Kling 4.0 goes live with 30-second native video at 10-bit HDR
Kuaishou's Kling AI has spent two years becoming the video model people actually finish projects with. Its new flagship, Kling 4.0, went fully live on October 15 — and this round is less about prettier frames than about the two things that make a generated shot usable in a real edit: length and control.
Kling 4.0 generates up to 30 seconds in a single native pass, outputs 10-bit HDR at 4K, 1080p and 21:9, and takes up to 15 reference assets in one task. That budget splits across ten images, five video clips and seven subjects, with up to ten keyframes for pinning exact moments, prompt input raised to 8,000 tokens, and repeated extension stretching a shot to two minutes. Kling's stated focus is the case where AI video usually falls apart: fast action, complex camera moves, and keeping a character consistent from one cut to the next. The company also added MCP and CLI hooks so agents can call the model directly, with API access for enterprise customers coming shortly.
The commercial logic is easy to see. Kuaishou's Q2 2026 results put Kling revenue above RMB 850 million for the quarter, up more than 200% year over year — roughly a $500 million annual run rate. The 100 million users, 224 countries and 50,000 enterprise customers Kling cites at launch are the company's own figures, first circulated in June and not independently audited; the revenue number is the one with a filing behind it. Worth noting: the launch trail runs through Chinese outlets and Kling's own site — 界面新闻, AIBase, 快科技 — and the English tech press has largely not picked it up yet, which is unusual for a model this widely used. One third-party aggregator still lists 4K and HDR as "coming soon," the only discrepancy in an otherwise consistent spec sheet.
Taiwan's government plans to build industrial parks in the United States costing $2 billion to $4 billion each, starting with a Texas complex priced at up to $4 billion. Economics Minister Gong Ming-hsin told Bloomberg the parks exist to give smaller TSMC and Foxconn suppliers — chip-packaging substrate and thermal-module makers — a home in the same country as their Nvidia and AMD customers, rather than shipping parts across the Pacific. Funding is expected to come from industry through TEEMA, the manufacturers' association chaired by Foxconn's Young Liu, with the government supplying credit guarantees, help with visas and land negotiations; Taiwan's Economic Ministry confirmed the same day that a single park runs $2–4 billion and is industry-funded.
Two caveats matter. This is a ministerial plan, not a program — no budget line, no construction date, and the Texas land acquisition rests on unnamed sources the ministry declined to confirm. And it is separate from TSMC's own capex: chairman C.C. Wei said on October 15 that TSMC is evaluating a Texas site of its own. A government-financed supplier park is a different beast from a fab, but both point the same direction — the AI supply chain's centre of gravity is being pulled toward US soil, one government subsidy at a time.
What to watch: whether TEEMA capitalizes the vehicle before year end, and whether the "more than one park" promise survives the first one's cost.
Does the US need Taiwan to build its supply chain on American soil — or should the chips come to Taiwan's ecosystem instead? Tell us in the comments.
