The Meta warning: When AI spending becomes a liability

What the broader market context means for enterprises

Nicholson expects to see more targeted AI investment in the coming years.

"Within the next two years, we're going to see that not everyone is going to be willing to maintain this level of investment because they're not going to be able to demonstrate positive ROI coming out of it," he said.

The market will become savvier at differentiating between speculative and sustainable AI strategies and invest accordingly.

"If you were to draw a line from now to 10 to 20 years from now, it's going to look like upward sloping, continuous growth," said Nicholson. "The closer you get to it, you're going to see some serious dips for individual organizations."

As enterprise boards observe spending trends at big tech companies and the market's reaction, their view of AI investment could shift from more spending to a focus on coming out ahead.

"We've definitely gone from fear of missing out to fear of screwing up," Nicholson said. "The majority of existing enterprises are finding that the fast follower model is absolutely the best thing to pursue."

Warning signs AI spending has become a liability

The AI investment narrative has shifted. Experimentation is over, and it's time to deliver value. While market expectations may be driving that change, there is still plenty of trial and error ahead.

"If you're treating your AI spend, your pilots, your pre-production experiments as an R&D program, you're comfortable with the idea that you're going to spend some money and find out what doesn't work just as much as you find out what does," said Alex Bakker, distinguished analyst and director of primary research at ISG (Information Services Group).

But that doesn't mean enterprises can get too comfortable. CIOs are being asked tough questions, such as when AI investment tips from strategic necessity to an enterprise liability.

When Capex continues to balloon without measurable outcomes, boards and investors will see a red flag. A nebulous promise of value with no metrics to demonstrate progress between the 'then and now' is not a solid foundation for more investment.

"There are countless examples of 18-month failures where millions of dollars have been spent, and nothing has been accomplished," Nicholson said.

Spending without cost controls sets enterprises up for failure; costs could continue to creep up, either outpacing any value or going up while value fails to materialize.

"Token generation can get out of hand really quickly. As soon as you have agents that are leveraging models on your behalf…If you thought that cloud cost creep was a problem, we haven't seen anything yet with the way that tokens are being generated," Nicholson said.

CIOs need to be able to articulate the business case for AI and tie that to spending before they start accumulating technical debt, according to Ravi Soin, CIO and CISO of Smartsheet, a work management platform.

"If you don't have that in spend control without clarity on the use case, you're essentially accumulating that debt disguised as innovation or velocity," he said.  

There is more than one way enterprises might disguise AI spending that is getting out of control. Layoffs are currently a major trend in the tech sector. Meta is set to cut 8,000 jobs on May 20, 2026. The New York Times reported that the company's Chief People Officer, Janelle Gale, said the cuts are a "part of our continued effort to run the company more efficiently and to allow us to offset the other investments we're making," in a memo to Meta employees.

Do big layoffs like these mean tech companies are getting massive AI efficiency gains, or are they "AI washing?" Meta employee Arnav Gupta, who is waiting to find out if he is among the laid off, argues in an X post that these layoffs are a consequence of companies floundering. "These layoffs will continue till we learn to use AI. Till we learn to convert AI-tokens into outcomes and not just input."

Building a framework for sustainable AI investment

Managing AI costs differs in many ways from managing the costs of other technologies, but enterprise leaders have plenty of experience in prudent purchasing and investing. How can they apply that experience in the fast-paced and novel world of AI?

Deciding to build, buy or partner

Big tech companies are building their own AI infrastructure, but that doesn't mean every enterprise should be doing the same. Enterprises have to determine if they will build AI capabilities in-house, use outside vendors or take a hybrid approach. Each option has unique cost considerations.

For some enterprises, the upfront cost of building their own infrastructure may make sense. But those companies are not in the majority, according to Nicholson.

"I counsel my students that unless you're trying to develop something to sell to other people specifically, the risk associated with trying to pioneer something really, truly new probably isn't worth it," he said. "If you're an Oracle shop and you want to deploy this in your database environment, let Oracle figure that out. It'll only take them a few months.

Understanding your costs

ROI remains the most critical measure of AI success; enterprises are spending so much on AI with the expectation that it will all be worth it in the end. But getting there requires truly understanding the total cost.

"Positive ROI? First, figure out the 'I.' And that's tough in the token era," said Nicholson.

Soin expects many organizations to be surprised by the total cost. "The license is the smallest cost in any of these AI functionalities," he said.

CIOs also must account for costs associated with implementation, change management, systems integration and ongoing prompt engineering. Enterprises that opt to build also have to add up their infrastructure costs.

Tying spend to value

CIOs need metrics to quantify the value of AI investment. Big, ambitious projects with multi-year horizons are proliferating. But it is more difficult to measure the effect of these initiatives and justify the ongoing expenses.

"It's a disciplined, limited proof-of-value exercise that will save CIOs their jobs," Nicholson said. "Small proof-of-value tests can teach you a lot about what makes sense to invest in."

Setting realistic timelines for ROI

Among the CIOs and CTOs Nicholson works with, 18 to 24 months appears to be the typical time horizon they are given for delivering ROI on AI investments. Setting a timeline and sticking to it can help enterprises avoid spending more on a use case that won't work as intended.

"Be honest with yourself and with your measures," Bakker said. "If you're not seeing it move, don't keep going. Cut the spend, try again, allocate tokens and effort and human time and energy to a new use case."

Monitoring spending in real time

AI spending demands disciplined oversight, not a set-it-and-forget-it budget.

"You need real time visibility, not this quarterly checkpoint that you may have for these AI cost models," said Soin. "Do you have the right framework and guardrails that are associated with it to trigger an alert or isolation when those cost overruns are happening on that workload level?"

Stress testing investment strategy

"[Look] at the overall AI spend as a percentage of that budget and how it's trended year over year. I think that establishes the scale and trajectory which an organization is going to be facing," Soin said.

Is that percentage sustainable? Can it be adjusted? CIOs need to be able to think about how enterprise AI investment strategy can adapt to changes in vendor capabilities, business needs, use case outcomes and enterprise revenue.

Communicating to boards and investors

Enterprise boards and investors are seeing the enormous investment in AI and the promise that the technology will drive measurable gains in efficiency and productivity.

"They are being dragged along by the global hype, and those expectations are rolling down on top of the people that work in their organizations," Nicholson said.

CIOs are at the frontlines of delivering on and managing those expectations. They are well-positioned to educate their boards on AI and how it can drive business value, which requires CIOs to talk to their boards as if they were investors, according to Bakker. Investors understand spending -- and they understand risk. It is the CIO's job to talk to board members about aligning the two.

"If you're a high-dividend business with a very steady cash flow and very steady businesses, maybe your AI strategy should be sensitive to that," Bakker said. "If you are a super high-growth company, maybe you have to be taking a bigger risk right now."

The opportunity for CIOs

Big tech companies like Meta have billions to funnel into AI spending, but they are still responsible to investors. Enterprises and their CIOs can monitor how markets respond to AI spending trends as they continue to develop their investment strategies.

CIOs can help their enterprises develop intentional investment frameworks that tie AI to business value.

"The CIOs, in my mind, who come out ahead aren't the ones who are spending the most on AI," Soin said. "Can you point to deployed solutions with measurable value and governance structure that lets them scale responsibly?"

Carrie Pallardy is a freelance journalist with experience writing in cybersecurity, technology and healthcare. She currently covers a wide range of issues relevant to today's CIOs and IT leaders.