AI bubble watch: OpenAI to burn through $115B by 2029

The AI emperor is not wearing any clothes — I mean, seriously, he’s as naked as a jaybird. Or to be more precise, the big-name AI companies are burning through unprecedented amounts of money this year. With no clear plan to make a profit anytime soon. (Sure, Nvidia is coining cash with its chip foundries, but the AI software companies are another matter.)

For example, we now know — thanks to analysis from The Information — that OpenAI will burn $115 billion (that’s billion with a capital B) by 2029. That’s up $80 billion from previous estimates. On top of that, OpenAI has ordered $10 billion of Broadcom’s yet-to-ship AI chips for its yet-to-break-ground proprietary data centers. Oh, and there’s the $500 billion that OpenAI and friends SoftBank, Oracle, and MGX are already committed to spending on the Stargate Project AI data centers.

It’s not just OpenAI. Meta, Amazon, Alphabet, and Microsoft will collectively spend up to $320 billion in 2025 on AI-related technologies. Amazon alone aims to invest more than $100 billion on its AI initiatives, while Microsoft will dedicate $80 billion to datacenter expansion for AI workloads. And Meta’s CEO has set an AI budget of around $60 billion for this year. 

It all adds up, and sooner or later you’re talking real money! 

Excuse me, but when companies are throwing money around like this, I’d really like to see a business plan that’s more than 1) AI! 2) Magic happens! 3) Profit! Because without that profit, something somewhere has to give.

Sure, OpenAI is now forecasting it will make $12 billion in revenue for 2025, escalating to more than $100 billion by 2029. But are those numbers real? It’s a private company, so all we really have is a belief that everything will all work out.

Interesting, isn’t it, though, how The Information also just announced that Microsoft, OpenAI’s godfather, has just admitted it’ll soon be buying services from Anthropic.

It’s true, venture capital and the stock market are still pouring cash into AI companies like there’s no tomorrow. They clearly are drinking the AI Kool-Aid.  

The bad news is that tomorrow does come, and people expect a profit. Sure, they all see growth is the most valuable aspect of tech businesses. That’s why mature firms like Amazon, Google, and Microsoft all like to pretend they’re growth companies. (Amazon CEO Andy Jassy can say Amazon is the “world’s largest startup” all he wants. It’s not.)

At the end of the day, though, someone has to pay cold, hard cash for your services. So tell me, who is going to be paying for AI? Yeah, I know the hype and I hear the cheerleaders shouting: “AI! AI! AI!”

I also know studies have shown that developers are actually less effective when they’re using AI tools. Yes, they produce more lines of code (LoC) in less time, but LoC has always been a lousy metric for measuring productivity.

As programming genius Edsger Dijkstra said back in 1988, LoC gives us “the reassuring illusion that programs are just devices like any others, the only difference admitted being that their manufacture might require a new type of craftsmen, viz., programmers. From there, it is only a small step to measuring ‘programmer productivity’ in terms of ‘number of lines of code produced per month.’ This is a very costly measuring unit because it encourages the writing of insipid code.”

Today, for “insipid” read “bad” code. It turns out AI creates poor-quality code, and you need to clean it up and find and fix security holes before you can send it to production. The time you save by vibe-coding your project is lost to fixing its mistakes. 

At least experienced programmers can find and fix those bugs. As technical writer Kaustubh Saini observed, “the core problem with vibe coding: it produces developers who can generate code but can’t understand, debug, or maintain it. When AI-generated code breaks, these developers are helpless.”

That’s just one example of how the hype around AI is illusory. In the short term, the technology looks great. In the long term, not so much.

You don’t need to believe me or think coding is just one area where AI falls short. According to MIT’s NANDA (Networked Agents and Decentralized AI) report, 95% of companies that have adopted AI have yet to see any meaningful return on their investment. 

The result, according to the US Census Bureau’s Business Trends and Outlooks Survey (BTOS), is that “AI adoption has been declining among companies with more than 250 employees.” You know, the ones big AI firms need if they’re ever to see a profit. 

Even if you’re willing to pay $20 a month per employee, even without any real return, you’re currently paying loss-leader rates for AI. We’re already seeing companies such as AnySphere and Anthropic raising their introductory $20 a month plans to $200 a month. They still aren’t charging enough to break even. 

Still, OpenAI ChatGPT, the face of the AI revolution, has about 700 million weekly users. Sounds good, doesn’t it? Guess what? A mere 5 million pay for it. I can’t make that math work. 

As Ed Zitron, a fellow tech journalist, notes in his popular AI blog, the costs of running generative AI do not make sense. And they’re only going to get worse. 

I’m calling it now. We’re in the middle of an AI is a bubble and at some point it’s going to pop. 

It’s not that AI isn’t useful at all. It is. It’s just like the internet. Yes, the dot-com bubble was as real as a slap to the face, but eventually the internet became the foundation of our entire economy. The keyword here is “eventually.” 

It took nearly six years for the US stock market and economy to recover from the dot-com crash.  Some technology indices, such as the Nasdaq, took even longer, with a full recovery only in 2015, needing 15 years to surpass their dot-com peak. Let’s hope our recovery from our current irrational AI bubble doesn’t take that long. Apple is making MDM migration so much easier – ComputerworldRead More