Opinion
Founders keep asking me if this is a bubble. That's the wrong question.
Market cycles rarely result in total collapse or complete stability. What fails is tolerance for unclear value propositions and unproven timelines. What endures is genuine utility, clear monetization, and teams with a precise understanding of the problems they address.
Every few weeks, a founder asks me some version of it. Sometimes it's direct - "Do you think this is sustainable?" Sometimes it's dressed up as strategy - "How are you thinking about the macro environment?" But the underlying question is always the same: Is this a bubble?
This concern is understandable. The dot-com era had a lasting impact on the ecosystem. In Israel, many companies with strong technology and teams failed when capital conditions changed more rapidly than they could adapt. For many, this was more than a market correction; it was a personal loss.
So the question comes from a real place. But it's still the wrong one.
The data presents a mixed picture. The dot-com Nasdaq reached a forward P/E above 60, while today's S&P 500 is around 23-elevated, but not at 1999 levels. Notably, AI stock P/E ratios have declined from about 40 in 2023 to approximately 34 by the end of 2025, despite significant price increases (Amundi Investment Institute, April 2026). In contrast, TMT valuations during the dot-com era expanded rapidly until the crash. This difference is significant.
The risk is real, but it is not where most expect. A small group of AI-related stocks now accounts for over 30 percent of the S&P 500, concentrated in about 45 companies (Amundi Investment Institute, April 2026). At the dot-com peak, nearly 100 TMT stocks held similar weight. The current AI rally is built on a narrower base, so any correction may be sharper and faster than headline figures indicate. This concentration risk is critical to understand.
Capital flows reflect a similar trend. AI now attracts 61 percent of global venture capital, a sectoral concentration not seen in 1999. However, most of this investment targets companies with real revenue growth. At the 2000 peak, only 14 percent of Nasdaq-listed companies were profitable. In contrast, today's leading AI infrastructure firms report substantial earnings. For example, NVIDIA generated $216 billion in revenue last fiscal year with a 53 percent net margin. Such performance was absent among dot-com companies in 2000.
What's different this time is the layer of real deployment underneath the hype. Enterprise adoption of generative AI has reached 71 percent of organizations using it in at least one business function. That's the floor, not the ceiling. It measures systems already in production rather than intentions or pilot programs (IntuitionLabs, 2026). We're already seeing companies build in verticalized applications, agentic infrastructure, and security-adjacent tooling where the local talent base is genuinely differentiated. These are bets on durable structural advantages.
However, some elements of AI investment mirror 1999. Some startups are valued between $400 million and $1.2 billion per employee (Reuters, 2025), an unprecedented ratio. Founders are raising capital based on platform narratives before proving sustainable, profitable value. While Nvidia is profitable, it is unclear whether others will achieve similar margins.
The main forward-looking risk is execution rather than valuation. As of early 2026, AI portfolio companies are investing in capital expenditures at twice the rate of non-AI firms (Amundi Investment Institute, April 2026). The critical question is whether revenue and gross margins will justify these investments within the expected timelines.
In my experience, the most innovative founders focus not on market cycles, but on whether their specific strategies are sound, regardless of broader trends. This distinction is important.
The dot-com era produced infrastructure that lasted beyond the crash. Surviving companies built meaningful products, independent of short-term market valuations. AI is following a similar trajectory, but at a faster pace. Foundation models are established, cost structures are improving, and enterprise adoption is significant.
Viewing the risk as a binary bubble oversimplifies the situation. Market cycles rarely result in total collapse or complete stability. What fails is tolerance for unclear value propositions and unproven timelines. What endures is genuine utility, clear monetization, and teams with a precise understanding of the problems they address.
This is the question that truly matters.
Moshe Zilberstein is a General Partner at N47.
