CEOs Are Spending Billions on AI But 56% of Companies Admit the Tech Isn’t Actually Making Them Any Money Yet

Credit: ZME Science/AI-generated NanoBanana 2.

Despite the relentless hype and massive capital expenditures, the corporate world is facing a harsh truth. Artificial intelligence is simply not paying the bills yet.

A new survey from PwC, which gathered responses from 4,454 chief executives across 95 countries, found more than half of the surveyed executives — 56 percent, to be exact — admit their companies have realized neither higher revenues nor lower costs from their AI initiatives.

Some companies do report marginal wins. Roughly 30 percent of executives note increased revenue tied to AI over the past year. Another 26 percent say they have managed to decrease costs.

But the holy grail of winning with an AI-first strategy — achieving both cost reduction and revenue growth simultaneously — is rare. Only one in eight CEOs report hitting this dual milestone.

So, why the massive gap between Silicon Valley’s promises and corporate reality?

Why Is the Payoff So Elusive?

We are witnessing the largest privately financed technological wave in human history, an effort that financially dwarfs the Apollo Program and the Manhattan Project combined. According to Stanford University’s 2025 AI Index Report, global corporate investment in artificial intelligence reached a staggering $1.6 trillion between 2013 and 2024, with $252.3 billion spent in 2024 alone.

And this financial firehose is only opening wider.

Technology research firm Gartner projects that worldwide enterprise AI spending will skyrocket to an incomprehensible $2.5 trillion by the end of 2026. Companies are buying up software, training custom models, and hiring expensive talent in a desperate scramble to secure a competitive edge.

×

Thank you! One more thing…

Please check your inbox and confirm your subscription.

Then there are the tech giants building the physical plumbing of the AI revolution. Companies like Amazon, Microsoft, Alphabet, and Meta are locked in an unprecedented infrastructure arms race, buying up land, specialized silicon, and power grids to construct hyperscale data centers. In 2025, these tech behemoths collectively spent over $400 billion on CapEx, and financial analysts project that figure will surge past $600 billion — and potentially hit $650 billion — in 2026. To put that into perspective, just four technology companies are now spending more on physical infrastructure in a single year than the entire advanced economy of Sweden produces in its annual GDP.

It is a high-stakes gamble that controlling the underlying architecture of artificial intelligence will eventually yield historic profits. Being late in the AI game could be corporate suicide. But what if you’re too early?

True enterprise adoption requires rewiring deeply entrenched human workflows and corporate cultures, a lot of the value that AI could unlock is currently in limbo, waiting for the human side of the equation to catch up.

Currently, AI adoption remains largely experimental. PwC notes that companies are deploying AI to a “large or very large extent” in only a few specific areas, like demand generation (22 percent) or support services (20 percent).

Furthermore, the actual human beings doing the work aren’t using the tools as much as you might think. A separate PwC workforce study found that a mere 14 percent of workers use generative AI on a daily basis.

The consulting giant argues that companies fail because they dabble. PwC claims that isolated, tactical AI projects often fail to deliver measurable value. Tangible returns, they argue, only come when a company deploys the technology at an enterprise scale, deeply aligning it with their core business strategy.

“A small group of companies are already turning AI into measurable financial returns, while many others are still struggling to move beyond pilots,” Mohamed Kande, PwC’s global chairman, told Business Insider.

However, your instinct is right if you’re skeptical of hearing this sort of go big or go home advice. Pilot projects exist precisely to test a concept safely before risking a massive, expensive rollout. Pushing companies to scale up failing pilot projects requires an almost religious faith in the technology.

Yet, data from outside the PwC ecosystem supports the idea that scaling up AI is incredibly difficult. Recent MIT research indicates that only 5 percent of enterprises have successfully implemented AI tools at scale. The remaining 95 percent saw zero return on their investments.

Similarly, an EY survey found companies are missing out on 40 percent of the potential productivity gains AI could offer simply because they lack the right underlying data architecture and talent strategy.

And then there’s the elephant in the room: maybe AI is just not good enough yet. And that may be true for many use cases by the looks of it.

