AI Bubble Or Not? Where Opportunities Could Lie For Founders
By Joash Lee, Forbes Councils Member.
As Published on Forbes: Sep 03, 2026, 08:00am EDT
Joash Lee is the Founder and CEO of Sedifly, a global EdTech firm democratizing access to education.
With SpaceX’s record-breaking initial public offering (IPO) in mid-June and frontier labs such as OpenAI and Anthropic edging toward public listings, the question, “Are we in an AI bubble?” has returned with full force. (Full disclosure: I’m an early investor in SpaceX.)
The infrastructure layer—chips, cloud giants and frontier model labs—is posting eye-watering numbers while showing signs of froth. The application layer—tools built for specific industries and workflows—is moving slow but sure. Amid the hype, many investors and founders might be asking, “Where do the real opportunities lie?”
Infrastructure: Real Money, Big Bets, Lots Of Hype
Nvidia reported full-year fiscal 2026 revenue of $215.9 billion, up 65% year over year, with data center sales alone hitting roughly $62 billion in the fourth quarter. And the pace has since accelerated: For the first quarter of fiscal 2027, revenue reached $81.6 billion, up 85% from a year earlier.
On the private-market side, AI captured more than half of all global venture funding in 2025. In the first quarter of 2026, a handful of large deals accounted for the majority of capital deployed. Concentration at this scale can create circular dynamics: Large players invest in or commit spend to startups that, in turn, buy their chips and cloud capacity. SpaceX listed this June at a $1.77 trillion valuation, pricing in years of future dominance. Even as its shares slid, some believers still load their boats with the conviction it’ll be a multigenerational company in years to come.
Valuation multiples today are elevated but not at dot-com extremes. The Nasdaq-100 traded at forward price-to-earnings (P/E) ratios around 60 times at the 2000 peak, while today’s forward P/E sits at 23 times, less than half that level, according to IntuitionLabs. Trailing multiples are higher, but earnings quality is also stronger: Nasdaq-100 index companies are seeing strong earnings and growth forecasts, unlike the ‘90s era when “nearly three-fourths of the Nasdaq-100 traded either at a P/E north of 60 or was unprofitable,” according to Nasdaq.
The structural risk is not zero earnings, but the gap between infrastructure spend and realized enterprise ROI. A widely cited MIT NANDA report found that 95% of generative AI pilots fail to deliver measurable profit-and-loss impact. McKinsey’s 2025 survey similarly shows most organizations still in pilot mode, with only about 37% attributing any EBIT impact to AI; for most of those, the contribution remains under 5%.
If enterprise ROI continues to lag while capital expenditure soars, the market will eventually have to answer who pays for the next phase of build-out. I believe that tension, more than a single valuation multiple, is what makes the infrastructure layer feel bubbly at times.
Applications: Deep Expertise, Seamless Fit, Peace Of Mind
In my view, the applications layer has upside potential. In regulated, high-stakes industries including law, finance, healthcare and education, customers are not buying another chatbot. They are buying reduced risk and measurable outcomes.
General-purpose models improve every quarter, but they often struggle with industry-specific rules, proprietary data, explainability requirements and liability. The winners in these verticals tend to share a few traits: They embed deeply into existing workflows, layer on domain expertise and proprietary data that general models cannot easily replicate, deliver differentiated outcomes, and build economics that support strong margins with relatively modest capital.
My sector, education, is a textbook case. College admissions is a high-stakes decision with clear, measurable results and heavy parental demand for accountability. While generic tools might be able to draft essays, they cannot replicate the judgment, relationships and outcomes that come from years of experience. Parents and students are effectively buying “peace of mind” and shifted risk, not software licenses.
The same logic applies in other verticals. Tools that sit at the most consequential step of a professional workflow, take on liability through contracts and compliance, and create dependency through integration tend to build real moats.
What This Means For Founders
I believe we are simultaneously in a build-out super-cycle at the infrastructure layer and the early stages of an application-layer industrial shift. The former will likely see continued expansion punctuated by corrections as IPOs test narrative premiums and as enterprises demand clearer ROI. The latter can compound more steadily because it is tied to specific, already-budgeted workflows and measurable business outcomes.
From my vantage point, the current concentration of capital and private valuations for frontier labs do carry bubble-like characteristics, especially with the wave of IPOs priced on future stories. However, unlike 2000, the core technology is delivering growth for leaders and founders, and the productivity gains from better models and cheaper inference are real and continuing.
In my view, the bigger opportunity lies in startups that take these powerful models and embed them into regulated, data-rich vertical workflows where generic AI falls short.
For founders, the signal is clear: Chase depth over breadth. Build companies that solve painful, regulated or high-stakes problems inside existing industry software and processes. The infrastructure wave will keep delivering powerful, cheaper tools; the companies that turn those tools into trusted workflow engines inside specific verticals are the ones positioned to weather the inevitable cycles.
The bubble question is really twofold. At the foundation and infrastructure layer, parts of the market are pricing in perfection and will likely face digestion. At the application layer—where AI meets workflows, data and accountability—the more interesting and durable chapter is just unfolding.
The information provided here is not investment, tax or financial advice. You should consult with a licensed professional for advice concerning your specific situation.
View the article on Forbes here.
Comments