Top of the slop? The Sora slide could trigger an AI avalanche

OpenAI has unexpectedly shut down Sora, its generative AI video model. Despite a surge of initial popularity, the tech was dogged by controversy and copyright issues and proved prohibitively expensive.

In a significant blow to OpenAI, a $1 billion December 2025 deal to allow Sora users to generate AI videos featuring Disney characters was canceled at the same time Sora was jettisoned.

The company gave no reason for the shutdown of Sora, although analysts point to the dismal economics. With AI in the midst of what has all the appearances of a bubble, many wonder if the downfall of Sora could be the beginning of the end, especially as the war on Iran pressures supply chains and drives energy costs higher.

What was Sora?

Sora was an AI-powered text-to-video platform developed by OpenAI that was able to develop cinematic-quality high-definition videos from written prompts. Unveiled in 2024 and launched as a standalone app in September 2025, it quickly soared to the top spot on Apple’s US App Store.

The clips it produced were by all objective measures stunning. Also, at a time when most AI video tools were producing four-second clips that barely held together, Sora could generate up to 60 seconds of visually coherent footage.

The app allowed for the insertion of real people into AI-generated videos through a feature known as cameos. This feature, however, courted controversy from the beginning. In addition to skirting or outright violating copyright laws, many users exploited the tool to create absurd, and in many cases offensive, videos. One video, for example, featured Adolph Hitler arguing with Michael Jackson about who created the moonwalk.

The emergence of Sora (and especially the new and improved Sora 2) sent the creative industries in and around Hollywood into a panic. Suddenly, entire performances could be summoned at the push of a button without having to pay or even credit the people involved. Ultimately, however, this type of content, while popular with many users, struggled to find a viable economic niche, and many analysts argued that once the novelty wore off, there was little to sustain the momentum.

A media outlet called 404 Media was excoriating: “The complete and utter failure of both Sora and Disney’s dalliance with AI garbage suggests AI slop is indeed not the future of Hollywood. Disney did not even get to the point [where] it allowed people to build anything with Disney characters before pulling the plug on the whole endeavor and its investment.”

What do analysts believe is the real reason Sora was axed?

The decision appears to have been driven by financial considerations as OpenAI looks to streamline operations in preparation for an initial public offering as early as this year.

The platform was extraordinarily expensive. In November, one analyst suggested that it cost OpenAI $1.30 to generate a single ten-second video. Based on the 11.3 million daily videos that he estimated Sora produced, the analyst said this would cost the company about $15 million every day.

Just a few weeks prior to that, Sora head Bill Peebles admitted that the platform’s economics were “completely unsustainable.”

According to Business Insider, there was another problem that is proving to be an intractable constraint across the AI industry: Computing power. Video generation is the most energy-intensive form of AI currently in use.

“Given the frantic search for more compute across the industry, OpenAI is prioritizing its greatest growth engine – ChatGPT,” according to Navica CEO Bernard Golden.

As demand for AI compute surges, supply isn’t keeping up, especially as building new data centers has become harder due to increasingly vocal local opposition, insufficient grid capacity, and shortages of key components such as memory chips.

How frothy is the AI bubble?

That AI has all the hallmarks of a bubble is a fact even admitted in Silicon Valley. AI stocks are trading at extreme multiples even as many of the companies don’t turn a profit. OpenAI is currently worth more than Toyota, Coca-Cola, and Disney combined.

Time Magazine characterized the fundamental problem as being a “mismatch between the trillions being invested in the infrastructure to develop AI and the billions people and companies are spending to use AI.” Just this year alone, four tech companies – Amazon, Alphabet, Meta and Microsoft – plan to invest $670 billion on AI infrastructure. The data center buildout is bigger than the railroad expansion of the 1850s.

High leverage levels have been an ingredient in pretty much every major bubble in the past and AI is no exception. A particularly debt-heavy aspect of the AI story is the construction of data centers. Companies such as CoreWeave have taken on staggering levels of leverage to finance their infrastructure, betting that AI demand will grow fast enough to service that debt.

Then there’s the problem of circular investments. A good chunk of the AI revenue is being generated by AI companies selling to other AI companies, which creates a significant churn but doesn’t represent much in the way of organic growth. In fact, relatively little revenue is generated from the actual end users of AI. This also means the risks are highly concentrated if anything goes south.

Why might Sora be the beginning of the end?

The basic problem is that OpenAI has committed to a staggering $1.15 trillion over the next five years, whereas it made only made $13 billion in 2025. Of the roughly 800-900 million users of ChatGPT, only 5% are actually paying for it. OpenAI doesn’t have a whole lot of other options for generating significant revenue. Monetization of actual use has become a critical issue.

OpenAI needs to find a way to boost revenue and Sora seemed to be a promising path. The deal with Disney seemed like a foot in the door in Hollywood. But Disney has bailed and other companies could do so as well, especially as the economic reality of AI – especially in light of compute shortages – starts to catch up with the hype.

This also comes as the war on Iran is revealing fragilities in the AI supply chain, not least because the East Asian nations who dominate semiconductor production have been hit by major energy shocks.

If OpenAI finds itself in trouble, the whole AI industry would almost certainly follow. This could be the moment of truth for an industry that is almost single-handedly propping up the US stock market.

The Nasdaq is already in correction territory. The index is in fact approaching the so-called Death Cross as its 50-day moving average drops toward the 200-day moving average. This well regarded indicator suggests a loss of momentum and increased selling pressure.

Panic selling is unpredictable but history shows that once it starts, it takes on a life of its own.