Venue assessment, gap diagnosis | Listing path report + remediation list |
| 2 | Joint-stock conversion, history clean-up | Conversion plan + chain-of-title file |
| 3 | Compliance remediation (data / algorithm / IP / employment) | Workstream reports + remediation evidence |
| 4 | Drafting legal sections of the filing | Lawyers’ work report + legal opinion |
| 5 | Question responses and inspection support | Reply letters + special explanations |
Frequently asked
A-shares or Hong Kong for an AI company? It depends on revenue mix and commercialisation stage. With scaled revenue and mainly onshore enterprise or government clients, STAR Market or ChiNext fits better. Pre-profit with heavy R&D suits Hong Kong 18C. Global-facing, USD-denominated revenue still points to the US.
When should listing compliance begin? 24 months before filing. Data remediation, algorithm filing, equity clean-up and capital contribution issues all take time; the later you start, the fewer options remain.
Is algorithm filing a precondition for listing? For providers of generative AI services it is a statutory obligation. Non-compliance is treated as a defect requiring remediation and disclosure before filing.
Do we still need counsel after listing? Yes. Continuous disclosure, equity incentives, refinancing, M&A and related-party approvals all occur post-listing; compliance cost does not fall after the bell.
Legal basis
- Securities Law; Measures for the Registration of Initial Public Offerings
- STAR Market rules (SSE) and ChiNext rules (SZSE)
- HKEX Listing Rules, Chapter 18C (Specialist Technology Companies)
- Interim Measures for the Administration of Generative AI Services; Provisions on the Administration of Algorithm-Generated Recommendation Services
- Measures for Labelling AI-Generated Synthetic Content
- Data Security Law; Personal Information Protection Law and export rules