CEO Yang’s View | Sora’s Exit Doesn’t Mean AI Video Lost: What the Market Is Really Weeding Out Is a False Prosperity of Burning Compute for Applause

On March 24, OpenAI shut down the Sora consumer app and related products. Multiple media reports point out that the main reasons were not that the technology wasn’t good enough, but high compute costs, a business model that was not solid enough, and OpenAI refocusing its resources on more core directions that are easier to monetize, such as coding, enterprise, robotics and world simulation.

On seeing this news, many people’s first reaction was: “See, AI video is just a bubble.”

But I don’t see it that way. What I see is not a myth shattering. What I see is the market finally starting to use true business logic to put on trial a product that once stood at the center of the spotlight.

The easiest mistake this world makes is to mistake a moment of amazement for a long-term victory. When something first comes out, everyone is stunned by it. The visuals are impressive. The effects are outrageous. The community is feverish. The user numbers are staggering. But the capital market has never looked at how sensational your first appearance was. The capital market reads a different language: is your retention stable enough? Do your unit economics work? Can your capital expenditure be recovered? Is your product a genuine need, or collective excitement?

This is not an emotional issue. It is a structural issue.

From the standpoint of a listed-company director, a product’s real value lies not in whether it creates buzz, but in whether it can clear three thresholds.

The first threshold is cost. The second is retention. The third is the business closed loop.

Sora’s exit was not because it wasn’t dazzling enough. Quite the opposite — it was because it was so dazzling that the whole world saw the central contradiction of this generation of AI video: the higher the visual quality, the heavier the compute and inference costs behind it; the more the product resembles the future, the more the financial statements resemble an alarm.

This is something I often say: a victory for technology does not mean a business has been established. Nor does it mean capital’s patience.

Judging from the international settings PSF has walked through in recent years, this is actually easy to understand. Stand in Japan and you feel a deep sense of order. The Japanese market will not embrace something fully just because it is new; it looks for stability, credibility and long-term operability. Stand in Singapore and you feel institutions and capital efficiency. Singapore doesn’t believe in stories; what it likes is the auditable, governable and scalable. Stand in Dubai and you see capital’s high sensitivity to future narratives, but however large the capital, it will keep betting only on narratives that can land. Stand in the Czech Republic and you see another problem: talent, infrastructure, national-level cooperation and industrial absorptive capacity — none can be missing.

So it has never been that “the technology is strong enough” and you win. It is whether this technology has met soil that can take it on. This is what Buddhism calls “cause, conditions and result.”

Many people simplify a product’s success into a strong “cause.” As if a strong model means the product will certainly succeed. That sees the world too thinly. Buddhism does not see it this way; Buddhism speaks of dependent origination.
Nothing arises from a single cause and a single result; everything arises from the coming together of many causes and many conditions.
What is Sora’s “cause”? It is OpenAI’s powerful model capability, a group of the world’s very best talent, the market’s high expectations for AI video and the inevitable trend of generative AI moving from text and images to audio and video. These causes are themselves very strong. No problem. But a strong “cause” does not mean the “result” will necessarily come, because in between you must look at the “conditions.”
What Sora lacked was not the cause. What it was truly short of was the conditions.
What are conditions? They are the circumstances that allow a product to go from a technology demo to a living commercial organism.
For example: can compute costs be borne, can legal and copyright risks be controlled, can content be produced continuously, are users willing to return, are enterprises willing to pay, is management willing to keep investing resources at this moment. If the cause is strong but the conditions are not fulfilled, the result still does not arise. If the cause is merely adequate but the conditions are very complete, the result may in fact ripen first. This is the core of the divergence between China and the US in AI video today.

Where is the US strong? In frontier models, basic research, world models, enterprise integration and high-end capability. OpenAI closing Sora this time is not exiting AI video but refocusing its firepower on higher-value, longer-term main battlefields, including more core products and world-simulation research.
Where is China strong? In another fulfillment of conditions. It looks not only at the model itself but at platforms, traffic, the creator ecosystem, business scenarios, content distribution, advertising demand and e-commerce conversion — this whole set of conditions all in place at once.
Take Kuaishou’s Kling as an example: official figures show that as of December 2025, Kling had served more than 60 million creators, generated more than 600 million videos, and worked with more than 30,000 enterprises, with monthly revenue in December 2025 exceeding US$20 million, corresponding to an annualized run rate of about US$240 million. The most important meaning of these numbers is not to boast, but that they prove one thing: AI video in China is no longer merely cool-looking; it has begun to form a real business closed loop.

This is the judgment I really want to state: Sora’s exit does not mean AI video has failed. Sora’s exit means the market is beginning to weed out a product form. Which product form? The kind that is good at producing gasps but can’t hold cash flow. The kind that looks like a revolution on social media but looks like a consumable on the financial statements. The kind that lives on buzz but has not truly embedded itself in users’ workflows or willingness to pay. That is the essence of the problem.

