New AI models usually create excitement among the true believers and this year hasn’t been an exception. OpenAI and Anthropic released GPT-5.3 Codex and Claude Opus 4.6 respectively on the same day, February 5. The Opus 4.6 included new add-ons that enable Claude Code to perform a range of functions typically filled by software providers, which came on top of predictions by Anthropic’s CEO, Dario Amodei, that 50% of entry-level white-collar jobs could vanish in 5 years as AI takes over workplaces.

The market reacted very quickly. Shares of software-as-a-service companies like Adobe, Intuit, and Salesforce declined sharply on fears that AI tools might chip away at their business. Tech giants with large AI businesses like Microsoft, Amazon, and Google were also hit hard. A trillion dollars in market cap has been recently wiped out.

The Hype A recent post on X packaged these and other events into a great story, “Something Big Is Happening,” and it has received 70 million views in the last few days. The post places these events within the context of how much generative AI has changed since OpenAI released ChatGPT in 2022, and yes, those changes are remarkable.

But he leaves a lot out, unfortunately. First, generative neural networks have been around since the late 2000s, and big companies were pushing them on selected customers several years before Sam Altman released ChatGPT to everyone at prices far below costs, which is why OpenAI, Anthropic, and almost every other AI software company is losing unprecedented amounts of money.

Second, the blog cites a study by METR, “that tracks the length of real-world tasks (measured by how long they take a human expert) that a model can complete successfully end-to-end without human help. About a year ago, the answer was roughly ten minutes. Then it was an hour. Then several hours. The most recent measurement (Claude Opus 4.5, from November) showed the AI completing tasks that take a human expert nearly five hours. And that number is doubling approximately every seven months, with recent data suggesting it may be accelerating to as fast as every four months.”

Sounds, great doesn’t it? But the blog, and many citations of METR’s work, doesn’t mention that the model only has to complete the tasks with 50% accuracy, which is unacceptable for almost every job in the world.

Why doesn’t METR or somebody else publish data on how long it is taking these AI models to perform complete these tasks at 99% accuracy or to go from 50% to 99% or higher? Because it would take a long time for these improvements to occur even if there wasn’t diminishing returns, which there are.

I am telling you this because I devoted many years to understanding rates of improvement in new technologies (including AGI). Only a small number of technologies experience the type of improvements mentioned in the blog post of let’s say more than 30% per year, either before or after commercialization occurs, and they do help us understand why new systems such as the iPhone emerged when they did.

The bottom line: Even with a 50% annual reduction in errors, and no diminishing returns, it would take 10 years to reach 99.95%, and a 50% annual reduction in errors will not occur.

What’s Behind the Hype? Meanwhile, in a seemingly separate announcement, CNBC published an article entitled Google, Microsoft pay creators $500,000 and more to promote AI. Nice salary. Many of us have known that tech companies pay media sites to publish articles and pay individuals to post, like, and in other ways hype tech company announcements. I discussed this in a 2019 article in Issues in Science & Technology, “What’s Behind Technological Hype.”

Why is the latest round of hype happening now and why do so many people want to hear this hype? Because the tech sector is scared. The AI bubble is purported to be worth $35 trillion and cracks are spreading in the edifice that supports the bubble.

As many readers know, when users access ChatGPT, OpenAI must pay the cloud center providers such as Oracle, Microsoft, and Google. But OpenAI, Anthropic, and other AI software companies can’t pay because they have set their prices too low in a successful effort to become the fastest growing technology of all time.

One solution is to have companies invest in OpenAI in return for buying their chips. Suppliers of GPUs such as Nvidia and AMD do this as well as cloud companies such as Oracle and CoreWeave. Called circular financing, it is merely a short-term solution.

The Stock Market is Getting Nervous

The stock market is beginning to question the circular financing and the share prices of the big investors in AI software companies. Softbank’s stock is down 1/3 in three months from its investments in OpenAI. Microsoft’s share price is down 10% in a week and 20% in three months because 45% of Microsoft’s $625 billion in bookings of future cloud contracts comes from OpenAI. That’s roughly $280 billion in promised payments from a company that lost $12 billion in the third quarter of 2025, and it expects to have $115 billion in cumulative losses by 2029.

There are also concerns about cloud companies in general because they are shouldering much of the debt. Bloomberg claims that “AI related companies have accrued at least $200 billion in debt,” and the figure is likely considerably higher because that estimate doesn’t count undisclosed private deals.

However, instead of putting all the debt on the parent balance sheet, cloud center suppliers such as Meta, Google, Oracle, and Salesforce often use special purpose vehicles (SPVs) that borrow against long leases, so repayment depends on contracted rent rather than the developer’s survival. This makes the company’s balance sheets look debt free, when they are not.

These problems have been widely known for at least a year, but it is just now that the numbers have begun to impact greatly on a few companies such as Oracle, Softbank, and Microsoft (and less so with others). Those are the first cracks, much as the first hints of the popping of the subprime mortgage came from the problems at Bear Stearns. Could the venerable Oracle actually go bankrupt?

Oracle’s latest activities suggest that it recognizes the urgency in a strange way. It announced this month that it’s raising a staggering $45 billion to $50 billion in debt and equity sales to build additional cloud infrastructure capacity, plans that suggest it believes the only forward is to do more of the same, and pay higher than normal interest rates. Bondholders are suing.

Nvidia has managed to avoid share price declines probably because it keeps waffling about funding OpenAI. Last week, almost every day brought a new nuance to Jensen Huang’s dance around this investment. He has been happy to receive orders from OpenAI’s cloud suppliers (e.g., Microsoft, Oracle) even as smart people know that if OpenAI isn’t funded, Nvidia’s revenues will also dry up. It will be a great case study in the future.

The final part of this story is the merger between xAI and Space X in preparation for an IPO (OpenAI and Anthropic are also planning IPOs). One reason for the merger is because CEO Elon Musk needs more funds to send data centers to space. That plan may require more money than anything mentioned above. Get ready for more excitement, but don’t mistake excitement for big advances in productivity.