ChatGPT is the Blockbuster of AI. You're the customer with the loyalty card.

Posted on: 23 January 2026

There's a category of professionals who've spent the past two years building entire personal brands around the ability to write prompts for ChatGPT. Courses, webinars, ebooks, consultancies. "Become an AI expert", "Multiply your productivity", "The perfect prompt for every need". Marketable skills, they said. The future of work, they said.

OpenAI could run out of cash by mid-2027.

This isn't some conspiracy blogger talking. It's Sebastian Mallaby, economist at the Council on Foreign Relations, writing in the New York Times last week. The numbers are public, verifiable, and tell a story that nobody in the digital marketing circus seems willing to hear. Too busy selling the next course on how to use a tool that might not exist in eighteen months.

Eight billion dollars burned in 2025. Fourteen billion projected for 2026. Forty billion by 2028 if the trajectory doesn't change. Sam Altman raised $40 billion in his last funding round, the largest in history for a private company, but has committed $1.4 trillion to data centres. The maths doesn't work. It won't work.

Ninety-five percent of ChatGPT's 800 million users don't pay a penny. Read that again: ninety-five percent. The introduction of advertising on the platform, announced with the usual garnish of visionary storytelling, was universally read as a panic signal. A venture capitalist told The Economist with perfect brevity: "This is the WeWork story on steroids." For those who don't remember, WeWork was the company that would revolutionise work, valued at $47 billion, bankrupt in 2023 after burning through investor money chasing a vision that didn't stand up.

But the truly interesting part isn't OpenAI's potential collapse. It's what's happening on the other side of the world while you're perfecting your prompts.

In Hangzhou, a Chinese city most British marketers couldn't locate on a map, a startup called DeepSeek has just published in Nature, the most prestigious scientific journal on the planet, a paper demonstrating how its R1 model, competitive with GPT-4, was developed with an infinitesimal fraction of the resources. They didn't copy. They didn't "distil" from Western models, as detractors claimed. They innovated. Under constraint.

The American chip ban, designed to slow Chinese AI, produced the opposite effect. It forced DeepSeek to develop more efficient architectures, models requiring less computing power, approaches that American labs, drowning in unlimited budgets, had no incentive to explore. Why optimise when you can simply buy more servers?

The penetration numbers tell a story nobody in the West wants to hear. According to a Microsoft report, DeepSeek holds 89% of the AI market in China. 56% in Belarus. 49% in Cuba. 43% in Russia. It dominates in Syria, Iran, Ethiopia, Zimbabwe, Uganda, Niger. Not for ideology, for accessibility. The models are open source, costs are a fraction, and they run on hardware the West considers obsolete. While you're paying £16 a month for ChatGPT Plus feeling cutting-edge, half a billion people are using something comparable for free.

The pattern is as old as innovation itself. Those operating under constraints develop solutions that those operating in abundance don't even look for. Toyota in the 1950s, with production lines compressed by the limited space of Japanese factories, invented the just-in-time system that would tear apart the American automotive industry. Detroit's Big Three, with their sprawling plants, didn't need to optimise. They could simply expand. Until they couldn't, and by then it was too late.

OpenAI today is exactly in the position of General Motors in 1970. Undisputed leader, apparently solid margins, total blindness to the disruption coming from below. With one crucial difference: GM had a business model that generated profits. OpenAI doesn't.

Google and Meta can afford to burn money on AI for a decade without feeling it. Google with £245 billion in annual revenue from other sources. Meta with £107 billion. They have the luxury of experimenting, failing, starting again. OpenAI depends entirely on venture funding and the Microsoft partnership. When the cash runs out, there are two options: get absorbed by a giant, or disappear. Tertium non datur.

And then there's Anthropic, the case nobody's telling. Founded in 2021 by former OpenAI research executives, it burns $3 billion a year against OpenAI's $9 billion. It expects to break even in 2028, two years before its competitor. The burn rate will drop to 9% of revenue by 2027, while OpenAI's will remain stuck at 57%. OpenAI will burn fourteen times more cash than Anthropic before becoming profitable. Fourteen times.

