AMZN, MSFT, NVDA, SFTBY setting $100 billion on fire

Amazon (AMZN), Microsoft (MSFT), Nvidia (NVDA), and SoftBank (SFTBY) are close to throwing a lot of money into a furnace, that is, OpenAI.

The first phase of the company’s latest funding round is almost complete, and if the companies invest near the highest ranges of what has been hashed out, the total investment will be close to $100 billion, according to Bloomberg.

Amazon is expected to invest up to $50 billion, SoftBank as much as $30 billion, and Nvidia $20 billion, according to Bloomberg, but Reuters reports that Nvidia will invest $30 billion.

Why am I comparing OpenAI with a money-burning furnace? Well, quite simply, it is not profitable, and it has no clear path to profitability, only ridiculous spending plans.

To understand why this funding is like lighting money on fire, we need to take a look at OpenAI and then examine specific problems with each of the investing companies.

OpenAI’s spending plans make no sense

Let’s begin by defining ridiculous spending plans.

“We expect to end this year above $20 billion in annualized revenue run rate and grow to hundreds of billion by 2030,” OpenAI CEO Sam Altmanposted on X (formerly Twitter) in November 2025. “We are looking at commitments of about $1.4 trillion over the next 8 years.” 

Tomasz Tunguz, general partner at Theory Ventures, did the math regarding this plan. He started with a premise that OpenAI has committed to spending $1.15 trillion on hardware & cloud infrastructure between 2025 & 2035, because the information Altman posted wasn’t available yet. So we know that Altman has committed to spending more money in less time. Keep that in mind.

Tunguz’s calculation suggested that OpenAI would need to grow from approximately $10 billion in 2024 revenue to $577 billion by 2029. This is roughly the size of Google’s revenue in the same year, he points out, and this is all with a more optimistic timeframe and less total spending.

Internal OpenAI forecasts predict the for-profit part of the company will hit $100 billion in annual revenues in 2029, according to The Information, as reported by Yahoo Finance.

An immense difference exists between what OpenAI will be making and how much it will need. You’d have to believe in unicorns to think OpenAI will achieve its $1.4 trillion spending. It’s obvious the company will face renegotiation of existing infrastructure contracts.

OpenAI doesn’t have a path to profitability

How long would it be reasonable to wait for the company to become profitable? OpenAI was founded in 2015, and HSBC Global Investment Research projects that OpenAI still won’t be profitable by 2030, according to Fortune.

Is 15 years of unprofitability sound business? Along with plans to spend $1.4 trillion? I don’t think so.

The company has recently made a U-turn on its plans to keep ChatGPT ad-free. This shows that the subscription-based model isn’t growing fast enough.

There is a huge problem with the new change. Users of services that also serve ads often tolerate them because they expect a certain level of quality and reliability that gets them “hooked” on the service.

OpenAI had an ace up its sleeve with ChatGPT 4o, as it made the model very sycophantic. When the model launched, Altman’s very short post on X simply said, “her.” He seemed to imply that 4o was built to foster relationships like the AI in the movie “Her.”

More Tech Stocks:

• Morgan Stanley sets jaw-dropping Micron price target after event
• Nvidia’s China chip problem isn’t what most investors think
• Quantum Computing makes $110 million move nobody saw coming

Unfortunately for OpenAI, its sycophantic AI model prompted a series of lawsuits. The most recent one, filed in January, was from a college student alleging that ChatGPT “convinced him that he was an oracle” and “pushed him into psychosis,” as reported by Ars Technica.

The company shut down the model on February 13.

This means the company no longer has a model on which people would tolerate seeing ads. Meanwhile, its competitor, Perplexity, is scrapping plans to put ads in its AI search product, to avoid provoking people’s mistrust, reported Wired.

Another OpenAI competitor, Anthropic, recently spent millions to mock OpenAI’s plan to introduce ads, as reported by Reuters.

Even if the introduction of ads to ChatGPT doesn’t result in users migrating to other LLMs, that doesn’t mean ads will suddenly turn OpenAI into a profitable company. “A path to generating several billion dollars in ad revenue in 2026, going to $25B+ by 2030, seems reasonable,” said Evercore ISI’s analyst Mark Mahaney, as Business Insider reported.

That $25 billion number would indeed be impressive if OpenAI didn’t need to spend so much on infrastructure and on training its AI models, and if it didn’t have competitors at all.

To understand how futile all this is, we need to think about profit margins. As AI critic Ed Zitron rightly points out in his blog, AI companies do not factor AI training into their margins, which, as he writes, “is inherently deceptive.”

