There is a temptation in moments of technological change, to ask the wrong question.

The question is not whether artificial intelligence matters. It clearly does. AI is real, powerful, and already reshaping how work gets done across industries. The more interesting — and more dangerous — question is what happens when nearly all attention, capital, and imagination collapse into a single narrative at once.

That is not an AI problem.

It is a capital allocation problem.

The Warning Signal Isn’t AI — It’s Homogeneity

One of the most striking patterns we keep encountering recently is not the rise of AI‑enabled companies, but the near‑total uniformity of deal flow. When every pitch, every update, and every portfolio discussion leads with the same framing, something important has changed.

Five years ago, venture and growth markets reflected dispersion. Capital flowed across software, healthcare, industrials, hardware, logistics, defense, manufacturing, biotech, energy, and services. Cycles favored some sectors over others, but variety remained.

Today, the question increasingly is not whether a company uses AI — it is how prominently it can brand itself around it. When investors start seeing “100 out of 100 deals are AI,” that is not evidence of superior opportunity density. It is evidence of attention collapse — the tendency for time, capital, and mental bandwidth to concentrate around a single idea at the expense of everything else.

Recent data suggest the shift is not just anecdotal: AI-related companies captured roughly half of all global venture funding in 2025 — up from one-third in 2024 and just over a quarter in 2023.

And attention collapse has consequences.

Where Did All the Other Deals Go?

A few years ago, “what are you seeing?” across venture and growth meant a map of the real economy: vertical SaaS, healthcare IT, manufacturing tech, CPG brands, logistics platforms, industrial automation, fintech, and an assortment of services. AI showed up, but as one ingredient among many.

Today, the same question often gets a very different answer. In more than one recent annual meeting, we have heard some version of: “Right now, essentially 100 out of 100 deals we see are AI.” When we ask what that mix looked like three to five years ago, the response is usually that the firm has “evolved its thesis” to where the market is going. On paper, a manager that once differentiated through sector breadth, niche sourcing, and idiosyncratic opportunities is now described in marketing language as an AI‑focused platform, or a fund “leaning heavily into AI” for the new vintage.

The part that bothers us is not the interest in AI. It is the casual assumption that a near‑total re‑labeling of deal flow is normal.

Because here is the real question: if every deal is now “AI,” what happened to everything that used to not be?

The Starting Line Moved Up a Layer

We do not think those businesses disappeared. Healthcare, CPG, logistics, manufacturing, SaaS, and services are still out there. Many of them now describe themselves as “AI‑powered” or “AI‑first,” sometimes for real strategic reasons, sometimes for a valuation bump.

The risk is not that a manufacturing company uses AI in its routing or forecasting. The risk is treating it as if it suddenly became an AI company and paying AI‑company prices for manufacturing‑company economics. At the end of the day, it is still a factory, a logistics network, or a niche software vendor, with all the cyclicality, margins, and constraints that implies. We do not blame teams for putting “AI” on the deck; incentives push them to do exactly that. But if investors fail to bifurcate “AI‑enabled” from “AI‑native,” pricing discipline breaks down. We do not just mis‑label; we mis‑price — and that is where the slope turns dangerous.

What has changed is the hierarchy.

A few years ago, the “tip of the spear” for categorizing deals was sector:

• Healthcare

• Financials

• CPG and consumer

• Industrial and manufacturing

• Vertical SaaS

• Logistics, energy, and so on

Today, the first sort key is no longer the underlying industry. It is where the company sits on an AI spectrum:

• Pure‑play AI: AI is the product; without the model, there is no business.

• AI‑enabled: Existing sector economics with an AI layer on top (for example, a vertical SaaS product or a healthcare workflow tool with AI copilots).

• AI‑adjacent: Picks and shovels — data, compute, infra, tooling, integration, governance.

• Non‑AI: Businesses that create value without relying on AI as a central pillar.

Only after that sort does the old taxonomy reappear: AI‑enabled healthcare vs AI‑enabled logistics vs AI‑enabled CPG, and so on. The starting line moved up a layer, from “what sector is this?” to “how AI‑centric is this?”, and then back down into sector detail.

AI stopped being a vertical and became a primary organizing principle.

This is what attention collapse looks like in practice: not just more AI deals, but a classification system in which everything must pass through an “AI or not?” gateway before anyone asks what the business actually is.

When Capital Chases Narrative, Not Scarcity

Markets function best when capital is allocated toward scarcity — unmet needs, underexplored assets, misunderstood structures. But when capital follows narrative instead of scarcity, distortions emerge.

Three things tend to happen at once:

• Overfunding of the visible

  • Capital floods into companies that fit the dominant story, compressing forward returns and inflating expectations faster than fundamentals can justify.

• Underfunding of the invisible -

  • Entire categories — industrial processes, physical infrastructure, boring cash‑flow businesses, non‑AI services — receive less attention not because they disappeared, but because they no longer sound exciting.

• False signals of risk and safety

  • Narrative alignment begins to masquerade as quality, while deviation from the story begins to feel risky even when the underlying economics are sound.

This is how bubbles form — not necessarily because the story is wrong, but because too much capital arrives too quickly, with too little discrimination. The danger is less “this thesis is false” and more “this thesis is now fully priced — and then some.”

Professional Herds and the Illusion of Sophistication

Herd behavior is not a retail phenomenon. It is a human one.

Markets like to tell themselves that excess lives on Main Street — in meme stocks, day traders, and cocktail‑party tips. Institutional investors prefer to believe they are immune, guided by models, committees, and process. History suggests otherwise.

