AI: Bubble, Breakout, or Both?

Written By Faraaz Khan

I’ve had countless conversations over the past year centered around whether AI is in a bubble or not. In the past week, three pieces of research caught my eye: the MarTech 2026 report, Vista’s AI Investment Primer, and Peter Moore’s article on AI adoption relative to innovation. Together, along with other nuanced research like the Gartner Hype Cycle for AI, you get a picture that’s more interesting than the usual “AI is overheated” or “AI will change everything.” The reality sits somewhere in the middle. We’re in a moment where hype, real traction, and uneven adoption are all happening at once.

This matters for operators, founders, and especially for companies that didn’t start as AI-native but now need to decide what to build, what to buy, and how fast to move.

1. The Hype Cycle Isn’t a Single Curve Anymore

Every conversation about AI seems to assume there is one big wave. But Gartner’s Hype Cycle makes one thing clear. We aren’t dealing with “AI” as a single phenomenon. We’re dealing with dozens of underlying technologies that are maturing at different speeds.

Some examples:

  • AI agents are sitting right at the peak, surrounded by inflated expectations.

  • ModelOps and AI engineering are sliding into the trough, where companies realize that running production AI is harder than demoing it.

  • Model distillation, knowledge graphs, and cloud AI services are climbing the slope, delivering steady value.

  • Foundation models and synthetic data are still early and evolving fast.

This nuance matters because it explains the tension we’re all feeling. People sense a bubble forming in parts of the market, yet they also see breakthroughs happening weekly. Both are true because we’re looking at different parts of the curve.

And there’s another layer. Gartner’s point that AI is not just GenAI is important. Predictive models, decisioning engines, non-generative machine learning, optimization algorithms, and graph techniques have been quietly powering real outcomes for years. They’re not in the same hype cycle as text or image models. Treating them as one macro trend misses the story.

2. Are Valuations Inflated? Yes, and No. It Depends Where You Look.

Some parts of the market clearly look stretched. But others don’t. What we’re seeing today is different from the dot-com boom in three ways.

First: AI companies can scale revenue with extremely small teams.

Lovable is the standout example. It reportedly hit $200M ARR in 12 months with fewer than 100 people. That level of capital efficiency didn’t exist in the early 2000s. Software distribution now moves much faster, and AI products compound value through user activity and feedback loops.

Cursor is another example of how quickly real traction can materialize. It raised at a $29.3B valuation with an implied ~30x revenue multiple, based on recent news reports. You can argue whether that multiple is high, but the underlying usage and value being created are hard to ignore.

Second: This wave moves faster than previous cycles.

There are three reasons for this speed:

  1. The infrastructure layer is already built. Startups can scale compute, storage, and distribution instantly.

  2. The developer ecosystem is global. Adoption happens simultaneously, not sequentially.

  3. AI compounds in real time. As models improve, the products built on them improve automatically.

The Vista AI Investment Landscape report shows this clearly. AI software valuations have grown faster than any prior enterprise software cycle, but unlike the dot-com years, much of the growth is coming from companies with real usage, strong early monetization, and deep engagement curves.

Third: Fragmentation is extreme.

Unlike the early internet era, where a few categories dominated, the MarTech 2026 report shows that AI is creating thousands of micro-categories, tools, and agents. Adoption is real, but production use remains low. Only 23% of companies report having agents in full production.

Fragmentation creates hype, but it also creates opportunity. Most companies won’t scale simply because distribution in AI (or just generally software) is hard. Many founding teams have strong technology but weak go-to-market. Some will fail not because of poor product quality but due to timing, positioning, or lack of sales infrastructure.

This dynamic sets the stage for one of the most interesting M&A environments we’ve seen in years.

3. Adoption Isn’t Keeping Up With Innovation

One of the most insightful pieces I came across recently was an article by Peter Moore titled “In the digital world, the speed of AI innovation is far exceeding the speed of its successful adoption.” His framing captures something most operators already feel but haven’t articulated. We’re seeing unprecedented leaps in AI capabilities, yet the ability of organizations to absorb these capabilities is lagging far behind.

Moore highlights three “speed gaps” that explain the disconnect:

The Innovation Curve Is Exploding Upward

New models, new agent frameworks, new toolchains, new capabilities. Every quarter resets the state of the art. But this creates a moving target for teams trying to place strategic bets. Something that looked cutting-edge last January can look dated by June.