AI Reality Check

Salesforce recently learned a hard, expensive lesson about artificial intelligence. The enterprise software giant essentially had to put a leash on its autonomous AI product, Agentforce, after the bots started forgetting instructions and wandering off-topic during basic customer interactions.

As Sanjna Parulekar, Senior Vice President of Product Marketing at Salesforce, bluntly noted, “All of us were more confident about large language models a year ago.”

McDonald’s spent three years working with IBM to build an AI-powered drive-thru ordering system. The goal was obvious. The company wanted to speed up service, reduce human labor costs, and boost the bottom line.

Instead, the system became a viral joke. The AI misheard orders, frustrated hungry customers, and created massive operational inconsistencies. So, the new AI system was turned off.

It’s becoming a trope, honestly. Companies treat AI like traditional software that you can simply install and run. But AI operates on probability. When you ask traditional software to do something, it executes a specific command. When you ask a generative AI model to do something, it guesses the most likely correct response based on its training data. In a corporate environment, this guessing game can lead to disaster.

Take Air Canada, for example. In late 2023, a grieving passenger consulted the airline’s AI virtual assistant to ask about bereavement fare policies. The chatbot confidently gave the passenger entirely fabricated information, assuring him he could claim a refund after purchasing his tickets.

When the passenger tried to claim the refund, Air Canada refused, essentially arguing that the company wasn’t responsible for the actions of its own AI. A Canadian courthouse vehemently disagreed. They ordered the airline to pay damages, ruling that the company was liable for the negligent misrepresentations made by its chatbot.

Government agencies are falling into the same AI hype trap. Microsoft powered a chatbot for New York City called “MyCity,” designed to help local entrepreneurs navigate complex business regulations. Instead of helping, the bot confidently handed out illegal advice. It falsely told business owners they could take a cut of their workers’ tips, fire employees who complained about sexual harassment, and discriminate against tenants based on their source of income.

A Crisis of Executive Confidence

Credit: ZME Science/AI-generated NanoBanana 2.

While executives struggle to make AI work, they are simultaneously fighting fires on multiple global fronts.

CEOs are significantly less confident about their company’s short-term growth outlook compared to last year. Only 30 percent feel very or extremely confident about their revenue growth over the next 12 months. This is a sharp decline from 38 percent in last year’s survey, and a steep drop from the recent peak of 56 percent in 2022.

Cyber risks and macroeconomic volatility now tie as the top threats keeping executives awake at night. Nearly a third of CEOs report that their companies are highly or extremely exposed to the risk of significant financial loss from cyber threats this year.

Then there is the unpredictable nature of global trade. With governments increasingly using tax policy to secure supply chains, tariffs have become a major headache.

Almost a third of corporate leaders globally expect tariffs to reduce their net profit margins in the coming year. In the United States, 22 percent of CEOs say their corporations face high or extreme exposure to tariff risks.

Escaping the Short-Term Trap

So, corporations now face both technological uncertainty and geopolitical chaos as major challenges. Overcoming these hurdles requires breaking old patterns and thinking outside the box. Four in ten CEOs report that their organizations have started competing in completely new sectors over the last five years. This cross-pollination seems to pay off. Companies that generate a higher percentage of their revenue from new sectors enjoy bigger profit margins.

“The companies that succeed will be those willing to make bold decisions and invest with conviction in the capabilities that matter most,” Kande told Business Insider.

Yet, human psychology often works against this kind of bold, long-term thinking.

When humans face complex, immediate threats, we tend to develop tunnel vision. Corporate leaders are no different. CEOs report spending nearly half of their working time (47 percent) on issues with time horizons of less than one year. They dedicate a mere 16 percent of their schedule to activities focusing on the next five years or beyond.

This creates a paradox. The leaders who must guide massive, multi-year technological transitions — like overhauling corporate infrastructure for AI — are trapped fighting daily fires.

Until organizations can align the staggering capital costs of this technology with fundamental, long-term changes to their human workflows, the chasm between Silicon Valley’s soaring hype and the harsh reality of the balance sheet will only continue to widen.