The mechanism is familiar. A compelling narrative emerges. Early participants are rewarded. Success becomes visible. Skepticism gives way to participation. Participation becomes expectation. Eventually, not participating begins to feel irresponsible.

At that point, the herd is already formed — even if everyone in it believes they are acting independently.

What makes modern cycles more dangerous is not ignorance, but sophistication layered on top of herd behavior. Governance structures, benchmarking, and “career risk” all push professionals toward what everyone else already owns, especially when performance spreads widen. When consensus forms among investors who speak in probabilities, cite research, and manage other people’s money, it becomes harder to recognize it for what it is.

When “everyone sees it,” the risk is no longer misunderstanding the opportunity.

The risk is misunderstanding crowding.

When the Cab Driver Isn’t the Signal Anymore

The old trope was simple: when the cab driver gives you stock tips, it is time to sell. That heuristic worked when information diffused slowly and enthusiasm spilled outward.

Today, the echo chamber often forms inward.

By the time themes dominate retail headlines, they have usually already passed through institutional channels — pitch decks, annual meetings, strategy memos, portfolio updates. When nearly every professional conversation centers on the same framing, the cab driver test has already failed — because the crowd has moved upstream.

That is what makes monocultures dangerous. Not that the story is false, but that it becomes unchallenged.

Exponential Change Does Not Cancel Cycles

One of the most under-appreciated mistakes in modern investing is assuming that technological inevitability negates market cycles.

Exponential systems do break linear intuition. Progress compounds quietly, then suddenly. This framework is essential to understanding AI’s long‑term impact.

But exponential progress in technology does not eliminate cyclical behavior in capital.

Innovation can accelerate while valuations overshoot. Adoption can compound while returns compress. Entire categories can be transformed while capital is still misallocated within them.

In other words, the world can change rapidly while markets still behave the same way they always have.

Believing that “this time is different” because the technology is real is how investors confuse inevitability with investability.

How We Participate — Without Surrendering Discipline

We are not observers. We are participants. But we participate with structure.

That means:

• Favoring picks‑and‑shovels over point solutions

• Owning toll roads rather than guessing which product wins

• Making small, asymmetric bets instead of concentrated narrative exposure

• Preferring fee‑based or cash‑flowing economics where possible

• Avoiding dependence on a single technological assumption

Participation does not require maximal exposure.

It requires intentional exposure.

Just as importantly, we track ourselves.

We pay attention to the meetings we take, the ideas that recur, and the language that creeps into conversation. Theme drift and style drift are signals. When every discussion begins to sound the same, it is not confirmation — it is information.

Why This Matters for GPs and LPs

If the industry’s shared rubric has silently shifted from “healthcare vs manufacturing vs SaaS” to “AI‑native vs AI‑enabled vs AI‑adjacent vs non‑AI,” then investors who do not make their own rubric explicit risk inheriting the market’s blind spots.

Two things follow:

• First, the map can hide the territory. If every pitch, memo, and update is framed primarily as an AI bet, it becomes harder to see that we may actually be overweight a very narrow slice of the real economy (for example, AI‑enabled software) and underweight less fashionable, but durable, cash‑flow producers.

• Second, style drift can look like progress. A GP “shifting thesis” from sector‑specialist to AI‑generalist can be rational on its face, but from an LP’s perspective it may mean the thing we thought we owned — distinct sourcing, differentiated sectors, idiosyncratic deal mix — is quietly converging toward the same homogenous pipeline everyone else is now chasing.

What makes this more worrying is how casually it is often presented. Style drift is treated as sophistication, not as a change in risk. Everyone stays cool, calm, and fluent in the new vocabulary; almost no one says, “We used to do X, now we mostly do Y, and here is the cost of that shift.” In practice, nearly every manager we speak with is “leaning into AI” in some form, often after reading the same posts and recycling the same talking points. The performance of expertise becomes a substitute for lived domain edge in the sectors these companies actually operate in. That gap may not show up in a single fund; it tends to surface over cycles.

As allocators, we therefore try to maintain a deliberate hierarchy of our own. We do not stop at “is this AI?” We ask:

1. Where on the AI spectrum does this company really sit (pure‑play, enabled, adjacent, non‑AI)?

2. Within that, what is the underlying economic engine (healthcare services, industrial automation, CPG brand, infrastructure, and so on)?

3. Do we already have crowding at this combined layer (for example, multiple AI‑enabled SaaS tools selling into the same budget)?

Institutional advantage is not predicting the peak.

It is recognizing when dispersion is shrinking.

Where the Opportunity Often Lives

Periods of narrative dominance are paradoxical. They feel crowded, but they also create vacuum zones — areas starved of attention where fundamentals quietly improve.

These environments tend to favor:

• Businesses with real cash flow but low narrative appeal

• Structures that return capital rather than promise it

• Industries where progress is incremental, not exponential

• Assets whose value is economic, not conceptual

These opportunities rarely announce themselves loudly. They tend to surface only after attention rotates and the story loses its grip.

Allocating Capital Beyond the Story

The greatest risks in investing rarely come from betting on the wrong story. They come from betting on the same story as everyone else, at the same time, for the same reasons.

When capital stops asking “what is scarce?” and starts asking “what is trending?”, mispricing follows.

Our responsibility is not to reject progress, nor to worship it. It is to remain one step removed from collective certainty: aware of where we are in the cycle, how capital is behaving, and whether attention is creating blind spots.

Technology will continue to advance exponentially.

Capital will continue to move cyclically.

The edge lives in understanding the difference — and in allocating accordingly.