--- title: "MiniMax books 60.8% of revenue offshore; Kling's split stays a secret" date: 2026-11-02 category: Generative Media site: NeuroAI canonical: https://neuroai.site/a/na-aigc-video-models-offshore-revenue language: en --- # MiniMax books 60.8% of revenue offshore; Kling's split stays a secret > Chinese video generators are selling to the world first: MiniMax's interim filing puts 60.8% of first-half revenue outside mainland China, while Kuaishou reports fast Kling growth without ever breaking out geography. A subscription charge in Los Angeles. A credit top-up in Jakarta. A Seoul ad studio burning API credits overnight. The models behind all three were trained in Beijing and Shanghai — and that mismatch between where generative video is built and where it is paid for is now the most interesting line in the public filings. ## Key takeaways - MiniMax's Hong Kong interim filing shows revenue of US$116.6 million for the first half of 2026, up 283.1% year on year, with US$70.8 million — 60.8% — booked outside mainland China. - Kuaishou's Kling (可灵) reported revenue of over 850 million yuan (≈US$120 million) in the second quarter, up more than 200%, but the company does not publish where that money comes from. - For both companies, the faster-growing line is not the consumer app but the developer and enterprise channel, where API calls and token plans are billed in dollars. - Overseas revenue is also the more fragile revenue: it carries currency, platform-policy and regulatory exposure that domestic billings do not. ## The only geography split anyone has published MiniMax — the company behind Hailuo AI (海螺AI) and the Talkie character app, listed in Hong Kong — filed interim results for the six months to 30 June 2026. The numbers are unusually clean: - Total revenue: US$116.6 million, up 283.1% from US$30.4 million a year earlier, and already above the US$79.0 million it booked for the whole of 2025. - Mainland China: US$45.7 million, or 39.2% of revenue. Other regions: US$70.8 million, or 60.8%. - Both sides grew. Mainland revenue rose roughly fivefold; overseas revenue rose about threefold. The overseas share fell from 71.8% simply because home-market billings grew faster. - AI-native products — Hailuo AI, Talkie and siblings — brought in US$42.6 million, up 100.9%. - The open platform and enterprise AI services brought in US$73.9 million, up 703.1%, and now account for 63.4% of the company, versus 30.3% a year earlier. That last line matters more than the apps. A video tool sold to consumers is a subscription business with churn. A model sold through an API is a metered business, and metering scales with someone else's workflow. MiniMax says its products and services reach users in more than 230 countries and regions, and its CEO Yan Junjie has framed the strategy as pushing intelligence to more people at lower cost rather than chasing model benchmarks. The profitability picture is still unglamorous: gross margin was 17.9% for the half, and adjusted net loss widened to US$293.0 million. Cheap inference buys volume; it does not yet buy profit. ## Kling's growth, minus the geography Kuaishou tells a similar story from the other direction. In its second-quarter results, Kling (可灵) posted revenue of over 850 million yuan (≈US$120 million), up more than 200% year on year, after a first quarter above 650 million yuan (≈US$92 million) that grew over 300%. By June, Kling had passed 100 million global users across 224 countries and regions and counted close to 50,000 enterprise clients — both up about 67% from the end of 2025. The product roadmap has followed the money: native 4K output aimed at film and advertising professionals, a Turbo tier tuned for speed and audio sync, and MCP and CLI interfaces so that agent systems can batch-produce clips without a human clicking through a web UI. In July 2026 Kling raised close to US$3 billion at a valuation of about US$18 billion, with Kuaishou retaining control. What Kuaishou does not publish is Kling's revenue split by market. Group-level overseas revenue was small in the quarter — about 1.18 billion yuan against 35.5 billion yuan total, and that figure is mostly the Kwai app, not the model business. So anyone claiming a precise overseas share for Kling is estimating, not reporting. ## Why the foreign buyer pays first Three mechanics explain the pattern, and none of them require the model to be better abroad. **Price anchoring.** In China, users expect creative tools to be bundled into a free super-app and monetised later. Outside China, a monthly subscription for a production tool reads as a normal business expense. **Who signs the invoice.** Enterprise and developer revenue is the fastest-growing line at MiniMax and the explicit target of Kling's 50,000 corporate accounts. Agencies, ad studios and overseas short-drama operators pay per render, in dollars. **Distribution physics.** A Chinese model company's own apps face home-market competition it cannot avoid; abroad, app stores and API marketplaces are comparatively open channels. According to Li An, chief scientist at BrainNet (脑机网), the split is a signal about willingness to pay rather than about model quality: overseas buyers treat a generation tool as a line item with a budget, while domestic users treat it as a feature that should arrive for free inside something else. The risks sit on the same side of the ledger. A business with 60.8% of revenue offshore is exposed to platform-policy shifts, payment-friction, data-transfer rules and the possibility that a large foreign competitor prices inference below cost. MiniMax's own filings list regulatory and geopolitical uncertainty as a standing risk. Compute cost is the other half: gross margin of 17.9% means each additional render still costs a great deal to serve. ## Honest limitations This piece leans on two filings — MiniMax's Hong Kong interim announcement and Kuaishou's quarterly results — plus the companies' own product statements. Anything about Kling's geographic mix, its margins, or its overseas paying-user count is not disclosed and has been left out rather than estimated. Third-party app-store rankings and revenue league tables circulate widely but trace back to trackers whose methodology is not public, so they are not used here. Figures labelled "over" are company-rounded floors, not exact amounts. Currency conversions use an approximate rate and will drift. Finally, both companies are in a phase where revenue growth and loss growth move together; none of the numbers above say anything about durable profitability. ## What readers can do now - If you buy generation credits for a team, compare the API price per finished second against your actual re-render rate — the headline per-clip price is rarely the number you pay, because discarded takes are billed too. - If you are evaluating a vendor whose revenue is mostly offshore, ask where inference is served and what happens to your account if a market becomes unavailable; put the answer in the contract. - Read the filing, not the launch post: for any video model you depend on, find the parent company's latest results and check whether revenue is growing faster than gross margin. That gap tells you how long the current pricing lasts. --- Published by NeuroAI (https://neuroai.site/) — https://neuroai.site/a/na-aigc-video-models-offshore-revenue Free to quote with attribution and a link to the original.