The difference lies in strategy. Anthropic focuses on enterprise: 80% of revenue comes from 300,000 business customers. No expensive generative video like Sora, no browser, no humanoid robots, no hardware with famous designers. Just models that work, sold to people who use them to work. Less spectacle, more discipline. And backing from Amazon, Google, Microsoft and Nvidia: giants that can absorb losses indefinitely.

This is why Cogniquity chose Anthropic and Google as technology partners, not OpenAI. Not out of sympathy, out of risk analysis. Cogniquity is the latest venture in my 40-year journey through technological disruptions, from the birth of digital audio to the internet revolution in the creative industries, through organisational systems that produced real results across different sectors. When you've seen enough cycles of hype followed by collapse, you develop a certain sensitivity to the warning signs. And OpenAI's warning signs are flashing red.

When you build a working system that needs to function for years, you don't rest it on foundations that could collapse in eighteen months. You don't tie your business model to a company burning $14 billion a year without a credible path to profitability. You don't bet everything on whoever's playing the WeWork game of artificial intelligence.

Claude for complex reasoning and fact-checking. Gemini for pattern recognition and access to the Google ecosystem. Two providers, two different philosophies, no dependence on a single point of failure. Diversification isn't paranoia: it's operational hygiene in a market where today's leader could be tomorrow's Blockbuster.

This doesn't mean Anthropic is immune. They too burn cash, they too depend on investor confidence, they too could suffer if the bubble bursts. But there's a difference between driving with your seatbelt fastened and driving blindfolded at 120 mph convinced the road will always be straight. OpenAI is doing the latter. And millions of users are in the car with them without knowing it.

Mallaby, the economist, concludes that an OpenAI failure wouldn't be an indictment of artificial intelligence. It would merely be "the end of its most hype-driven builder." The technology will survive. The company perhaps not. And with it, the market value of all those skills built exclusively around its products.

DeepSeek has announced it will release a new model around Lunar New Year, mid-February. OpenAI, on its company blog, wrote that it expects another "seismic shock" from China. They know what's coming. It's not clear what they can do to stop it.

Meanwhile, millions of Western users continue using ChatGPT for free, unaware they're helping drain the coffers of a company haemorrhaging cash. And thousands of professionals continue building careers on mastery of a tool that could turn out to be the Betamax, the BlackBerry, the Blockbuster of the AI revolution.

Remember Blockbuster? Loyalty card, accumulated points, the Friday night ritual of choosing a film. It seemed impossible it could disappear. Then Netflix arrived, with a completely different model, built on budget constraints Blockbuster didn't have, and in five years nothing was left. The last Blockbuster in the world is in Bend, Oregon. It's become a tourist attraction. People go there to take selfies in front of the VHS tapes.

Your certified prompt engineering expertise could meet the same fate. A vintage curiosity. "Grandad, did you really write instructions by hand to make the AI work?"

The lesson, for those with eyes to see it, is always the same. The market leader isn't necessarily the long-term winner. First mover advantage is a myth that history disproves with relentless regularity. From Netscape to Nokia to Kodak. Whoever arrives first pays the cost of exploration. Whoever arrives later, with fewer resources and more constraints, is forced to find better paths. And often does.

This doesn't mean ChatGPT is useless today. It means building a professional identity entirely around a tool from a company burning $14 billion a year without a clear path to profitability is a high-risk bet. And most people making it don't even know they're betting.

The future of generative AI is being written elsewhere. In a Chinese city you can't pronounce, by a team that turned American sanctions into their competitive advantage. While you were busy optimising prompts to generate LinkedIn posts.

If history teaches anything, it's that constraints don't slow innovation. They accelerate it. And whoever always bets on the horse with the most resources often ends up wondering how they managed to lose.

But don't worry. Do carry on with the prompt engineering courses. I'm sure they'll work out splendidly.