There isn’t enough data available for OpenAI, but Zitron did the math using data from Anthropic.

Anthropic’s 2025 gross margins were 40%, according to Zitron, but once he added the cost of training, the margins sharply dropped to a negative 53%. That’s the true price tag for LLMs.

OpenAI’s margins are probably in the same negative ballpark. This is because LLMs need to be constantly patched for every possible prompt that could turn out to be malicious or could lead to dangerous “hallucinations.”

Models basically have to be trained until the end of time, since an infinite number of prompts could break them. It also means a constantly growing power bill.

The worst part may be that large language models have peaked, and in my opinion, they don’t even work that well.

The vast majority of 6000 executives see no impact from AI

The National Bureau of Economic Research recently did a survey on almost 6000 CFOs, CEOs, and executives from companies across the U.S., U.K., Germany, and Australia.

Here is what researchers found.

“On average, more than 90% of business managers across the four countries estimate no impact of AI on their employment over the past three years. 89% report no impact of AI on their labor productivity (measured as volume of sales per employee) over the last three years.”

Need more proof?

Center for AI Safety and Scale AI released a paper “Remote Labor Index: Measuring AI Automation of Remote Work,” in October 2025, showing some interesting stats.

They used what they call a Remote Labor Index (RLI), a multi-sector benchmark comprising real-world, economically valuable remote-work projects.

They updated the results since the paper’s release, and according to the latest numbers, the best-performing model is Anthropic’s Opusa 4.5, which achieved an automation rate of only 3.75%.

“This demonstrates that contemporary AI systems fail to complete the vast majority of projects at a quality level that would be accepted as commissioned work,” the researchers write.

Here’s how AI systems’ automation rates stack up

If you need convincing that the models have peaked, read Futurism’s article, “Scientists Are Getting Seriously Worried That We’ve Already Hit Peak AI.” 

There is one thing LLMs excel at, and I’ll admit that. They increase cybersecurity risks. Morgan Stanley agrees, and I wrote about that in my article “Morgan Stanley flags $45B hidden cybersecurity opportunity.”

And now, we need to take a look at the investors themselves.

The AI bubble must keep growing

Amazon is investing in OpenAI, yet it is also investing in its competitor, Anthropic. In addition, Amazon is working on its AI models.

How does investing in two unprofitable companies, while making a competing product, make sense? The best part is that Amazon will have to raise debt this year if it really wants to accomplish its spending plans. I’ve explained this in my article “Bank of America resets Amazon stock price target after earnings.”

Related: Analysts revamp Palantir stock rating

After my article posted, a Citi analyst also did the math and came to the same conclusion. Read the article “5-star analyst flags sharp contrast between Amazon, Google stock” by TheStreet’s Moz Farooque to explore the issue in more depth.

Microsoft AI CEO Mustafa Suleyman said the company is pursuing “true self-sufficiency” in AI by building its own powerful models and reducing its reliance on OpenAI, the Financial Times reported. Again, how does investing in an unprofitable competitor company make sense?

I covered Nvidia’s and Microsoft’s plans to invest in Anthropic in my article “Nvidia, Microsoft deal takes ‘circular’ financing to entirely new level.” So Nvidia’s situation is pretty much the same: investing in two unprofitable companies, and also making the same product competitors are making.

For the uninitiated, Nvidia offers a line of Nemotron open models aimed at organizations that want sovereign AI.

SoftBank is the only entity here without a conflict of interest, but it’s more like a gambler who placed a really bad bet and is now putting even more money in, hoping things will turn around.

The crazy part is that the company sold its Nvidia shares to complete the 2025 investments in OpenAI. SoftBank’s cash and cash equivalents at December 31, 2025, were ¥11.5 billion ($74.1 million), according to its Q3 earnings report.

After selling Nvidia, I wouldn’t be surprised if the company now sold its Intel stake to fund this “great” opportunity.

You might be asking: Why are these companies investing if it isn’t smart?

It’s actually quite simple. If they decline to invest, other investors will get spooked, and OpenAI runs out of money.

If OpenAI goes bust, the magical $1.4 trillion gets erased from AMZN, MSFT, NVDA, ORCL, AMD, AVGO, and CRWV backlogs. That is almost a total wipeout, and Anthropic’s next founding round also won’t happen after the OpenAI failure, as the panic will be unstoppable.

This would erase whatever money they have in their backlogs from Anthropic, and suddenly revenue projections don’t look that flattering, and the chain reaction continues.

Plugging the hole in the AI bubble is a terrible move, as the longer the bubble persists, the bigger the consequences for the economy once it bursts.

Related: Bank of America resets Nvidia stock forecast