Are we living in an AI bubble? Applying lessons from the dot-com era

We’ve seen this economy before. Rapid innovation followed by unprecedented growth, all fueled by investors hungry for ROI. Remembering the dot-com era and the housing market of 07-’08, today many are wondering: Is today’s AI boom another bubble? And if so, is it vulnerable enough to pop soon?  

In 2025 alone, roughly two-thirds of U.S. venture capital flowed into AI-related companies, with most of that funding concentrated in a remarkably small number of firms. That level of focus amplifies upside, but it also amplifies risk. 

For CIOs, the more productive question isn’t “Will the bubble burst?” It’s “What happens to our enterprise if it does, or if growth simply slows?” The responsibility of IT leadership isn’t to predict market corrections but to build organizations that remain operationally sound regardless of how the market evolves.  

Why this isn’t the dot-com bubble, and why that matters 

Comparisons to the dot-com era are inevitable, and they’re not wrong, but they are often incomplete. In the late 1990s, hundreds of newly formed internet companies went public with little or no profitability, driven largely by speculative enthusiasm. When the market corrected, many of those companies vanished. 

Today’s AI ecosystem looks vastly different. While startups have earned funding, much of the investment is flowing to highly profitable, deeply entrenched enterprises that are household names, with strong revenue streams and global reach. The capital being deployed is also of a different nature. AI investment today is heavily tied to tangible assets, like data centers, specialized chips and infrastructure, rather than purely speculative growth. 

At the same time, the risk is more concentrated. A small number of vendors, platforms and models underpin an outsized portion of enterprise AI strategies. If valuations are correct, funding tightens, or consolidation accelerates, the downstream impact won’t be evenly distributed. 

That’s the real parallel to previous bubbles: not the technology itself, but the organizations that tied their futures too tightly to a single assumption — that growth would continue uninterrupted. 

Stop asking “Will it burst?” and start planning for volatility 

Market corrections don’t need to look like catastrophic crashes to be disruptive. Revenue growth can slow. Pricing models can change. Vendors can consolidate, pivot or disappear. In each case, enterprises that assumed stability feel the impact most acutely. 

The lesson from the dot-com era wasn’t that the internet failed; it was that preparedness mattered. Companies that could absorb the shock kept building. Companies that couldn’t keep up were set back years or were forced to close their doors. 

AI is no different. It isn’t going away. But volatility is inevitable. CIOs who treat AI strategy as a resilience exercise are better positioned to navigate what comes next. 

Best practice #1: Enforce disciplined ROI early and continuously 

One of the fastest ways enterprises expose themselves to bubble risk is by suspending normal investment discipline. AI is often framed as so transformational that traditional ROI frameworks no longer apply. That’s a mistake. 

Business cases still matter, especially when capital is concentrated, and vendor landscapes are evolving. ROI expectations may shift over time, but they must remain. Enterprises that require clear value hypotheses, adoption milestones and measurable outcomes — like every other initiative or purchase — insulate themselves from hype-driven overextension. 

Best practice #2: Diversify vendors, architectures and assumptions 

Diversification is a technology risk strategy. Enterprises that commit fully to a single AI vendor, platform or model assume they’ve correctly identified the long-term winner. History suggests that’s a risky bet. Vendor consolidation is likely. Pricing structures may change. Smaller innovators may be acquired or disappear. 

A resilient tech strategy spreads exposure across vendors, use cases and architectures. It avoids hard dependencies where possible and adds flexibility to the IT infrastructure. Multivendor strategies, modular integrations and data portability are practical defenses against market disruption. 

Best practice #3: Treat adoption and change management as first-class investments 

AI initiatives fail quietly more often than they fail publicly. Tools are deployed, pilots succeed and then adoption stalls. 

Change management is frequently underfunded relative to development, even though it determines whether AI delivers value at scale. Training, process redesign, communication and ongoing enablement require sustained investment. 

A useful rule of thumb is to match AI development spend with adoption and scaling spend. Without that balance, AI remains an underutilized asset, particularly vulnerable during periods of cost scrutiny. 

This is especially important as AI shifts from efficiency gains to workforce augmentation. Premature assumptions about labor replacement can erode morale, eliminate critical institutional knowledge and undermine long-term ROI. 

Best practice #4: Invest in people before you bet against them 

One of the riskiest assumptions in any AI strategy is that people can be removed early and replaced entirely by automation. In practice, the opposite is often true. 

AI systems rely on human expertise for training, tuning, oversight and contextual judgment. Organizations that cut too deeply, too early, often find themselves rebuilding the same teams they dismantled at a higher cost and lower morale. 

A more resilient approach treats AI as an amplifier of human capability rather than a substitute. Enterprises that invest in upskilling, experimentation and collaborative workflows tend to realize more durable value and avoid costly reversals. 

Best practice #5: Make due diligence continuous, not transactional 

In volatile markets, vendor due diligence can’t be a one-time exercise. Financial stability, data ownership, governance models and cybersecurity postures evolve quickly, especially when valuations fluctuate. 

CIOs should regularly reassess AI partners, asking not only “Is this technology sound?” but also “Is this company structurally positioned to endure change?” Contractual safeguards around data ownership, exit rights and continuity planning are essential, particularly when vendors play central roles in core operations. 

Security considerations are equally critical. As AI systems gain deeper access to enterprise data and processes, third- and fourth-party risk expands. Continuous assessment becomes the differentiator between manageable risk and systemic exposure. 

AI resilience is now a strategic imperative 

Enterprise AI implementation is increasingly tied to both national security and the growing geopolitical innovation race. Governments are investing heavily in the same large vendors that enterprises rely on, reinforcing consolidation dynamics and raising the stakes of dependency. 

As we explored in our last CIO article on AI and regulatory urgency, enterprises can’t afford to wait for clarity before acting. But acting doesn’t mean overcommitting. It means preparing for change. 

The dot-com bubble didn’t reverse the internet’s trajectory; it clarified it. Organizations that survived emerged stronger, more focused and poised for growth. 

The same will be true for AI. Market volatility will test assumptions. Prepared companies will absorb the shock. Those who are unprepared will feel it deeply. 

This article is published as part of the Foundry Expert Contributor Network.
Want to join?