The Adoption Curve Is Slow and Nonlinear

Enterprises don’t change overnight. They move through Moore’s “digital adoption life cycle” in phases:

  • Experiments

  • Pilots

  • Operationalization

  • Transformation

Most companies are stuck in the first two stages. They can prove concepts, but scaling them across workflows, functions, and governance structures takes years. It’s rarely a technology problem. Its alignment, change management, budgeting, data quality, and capability building.

The Value Curve Only Turns Up When Adoption Catches Up

Moore shows that value creation doesn’t track innovation. Value inflects only when adoption becomes broad and integrated, and we aren’t there yet for most organizations. Tools are improving faster than teams, processes, or data environments can support.

This reinforces the point from the MarTech 2026 report. Even with high experimentation, only a small percentage of companies have AI agents deployed in full production. And many cite the same friction:

  • fragmented stacks

  • poor data quality

  • unclear responsibilities

  • talent constraints

  • lack of repeatable workflows

This is where the hype cycle meets the real world. The technology is sprinting. Organizations are jogging. And the value curve hasn’t yet bent upward for most companies.

This gap isn’t a sign of a bubble. It’s a sign that adoption will be uneven and that capability building is becoming the real moat.


4. What This Means for M&A in AI

The narrative often swings between two extremes: AI startups are too expensive, or AI startups are a bargain given their potential. The truth is simpler.

A handful of breakout companies are definitely out of reach.

They have no incentive to sell. Their valuations are high, but they’re also producing real revenue with small teams and strong margins. These companies behave like early platform plays, not tools.

But the long tail is enormous, and many strong companies will not scale.

This is the heart of the MarTech 2026 report. Even with record adoption of AI tools:

  • only a minority reach sustained production usage

  • distribution remains the bottleneck

  • monetization is uneven

  • many brilliant engineering-led teams never find a commercial chassis

This is not failure. It’s fragmentation. And fragmentation is fertile ground for M&A.

Why now is strategically interesting for acquirers

  1. Talent is scarce. Many companies don’t have the internal capability to build agentic workflows, inference infrastructure, or context engineering systems. Acquiring teams accelerates learning curves.

  2. Workflows matter more than models. Companies that figured out context, orchestration, or data governance are far more valuable than companies training yet another model.

  3. AI is expensive to run. As McKinsey highlights, costs often exceed early ROI until companies optimize. Efficient teams with real production experience are worth a premium.

  4. R&D capital is tightening. Vista’s analysis shows that while valuations have expanded, capital is increasingly selective. Pressure will build on companies that raised early but haven’t found distribution.

  5. Strategic multiples are not the same as financial multiples. Paying 15x revenue for a standalone business may look aggressive. Paying 15x revenue for a capability that increases your core company’s valuation multiple by one turn (say, taking a $1B revenue business from 5x to 6x) is rational. You’re not buying the revenue. You’re buying the acceleration.

This is the real takeaway:

You don’t need to acquire the breakout winners. You need to identify the teams with strong technology, strong engineering talent, and weak distribution. Those are the ones who will take your platform further than you can take it on your own.

But this only works if you have:

  • a clear vision of your future AI architecture

  • a map of which capabilities you need to own

  • a sourcing engine that surfaces opportunities early

  • an evaluation process that separates hype from genuine acceleration

If the bubble cools, these teams become cheaper. If valuations keep rising, they may still be worth acquiring because of what they do to your core business.

Final Thoughts

We’re not in a simple boom or bust. We’re in a stretched spectrum. Some AI categories are overheated. Others are underpriced. Some companies are creating enormous real value. Others are creating noise.

The important question isn’t “Is AI a bubble?”  The important questions are:

(1) where on the cycle is the part of AI that matters to your business?

(2) what impact would bringing that capability in-house have on your core business?

(3) can you afford to be reactive instead of proactive in the current environment?

If you want to play offense in this market, clarity matters more than speed.

If you’re building, buying, or operating in this space, I’d love to compare notes.

You can reach me at faraaz@inorganicedge.com or on LinkedIn.


Sources

Faraaz Khan
Previous

AI-Ready M&A: How Acquirers Should Evaluate AI Compatibility
Next

How to Build a Repeatable M&A Engine That Increases Enterprise Value for SaaS