Sora vs Kling: The Real Competition is ARR, Not Technology

There’s an interesting phenomenon in the AI video generation field.

Everyone is discussing “whose technology is better” — Sora’s physics simulation, Kling’s motion coherence, Runway’s artistic style. But there’s a more important metric that few talk about: ARR (Annual Recurring Revenue).

Kling’s ARR has reached $300 million. Sora? After six months, it still hasn’t found a viable commercialization path and was eventually shut down by OpenAI.

This contrast is thought-provoking.

From a pure technology perspective, Sora is undoubtedly strong. Its physics simulation capabilities, long video coherence, and detail handling are all industry-leading. But technology leadership doesn’t equal business success.

Kling’s winning formula is simple: fast iteration, affordable pricing, scenario-focused.

They didn’t pursue “perfect physics simulation” but focused on scenarios users actually need: e-commerce product videos, short video special effects, social media content. These scenarios don’t require Hollywood-level quality, but they do need generation speed and cost-effectiveness.

Kling’s pricing is about 1/10th of Sora’s. For most users, “good enough” quality at 1/10th the price is an easy choice.

More importantly, Kling found a flywheel: more users generate more data, more data improves the model, better models attract more users. This is the essence of internet products, not the “high-end boutique” route.

Sora’s failure gives us a lesson: in the AI era, the “best technology” doesn’t always win. What wins is the technology that can find product-market fit.

Of course, this doesn’t mean technology is unimportant. Kling’s technology is also strong; it’s just not “Sora-level” strong. But in business, “good enough” often beats “the best.”

Another noteworthy phenomenon: video generation is a track where Chinese companies have significant advantages. Besides Kling, there are strong players like PixVerse and Alibaba’s Wan series. In contrast, besides Sora, US companies haven’t produced many competitive products.

This is the opposite of the large model field. In LLMs, the US leads; in AI video, China leads.

Why? I think it has to do with application scenarios. China has a massive short video user base and a developed e-commerce ecosystem; demand for video generation is more刚性. The US market is relatively smaller, making it harder for video generation tools to find sustainable business models.

Anyway, Kling’s success proves one thing: AI products can make money. The key is finding the right scenarios and pricing strategies, not blindly pursuing technological perfection.

What do you think? Is technology more important or commercialization more important for AI video generation?