Sam Altman, chief executive officer of OpenAI Inc., during a media tour of the Stargate AI data center in Abilene, Texas, US, on Tuesday, Sept. 23, 2025.
Kyle Grillot | Bloomberg | Getty Images
With OpenAI’s recent release of its AI browser, the historic level of capital expenditures being made in the current AI arms race may accelerate even further, if that is possible.
From the reciprocal, and some have said circular, nature of hundreds of billions in commitments in investment, tied to future chip purchases, to the extent to which GDP growth is reliant on this boom, some have said this is a bubble. A Harvard economist estimates 92% of US GDP growth in the first half of 2025 was due to investment in AI.
But much more needs to be understood about the connection between the breakneck investment in AI and the business models that underpins the entire economy: the advertising technology (Ad Tech) industrial complex.
For the past 25 years the infrastructure of the internet has been engineered to extract advertising revenue. Search Engine Marketing, the advertising business model at the core of Google, is perhaps the greatest business model of all time. Meta’s advertising business, based on engagement and attribution, is a close second. And right behind both of these is Amazon’s advertising business, powered by its position as the largest online retailer. While a smaller portion of Amazon‘s topline, its highly profitable advertising business makes up a disproportionate percentage of Amazon’s profits. So much so that nearly every major retailer has spun up their own version of retail media networks, all driving significantly to the bottom lines and market capitalization of massive companies like Walmart, Kroger, Uber (and UberEats), Doordash and many more.
In fact, these platforms have been using AI to refine their advertising business models for years, in the form of algorithmic models that powered their search and recommendation engines, and to increase engagement and better predict purchase decision, seeking an ever-greater share of all commerce, not just what is typically thought of as “advertising.” These three multi-trillion-dollar market cap companies either wholly, or substantially, derive their profits from advertising. And now they are using some portion of those historically profitable advertising revenues to fuel infrastructure investments at a level the world has not seen outside of War Time spending by governments.
But at the same time, the latest wave of AI has the potential to disrupt the very same trillions in market cap that is fueling it. AI will, without question, change how people search (Google), shop (Amazon) and are entertained (Meta). Answers delivered without clicking around the web. AI-assisted shopping. Infinite personalized content creation.
If AI represents such a potential existential risk, why are Google, Meta and Amazon such a huge part of the current arms race to invest in AI? The “moonshot” outcome of would be that achieving Artificial General Intelligence, or Super Intelligence, AI that can do anything a human can, but better, would unlock so much value that it would dwarf any investment.
But there is more immediate urgency to protect, or disrupt, the advertising business model fueling the trillions in market cap and hundreds of billions of current investment, before someone else does. While the seminal paper that launched this phase of AI, “Attention is All You Need” was written by mostly Google researchers, it was OpenAI and Microsoft, and now Grok as well, that launched the current AI arms race. And they are not remotely as dependent on the current advertising industrial complex. In fact, Sam Altman has called the feeds of the major platforms using AI to maximize advertising dollars, “the first at-scale misaligned AIs.” He is clearly stating which businesses he believes OpenAI is trying to disrupt.
What comes next?
This time is different, but it also comes with different risks. The major difference with the current fever in infrastructure investment vs the dotcom bubble of 2000, is that in large part the companies funding it are among the most profitable companies in the world. And so far, there has not been indications of cracks in the business model of advertising that is both funding their investments, and their market capitalizations (along with so many massive companies people wouldn’t think about being in the advertising business).
But if AI does disrupt, or even break, the current advertising model, the shock to the economy and markets would be far greater than most could imagine.
Google, Meta and Amazon are still best positioned to create new business models, and as mentioned, have been using AI for far longer to support their advertising business models with great success.
However, fundamentally changing the way people interface with search, commerce and content online will require just that, entirely new revenue models, maybe, hopefully, some that are aligned, that are not advertising based. But whatever the model, perhaps it is helpful to consider that the justification in AI infrastructure spending may not be to just unlock new revenue, but to protect the business models that make up a much more significant portion of the market capitalization of public companies than most people are aware.
Meta CEO Mark Zuckerberg has repositioned the social media giant as an AI company.
Vincent Feuray | AFP | Getty Images
Meta Platforms shares popped about 4% higher on Thursday after Bloomberg reported that CEO Mark Zuckerberg was looking to make significant cuts to the company’s metaverse resources.
