Published: August 5, 2026 | Reading Time: ~7 minutes | Channel: techminute
Here's a sentence I didn't expect to write in 2026: the most interesting AI funding story of the moment isn't about a chatbot. It's about a company that turns your text prompts into video — and it just got $2.8 billion of China's best and brightest's money behind it.
Kuaishou's AI video arm, Kling AI, has raised over 19 billion yuan ($2.80 billion) from a who's-who of Asian tech and capital — Alibaba, Tencent, Baidu, and Abu Dhabi's BlueFive Capital — at a $15 billion pre-money valuation. For a division spun out of a short-video app most Western readers have never opened, that's a statement of intent so loud it should come with a volume warning.
And the timing? Perfectly brutal. Because halfway across the world, OpenAI just pulled the plug on Sora, the product that was supposed to own this entire category. Let's talk about what the Kling raise really means — and why the smartest money in Asia just bet on a video model when everyone's still freaking out about text models.
Let's start with the money, because the structure tells you how seriously everyone involved is taking this.
The round is $2.80 billion (over 19 billion yuan), and it's capped so Kling can bring in one more investor within the next two months (total cap: 20.45 billion yuan). After this injection, Kuaishou's stake in Kling AI dilutes from a controlling 100% down to about 68% — still in charge, but explicitly inviting the rest of China's tech aristocracy to the table.
That roster matters. It's rare for Alibaba, Tencent and Baidu to back the same horse — these three are usually trying to kick each other's buckets over. The fact they all piled in says less about Kling specifically and more about a shared, panicked realization: whoever owns AI video owns a giant slice of the next decade of content.
The revenue story backs up the excitement. Reuters reports Kling generated 650 million yuan in the March quarter — more than quadrupling year over year. Independent trackers put Kling's annualized recurring revenue at a shade under $500 million after surging from around $240 million in just three months — which puts this $15 billion pre-money at roughly a 30x revenue multiple, right in line with the fastest-growing AI software businesses on earth. Citi analysts called the investor roster "impressive" and flagged Kling's upcoming model upgrade as the thing everyone's watching.
You can't understand the Kling check without understanding what happened to its most famous competitor: OpenAI's Sora went from the most hyped AI product of 2024 to the most expensive failure of 2026.
The autopsy numbers are a case study in why wholesale hype is a terrible business plan. By March 2026, OpenAI had shut down Sora's app trajectory after the product was hemorrhaging an estimated $15 million per day in inference costs against roughly $2.1 million in total lifetime revenue. Each 10-second clip reportedly cost OpenAI around $1.30 to produce — compute so expensive that users simply weren't willing to cover it. Downloads cratered ~67% from their November 2025 peak, active users slid below 500,000, and a reported $1 billion Disney partnership — which would have unlocked 200+ Marvel, Pixar, and Star Wars characters — collapsed. The consumer app went dark on April 26, 2026, and the API is set to sunset on September 24, 2026.
OpenAI's pivot makes strategic sense (enterprise productivity ahead of a potential IPO), but the lesson is brutal and universal: an AI video model that burns more compute per frame than the market will pay for isn't a product — it's a glorified research demo with a subscription page.
That's precisely the gap Kling walked into. While Sora was bleeding out on photorealistic physics it couldn't monetize, Kling built a model that's good enough to ship at scale — and priced like it's actually trying to win.
So what does $2.8 billion buy you in a market where the incumbent just died of its own economics? Let's get into the tech, because this is where it stops being a funding story and starts being a war.
On the Artificial Analysis Video Arena — which ranks models by blind human preference across a large leaderboard — Kling 3.0's 1080p Pro tier sits at roughly Elo 1,104, the highest among models the public can actually buy, edging out Google's Veo 3.1 at around Elo 1,094. That makes Kling the de facto value champion of the post-Sora world — and it's not even close on price.
| Model | ~Per-second cost | Notes |
|---|---|---|
| Kling 3.0 | Best value, Elo ~1104 | |
| Google Veo 3.1 Lite | ~$0.05/s (720p) | Cheapest entry today |
| Google Veo 3 standard | ~$0.35–0.50/s | Cinematic tier |
| OpenAI Sora 2 Pro | ~$0.75/s | Priciest — and shutting down Sep 24 |
Independent analysis pegs Kling 3.0 as 65% cheaper than Sora was and 44% cheaper than Runway, while delivering output that holds up for real production work. One hands-on June 2026 comparison scored Kling's Pro tier at 91/100, ahead of Veo 3.1 (63/100) and Sora 2 Pro (59/100).
The features justify the "Omni" branding. Kling 3.0 generates multi-shot sequences of 2–6 scenes with automatic shot transitions and camera logic; clips run 3–15 seconds with native synchronized audio including character dialogue, SFX, and ambient noise with lip-sync; and it maintains character lock across scenes plus native text rendering. In other words, it's not just making pretty loops — it's trying to be a one-pass storytelling engine.
The secret sauce, as independent analyses note, is Kuaishou's decade of short-video data and video-compression expertise. That proprietary dataset is nearly impossible for a pure-play startup to replicate — which is exactly why Alibaba, Tencent and Baidu are all willing to compete with each other just to hold a piece of the pipeline.
Here's what I keep circling back to: the whole AI world spent 2025 arguing about which text model rules — GPT vs Claude vs Gemini vs the open-weight crowd. But the funding market is quietly telling you the next frontier is audiovisual — and that the economics there are unforgiving in a way text models never had to be.
Sora proved that raw quality without a viable cost structure is a tombstone, not a foundation. Kling's thesis is the opposite: build a model good enough for the 99% of use cases — ads, social, product demos, training videos — price it so the math works, and let the market come to you. Google's Veo is the enterprise-safe alternative with better photorealism and ecosystem integration, but at a premium price. Runway owns the quality/motion-control niche for professionals. Pika owns the fun, viral-creation lane.
Nobody owns all of it. But Kling just bought the biggest war chest in the category to try.
Also worth your attention: this is a Chinese company raising from Chinese giants to take on the global AI video market — the same week Washington was reportedly drafting restrictions on Chinese data-center hardware. The U.S.-China AI competition isn't just about chips anymore. It's moving downstream, into the content layer. Whoever wins AI video gets to define how a huge share of the world's moving images get made — and on whose infrastructure and whose terms.
I'm genuinely impressed by the Kling story, so let me be equally honest about the risks:
Sora proved that the AI video category has massive demand but unforgiving economics. Kling AI just placed the single biggest bet that a value-first, mass-market approach is the way to actually win it — with $2.8 billion and the three most powerful companies in Chinese tech backing that bet. The $15 billion valuation is the market's answer to a simple question: can you build AI video that ships at scale before your compute burns the whole thing down? Kling is betting an awful lot that the answer is yes. Honestly? I think it's the smartest bet being placed in generative AI right now.
The cameras are rolling. The money is on the table. And for the first time since Sora's spectacular collapse, the person holding the director's chair in this category might not be in Silicon Valley at all.
All claims verified against wire-tier (Reuters) and independent analysis (Silver/Bronze) sources; each URL was scraped and confirmed accessible. Funding, valuation, and revenue figures sourced from Reuters wire reporting and cross-confirmed across syndications. Benchmark Elo, per-second pricing, and Sora economics from independent analyses and disclosed as such. Last verified: August 5, 2026.
