The problem with the bubble theory is that customers keep showing up with real money, writes Dr. Gleb Tsipursky.
NVIDIA’s latest financial results put a hard stand on this argument, with record quarterly revenue of $81.6 billion and data center revenue of $75.2 billion driven by an AI investment boom. These numbers do not prove that all AI stocks are reasonably priced or that all company pilots will be profitable. These studies prove that the demand for AI infrastructure is not a social media hallucination, a boardroom fad, or a demonstration-driven burst of enthusiasm.
Smarter conclusions are more nuanced and more useful to executives. The AI bubble hypothesis explains glut, valuation anxiety, and copycat spending. It does not account for the scale of adoption, capital formation, or the evidence of new productivity that we now see across the economy.
Demand acts like infrastructure, not hype
Speculative bubbles tend to float above reality. AI is becoming more prevalent there. The US Federal Reserve (FRB) survey of corporate AI adoption in 2026 found that approximately 18% of US companies will have adopted AI by the end of 2025, and work-related generative AI use by individuals will have reached nearly 41%. Another senior leader survey cited in the same analysis estimated that 78% of the workforce works for companies that have implemented AI.
This width is important because bubbles depend on a widening gap between belief and use. AI is moving in the opposite direction. stanford university 2026 AI Index Report Enterprise investment, frontier model returns, cloud capital spending, and consumer value are all rising, indicating the ecosystem is moving from experimentation to construction.
The most important signal is not a single company’s earnings. This is an alignment between chip demand, cloud expansion, model deployment, enterprise experimentation, and employee usage. Spending on AI infrastructure looks more like the early stages of an industrial platform than Tulipmania. Railroads, electricity, broadband, and cloud computing all required large upfront investments before the most valuable applications became apparent. The AI follows that pattern, but at a faster pace and more erratic.
This difference should make leaders feel more confident and less reckless about investing in AI. The key is not to chase every shiny model or vendor claim. The key is to recognize that the foundation is becoming durable enough to warrant serious operational effort.
The real risk is not the AI itself but the weak implementation.
The best argument for being cautious is not that AI isn’t valuable, but that many organizations remain elusive. M.I.T. The state of AI in business in 2025 Our research on generative AI ROI highlights the significant difference between extensive pilots and successful production environments, especially for custom enterprise tools. This discovery should sober executives, but they shouldn’t freeze.
The lessons are not technical, but managerial. Companies that paste chatbots into broken workflows are usually a novelty. It leverages companies that redesign processes, connect AI to their own data, create accountability, and measure performance.
In McKinsey’s 2025 Global Survey, 88% of respondents reported regularly using AI in at least one business function, but only 39% reported an impact on company-level EBIT. The same study found that high performers are much more likely to redesign workflows, incorporate AI into business processes, track KPIs, and invest in changes to people, data, and operating models. In short, the business value of AI is not magic. It is being built.
This should further increase management’s trust. You will see a pattern of failure. A pattern of success is also emerging. Leaders don’t have to blindly trust AI. They need the discipline to stop funding orphan pilots and start supporting fewer, better programs tied to revenue growth, margin expansion, customer experience, risk reduction, and speed.
It also means changing the conversation around investing. A true AI leadership strategy starts with business architecture, not software procurement. The winner will not be the company with the longest list of AI tools. They will be the companies that reimagine the way work is done.
Confident leaders manage reinforcement, not avoid it.
Discussions about bubbles often tempt executives into the false choice of enthusiasm or paralysis. Neither is appropriate. While AI itself will become essential, the AI market may contain excess. The Internet gave rise to both Pets.com and Amazon. Cloud computing has created both wasteful migrations and lasting operational benefits. The same schism is already showing up in AI.
PwC Global AI Employment Barometer in 2025 Give leaders another reason to stay involved. An analysis of nearly 1 billion job advertisements found that after the adoption of generative AI, productivity grew nearly four times in industries with the most exposure to AI, and revenue per employee grew three times faster in industries with the most exposure to AI than in those with the least. These AI productivity gains are early, uneven, and still controversial, but they are too important to ignore.
The appropriate response is controlled acceleration. Boards should ask about use case economics, data readiness plans, cybersecurity controls, vendor-focused reviews, model risk protocols, and employee recruitment metrics. Finance teams need to request milestones. Business leaders need to own the results. Technology teams need to build reusable platforms, not one-off demos.
In this way, capital investments in AI become strategic rather than theatrical investments. The goal is not to spend because your competitors are spending, but to build capabilities that combine better data pipelines, faster product cycles, more responsive service operations, stronger forecasting, lower error rates, higher employee utilization, and more.
This is also why waiting for complete clarity is dangerous. By the time all AI use cases have clean benchmarks and all evaluation concerns are resolved, leading companies will have already redesigned their workflows, trained their teams, negotiated access to infrastructure, and learned from their mistakes. AI transformation rewards cumulative learning, but cumulative learning takes time.
conclusion
The AI bubble theory is not stupid. Markets can overestimate actual revolutions, and managers should be distrustful of strategies built on fear of missing out. But stronger evidence now points to something bigger than the hype: real adoption, real infrastructure demand, real productivity signs, and a widening gap between organizations casually experimenting and those seriously implementing.
AI should give leaders more confidence because investment deals are no longer based on imagination alone. It is increasingly built on operational evidence. The mission is not to spend blindly. It’s about investing with conviction, governing with rigor, and acting fast enough to learn before competitors turn AI from experimentation to advantage.
Dr. Gleb Chipursky, nicknamed the “office whisperer” new york timeshelps leaders turn AI hype into real-world results. He is the CEO of Disaster Avoidance Expert, a future of work consultancy, and the author of seven bestselling books, including: The psychology of generative AI adoption.
Want more articles from trusted news sources?Make HR Leader your preferred news source on Google.
Dr. Gleb Chipursky
Dr. Gleb Tsipursky, nicknamed the “office whisperer” new york timeshelps leaders turn AI hype into real-world results. He is the CEO of Disaster Avoidance Experts, a future of work consulting firm, and the author of seven bestselling books, including: The psychology of generative AI adoption.