As a director, much of the time you cannot be held hostage by a “technological spectacle.” What you look at is not “can it be done” but “is it worth the company doing it now.” Especially when a company enters the capital market’s view and faces higher revenue pressure, valuation pressure and resource allocation, many seemingly romantic side projects ultimately have to return to the most practical questions: can it serve the overall strategy? Can it turn resources into a bigger moat? Is it the main battlefield or a display piece?
If it is not the main battlefield, even something once watched by all may be shut down. That cannot be defined as failure. It is called capital discipline, and cutting losses in time is sometimes the smartest thing. And capital discipline has never been coldness; it merely acknowledges that all phenomena in the world are impermanent.

Buddhism says: all conditioned things are impermanent. Hearing this, many people assume it is a pessimistic statement. But people who truly understand impermanence can see one thing more clearly: every boom is only a temporary gathering of causes and conditions. When causes and conditions are complete it arises quickly; when they change it disperses just as quickly. The problem is not whether it disperses. The problem is whether you understand why it arose and why it ended. Sora arose because at that time the market needed a symbol. It symbolized that AI video had really arrived. It symbolized the technological leap from text to images and images to audio-video. It symbolized that OpenAI could not only do language but also touch a higher-dimensional world of generation. But Sora’s ending was not accidental either. When a product’s reason for existing begins to lose to the company’s larger logic of resource allocation, when its cost, risk and payback period cannot compete with other more core strategic projects, it will naturally withdraw. That is the ending of conditions.
It is not that the cause has disappeared. It is that the conditions that supported that result have changed. So seen from the Dharma, you will not view this as simple success or failure. You will view it as a complete flow: there is its cause, it meets its conditions, it bears its result; after the result forms, causes and conditions change again, and the aspect of the result ceases. A person with real judgment does not become fanatical on seeing something arise or pessimistic on seeing it end, but sees the whole structure of causes and conditions behind it.

Returning to the future of AI video, I actually think now is when it truly begins. Because in the previous period the whole market was comparing who was most astonishing. Next, the market will compare who survives best. Who can bring costs down. Who can build the tools into the high-frequency scenarios of e-commerce, advertising, short dramas, games, animation, brand content, training and corporate marketing. Who can get users not to play occasionally but to open it every day. Who can get enterprises from “feeling it’s impressive” to “willing to set a budget.” Who can make the model’s capability exist not merely as a demo feature but as an SOP.
From this angle, China is indeed running very fast right now. It isn’t simply overtaking on models but has embedded AI video into real demand earlier. This is also why the commercialization figures of products like Kling deserve special attention. But this doesn’t mean the US has lost; the US has simply changed its way of playing. It is pulling the front line to a higher dimension: not only making tools for creators to play with, but betting on larger foundational capabilities, deeper enterprise integration and a longer link between world simulation and robotics. So the most likely future pattern is not one party dominating but two lines in parallel.
One line extends toward the high end, toward enterprises, toward world models and toward physical-world AI.
The other line extends toward mass creation, the content industry, advertising e-commerce and the deployment of high-frequency audio-video production. One line lays the foundation; one line grabs the market. One line decides the future’s technological high ground; one line decides today’s user habits and the entry point of cash flow. These two lines do not conflict and may even eventually merge.

And for people like us who truly work in industry, capital and cross-border positioning, what matters most is not taking sides but reading the rhythm.

Japan taught me to look at order.
Singapore taught me to look at institutions.
Dubai taught me to look at how capital chases the future.
The Czech Republic let me see the importance of infrastructure and national cooperation.

And Taiwan makes me clearer still: however strong the technology supply chain, if the capability isn’t turned into your own industrial structure and business system, you easily end up stuck inside someone else’s narrative.

So my real conclusion on AI video is: with Sora shut down, what died is not AI video. What died is a product illusion that survives only on buzz and cannot cross business reality.

AI video has not ended. It has simply moved from an exhibition match to a knockout round. From the era of astonishment to the era of judgment. From “can it be done” to “can it live.” And looking deeper, this is not even only a question for AI video. It is a proposition the whole AI industry must face together. Those who survive in the future will not be the best at performing, but those who best understand conditions, best allocate resources, best pass through impermanence and best turn technology into systems.

Those who understand the arising of conditions will not lose themselves in a moment of prosperity.
Those who understand the ending of conditions will not become pessimistic at a moment’s withdrawal.
Because they know the world has never been linear. Industry is not, and the capital market even less so.

A true master doesn’t look at whether one product continues; he looks at what causes and conditions the next result is quietly taking shape in.