Bloomberg said that executives have considered budget cuts as high as 30% for the unit, citing people familiar with the talks.
The move would be notable for the Facebook parent company, which changed its name to Meta in October 2021 to signal its pivot beyond social media.
Zuckerberg said at the time that “the metaverse is the next frontier just like social networking was when we got started.”
The proposed cuts would likely include layoffs, according to Bloomberg, which said the cuts were part of budget planning for 2026. The cuts will likely hit the company’s virtual reality group.
Meta did not immediately respond to a request for comment.
Read more CNBC tech news
Meta’s Reality Labs unit, which develops the Quest family of VR headsets and Ray-Ban and Oakley AI smart glasses, reported a $4.4 billion loss in the company’s most recent quarter.
The division had recorded over $70 billion in cumulative losses since late 2020 as of the third-quarter report.
When Nvidia this week said it would take a $2 billion stake in chip design company Synopsys, it was just the latest in a string of massive investments announced by the chipmaker this year.
That doesn’t even include the biggest commitment of all: $100 billion to buy OpenAI shares over a number of years, although there is still no definitive agreement, Nvidia finance chief Colette Kress said Tuesday.
It’s a lot of money and a lot of deals, but Nvidia’s got the cash to write big checks.
At the end of October, Nvidia had $60.6 billion in cash and short-term investments. That’s up from $13.3 billion in January 2023, just after OpenAI released ChatGPT. That launch three years ago was key to making Nvidia’s chips the most valuable tech product.
As Nvidia has transformed from a maker of gaming technology into the most valuable U.S. company, its balance sheet has become a fortress, and investors are increasingly wondering what the company will do with its cash.
“No company has grown at the scale that we’re talking about,” said CEO Jensen Huang, when asked what the company plans to do with all its cash, on Nvidia’s earnings call last month.
Analysts polled by FactSet expect the company to generate $96.85 billion in free cash flow this year alone and $576 billion in free cash flow over the next three years.
Some analysts would like to see Nvidia spend more of its cash on share repurchases.
“Nvidia is set to generate over $600B in free cash flow over the next few years and it should have a lot left over for opportunistic buybacks,” wrote Melius Research analyst Ben Reitzes in a note on Monday.
The company’s board increased its share repurchase authorization in August, adding $60 billion to its total. In the first three quarters of the year, it spent $37 billion on share repurchases and dividends.
“We’re going to continue to do stock buybacks,” Huang said.
Nvidia is doing the buybacks, but it’s not stopping there.
Huang said that Nvidia’s balance sheet strength gives its customers and suppliers confidence that orders in the future, which he called offtake, will be filled.
“Our reputation and our credibility is incredible,” Huang said. “It takes a really strong balance sheet to do that, to support the level of growth and the rate of growth and the magnitude associated with that.”
Kress, Nvidia’s CFO, on Tuesday said that the company’s “largest focus” is making sure it has enough cash to deliver its next-generation products on time. Most of Nvidia’s largest suppliers are equipment manufacturers like Foxconn and Dell, which can require that Nvidia provide working capital to manage inventory and build additional manufacturing capacity.
Huang called his company’s strategic investments “really important work” and said that if companies like OpenAI grow, it drives additional consumption of AI and Nvidia’s chips. Nvidia has said that it does not require any of its investments to use its products, but they all do anyway.
“All of the investments that we’ve done so far — all of it, period — is associated with expanding the reach of Cuda, expanding the ecosystem,” Huang said, referring to the company’s AI software.
In an October filing, Nvidia said it had has already made $8.2 billion in investments in private companies. For Nvidia, those investments have replaced acquisitions.
Nvidia’s $7 billion acquisition of Mellanox in 2020 is the largest the company has ever made, and it laid the groundwork for its current AI products, which aren’t single chips but entire server racks that sell for around an estimated $3 million.
But the company faced regulatory issues when it tried to buy chip technology firm Arm for $40 billion in 2020.
Nvidia called off the deal before it could be completed after regulators in the U.S. and UK raised concerns about its effects on competition in the chip industry. Nvidia has purchased some smaller companies in recent years, to bolster its engineering teams, but it hasn’t completed a multi-billion acquisition since the Arm deal failed.
