Echoes of the Internet Boom
The current artificial intelligence surge bears a striking resemblance to the early internet era, and that parallel deserves more attention than it typically receives. In the dot-com years, the perceived winners were the companies that simply provided access to the web or offered the first take on it — names like Netscape and Yahoo dominated the imagination of investors. History reminds us how that story ended: Netscape vanished as a meaningful entity, while Yahoo struggled to maintain its early lead. The enduring giants turned out to be the companies that built durable products on top of the new infrastructure rather than those that merely supplied the infrastructure itself.
We are very likely still in the early innings of the AI arms race, and the same pattern may be unfolding now. Despite enormous capital investment, AI as an industry is not actually achieving meaningful differentiation among its major model providers. When products fail to differentiate at the foundational layer, the economic value tends to migrate downstream — to the companies that use AI to build something useful, rather than to those producing the AI itself.
The Elephant Under the Rug
There is a critical issue that almost no one is willing to discuss openly: the staggering and growing cost of producing AI. Each new generation of models consumes vastly more tokens than its predecessors, and the cost curves are moving in the wrong direction. For now, this reality is obscured because AI companies are willing to absorb enormous losses in exchange for market share — a classic land-grab strategy.
But that pricing dynamic cannot last forever. Eventually, AI services will have to be priced at something approximating their true cost of production. When that happens, the burden will land squarely on users, who will be forced to fundamentally rethink how they deploy these tools. Optimization will become essential. Sophisticated users will need to adopt a portfolio approach — applying AI selectively to the problems where it offers the most leverage, while handling other tasks through traditional rules, processes, and structured workflows. The era of treating AI as a free or near-free resource is approaching its end.
The Bifurcation Investors Are Missing
Capital markets have rewarded the obvious AI heavyweights — chipmakers and hyperscalers building the underlying infrastructure — while applying a much narrower lens to the broader software universe. The dominance of the AI theme is dampening the nuance with which other companies can be perceived. Investors often divide the world into a simple binary: AI versus not-AI.
That binary misses a vital category. There is going to be software inside the AI stack, and some software companies are actually being hyper-charged by the presence of AI rather than threatened by it. Recognizing that distinction is one of the more important analytical tasks ahead for investors looking for second- and third-tier opportunities. Strong fundamental performance — accelerating cloud growth, expanded buyback programs, robust forward earnings projections — can easily get drowned out when the conversation reduces everything to whether a company is or isn't an AI pure-play.
The Success Gap in Execution
Perhaps the most revealing data point about the current moment is the chasm between aspiration and execution within enterprises. Surveys indicate that essentially 100% of organizations are now engaged with AI in some form — they are researching, piloting, experimenting, or running it in limited production. Yet fewer than 10% have deployed AI to genuinely strategic functions.
The pattern repeats across nearly every dimension of responsible AI use. Approximately 92% of leaders acknowledge that AI requires guardrails, but only a small fraction have actually implemented them. Most leaders understand that AI must be connected to a defined business process, but only about 18% have achieved that integration. Organizations know what they need to do, but they are slow to do it — and until they do, they cannot responsibly entrust AI with consequential decisions.
The reason for this hesitation is not technical. It is psychological and organizational. While AI is exciting on the investor side of the table, on the user side there is fear. Companies are feeling their way through unfamiliar territory, looking for examples to follow, and waiting for someone else to go first. As a result, the vast majority of AI deployments today are confined to purely tactical behaviors rather than strategic transformations.
Labor and the Reshaping of Work
The implications for the labor market are real but more nuanced than the headlines suggest. AI will undoubtedly reorganize work. Recent corporate announcements — including significant job cuts tied to AI optimization at major technology firms — are early signals of that restructuring.
Counterintuitively, recent college graduates are probably better positioned than commonly believed. They have plenty of time to adapt, and their flexibility is an asset. The greater difficulty falls on those already established in careers that are about to undergo fundamental change. Retraining at mid-career is a genuine displacement, and those costs should not be minimized.
Taken in aggregate, however, AI is likely to create roughly as many jobs as it eliminates, and the new ones should be better — more elevated, more human, more reliant on judgment. The trajectory is constructive, but the path will involve real disruption along the way.
Where the Winners Will Be Found
If the obvious play is not the right play, where should attention focus? The most promising candidates are companies using AI to accomplish things that simply could not be done before. The interesting frontier lies in handling the most error-intolerant, highest-value tasks — the work where reliability matters more than novelty.
Doing that requires structure. AI on its own cannot be trusted with mission-critical decisions; it needs surrounding software that provides guardrails, process integration, and accountability. The companies that build the connective tissue allowing AI to operate safely in domains it otherwise could not enter are positioned to capture meaningful value as the broader market matures past its current infatuation with model providers.
A New Frontier
We are living through a remarkable convergence of transformative forces — artificial intelligence, a renewed space race, the resurgence of nuclear technology — and the pace of change is unlikely to slow. The investors and operators who will navigate this period most successfully are those willing to look past the obvious narratives, take seriously the unspoken economic realities of AI production, and identify the businesses applying these tools to do genuinely new things rather than simply layering them onto familiar ones.