“It’s hard to think about very significant, large types of M&A,” Kress this week said, speaking at an investor conference. “I wish one would come available, but it’s not going to be very easy to do so.”
A sign at a NYS Department Of Labor job fair at the Downtown Central Library in Buffalo, New York, US, on Wednesday, Aug. 27, 2025.
Lauren Petracca | Bloomberg | Getty Images
This is CNBC’s Morning Squawk newsletter. Subscribe here to receive future editions in your inbox.
Here are five key things investors need to know to start the trading day:
1. Silver linings playbook
Yesterday highlighted the relevance of a market adage: Bad news can actually be good news for investors. After private payroll data showed weakness in the labor market, stocks climbed as investors hoped the report would strengthen the case for an interest rate cut at the Federal Reserve’s meeting next week.
Here’s what to know:
The ADP reported a surprise decline of 32,000 jobs in November. Economists surveyed by Dow Jones were forecasting a gain of 40,000.
The Dow Jones Industrial Average rallied more than 400 points in Wednesday’s session, pulling the 30-stock index into positive territory for the week.
Traders are now pricing in a roughly 89% likelihood of a rate cut, up from under 70% a month ago, according to CME’s FedWatch tool.
Data released by Challenger, Gray & Christmas this morning also showed layoff announcements this year totaled the most since 2020, another sign of the labor market’s slowdown.
Speaking of tariffs, Treasury Secretary Scott Bessent said that the Trump administration can replicate the sweeping levies if the Supreme Court rules the president exceeded his authority to enact the duties.
Salesforceblew past earnings per share expectations for the third quarter, sending shares higher in today’s premarket. While the company’s quarterly revenue came in slightly under Wall Street’s consensus forecast, Salesforce offered stronger-than-anticipated revenue guidance for the current three-month period.
Salesforce also said annualized revenue from its Agentforce AI software jumped 330% year over year. The firm set a better-than-expected revenue target of $60 billion for fiscal 2030 for Agentforce.
3. Jensen’s jaunt
Nvidia President and CEO Jensen Huang speaks to the media as he arrives for a meeting with the Senate Banking Committee on Capitol Hill on December 3, 2025 in Washington, DC.
Anna Moneymaker | Getty Images
Nvidia CEO Jensen Huang returned to Washington, D.C. yesterday to meet with Trump and discuss chip export restrictions. Huang then went to Capitol Hill, where lawmakers are weighing whether to approve a rule that would limit AI chip exports.
Huang said the Guaranteeing Access and Innovation for National Artificial Intelligence Act — known as the GAIN AI Act — “is even more detrimental to the United States than the AI Diffusion Act.” Huang also broke with some of his fellow AI executives by slamming state-by-state AI regulation. Such oversight would “drag this industry into a halt” and would “create a national security concern,” he said.
Get Morning Squawk directly in your inbox
4. Vaccination vote
Massachusetts Institute of Technology professor Retsef Levi speaks during an Advisory Committee on Immunization Practices meeting at the Centers for Disease Control and Prevention in Atlanta, Sept. 19, 2025.
Alyssa Pointer | Reuters
Health and Human Services Secretary Robert F. Kennedy Jr.’s hand-picked Advisory Committee on Immunization Practices is slated to vote today. On the docket: whether to change its longstanding recommendation that babies gets the hepatitis B vaccination within 24 hours of birth.
While it’s unclear how the committee will rule, any change to the recommendation would have major impacts within public health. Some experts caution that doing away with the decades-old recommendation could lead to a higher rate of chronic infections in children.
5. New terrain
GM Chief Product Officer Sterling Anderson during the automaker’s “GM Forward” event on Oct. 22, 2025 in New York City.
Anderson’s remit includes overseeing “the end-to-end product lifecycle” of GM’s vehicles, according to the company. He told CNBC that the he wants to see a faster rate of innovation and create a “unified approach” to product.
Also helping General Motors: Trump’s decision to cut tariffs on South Korea. The company is the second-largest new vehicle importer from the country, behind South Korea-based Hyundai Motor.
Current-quarter earnings per share impact: 25 cents
— CNBC’s Sean Conlon, Jeff Cox, Kevin Breuninger, Jordan Novet, Annie Palmer, Ashley Capoot, Annika Kim Constantino, Mike Wayland and Leslie Josephs contributed to this report. Josephine Rozzelle edited this edition.