Artificial intelligence has entered a phase of rapid technological advancement, widespread experimentation, and unprecedented levels of capital deployment. Productivity gains are becoming visible across industries, from software development to healthcare and logistics. However, despite these encouraging signals, a fundamental disconnect has emerged between the potential of AI and the valuations investors are currently placing on companies operating in the space.
This article explores why AI-driven productivity does not necessarily justify soaring valuations, how different forms of investor behavior and FOMO are shaping markets, and why the eventual winners may not be the ones currently valued the highest. More importantly, it examines the layers of the ecosystem to determine where value is likely to accumulate and how investors are mitigating uncertainty through diversification.
Productivity Growth Does Not Automatically Justify Valuations
AI is widely expected to contribute to productivity growth at the macroeconomic level. Many experts anticipate annual productivity gains of around 1–1.5%, which would be at the upper end of historical improvements. Such gains would represent a transformative boost for developed economies that have struggled with stagnation since the mid-2000s.
However, even if those optimistic productivity outcomes materialize, the question remains: Can the profits of AI companies scale in a way that justifies their current valuations?
High Investment vs. Modest Revenue
The problem at the heart of the valuation debate is the massive gap between investment and revenue:
• The largest AI firms are deploying trillions of dollars in collective investment for compute, infrastructure, research, and model training.
• Yet the aggregate revenue from commercial AI products remains in the tens of billions, a gap too large to ignore.
For valuations to hold, investors must believe that:
1. AI companies can achieve sustainable and scalable monetization, and
2. Profit margins will eventually rise rather than compress under competitive pressure.
Competitive Pressure and Uncertain Revenue Models
The competitive dynamic adds another layer of difficulty. Users – both individual and enterprise – frequently switch between leading models such as GPT, Claude, or DeepSeek. This fluidity suggests:
• Low switching costs
• Weak vendor lock-in
• Commoditization risk for base models
Unlike cloud or enterprise software, where pricing and margins stabilized over time, AI revenue models remain unclear. Some companies rely on subscription access, others on usage billing, and many are experimenting with agent-based or embedded pricing models. Without clarity on how value will be captured, valuations become speculative rather than based on discounted cash flows.
Three Forms of FOMO Driving AI Investment Behavior
Several observers describe the AI boom not as a bubble, but as a cycle driven heavily by different forms of FOMO – “fear of missing out.” These behavioral forces are shaping how capital enters the sector and how valuations become inflated.
1. Capacity FOMO
Capacity FOMO centers around the race to acquire compute, GPUs, data centers, and model training infrastructure. The logic is that:
• Scarcity of compute resources provides temporary pricing power
• Model scale and performance correlate with training capacity
• Early movers gain strategic advantage
This dynamic has driven hyperscalers and model companies to raise unprecedented sums of capital in short time periods.
2. Application FOMO
Application FOMO captures the belief that AI will permeate nearly every industry, unlocking new business models and use cases. Investors expect that:
• Every enterprise will integrate AI into workflows
• Adoption will be broad and fast
• A future ecosystem of AI-native applications will emerge
This belief pushes money into startups building agents, copilots, and vertical AI platforms well before monetization proves durable.
3. Investment FOMO
Investment FOMO is more psychological and macro-level. Investors fear being late to a technological revolution of the same magnitude as:
• the Internet
• mobile computing
• cloud adoption
This is similar to how investors piled into early dot-com companies, although today’s AI giants are much more profitable and diversified. Still, the mismatch between timing of investment and timing of monetization remains a key valuation risk.
Breaking the AI Ecosystem Into Three Layers
To understand where valuations may be justified – and where bubbles may be forming – it is useful to break the ecosystem into three constituencies:
Layer A: Token Manufacturers (Core LLM Providers)
These are the companies producing tokens and base model output – effectively the raw compute and intelligence. They are:
• capital-intensive
• infrastructure-heavy
• increasingly commoditized
Examples include foundation model companies and hyperscalers.
Over time, commoditization forces may drive margins down. If one company fails, its capacity can be absorbed by others, similar to how semiconductor fabs absorb failed peers. Therefore, while these companies are currently achieving the highest valuations, they may not capture the majority of long-term value.
Layer B: Application & Ecosystem Layer
This layer consists of companies building:
• agents
• copilots
• workflows
• AI-native enterprise tools
• integrated automation platforms
This is where much of the true innovation occurs and where demand will likely be sticky. These companies benefit from:
• differentiated functionality
• vertical specialization
• integration into customer workflows
In the long run, this layer may capture more sustainable economic value than the raw model providers, even though current valuations may not fully reflect that future.
Layer C: End Users (Enterprise & Consumer)
The end users will ultimately compound the productivity gains. Unlike the dot-com era, where consumer adoption preceded enterprise productivity, AI offers immediate productivity benefits to:
• developers
• analysts
• designers
• researchers
• corporate back office functions
This could yield productivity increases of 1–5% annually, compounding over time. From a macroeconomic perspective, this is where the true value of AI manifests – not necessarily in the valuations of the companies that build the technology.
Historical Parallels and Timing Risk
Looking back at transformative technologies provides context. The comparison to the railroad build-out is instructive:
• Railroads transformed commerce and mobility
• Massive economic value was created for society
• Many railroad investors and firms failed financially
This is a case where economic value and investment returns diverged sharply. Investors often:
• overestimate short-term financial returns
• underestimate long-term societal impact
AI appears to be following a similar trajectory.
Diversification as a Rational Investment Strategy
Given valuation uncertainty and timing mismatch, investors are increasingly turning to diversification.
Geographic and Asset Allocation Diversification
Investors overweight in U.S. technology are now exploring:
• Europe
• private markets
• infrastructure assets
The incremental capital is not always flowing into the same category as the highest-flying AI names, reflecting caution.
Strategy Diversification Within the AI Sector
Within AI itself, investors are seeking decorrelated return streams, such as:
• data centers
• energy infrastructure
• airports
• communication assets
• mission-critical real estate
For example, a data center fully contracted for 15 years with reliable counterparties (e.g., hyperscalers or governments) may offer strong returns even if AI startup valuations correct.
Conclusion: AI Is Not a Monolith – And Neither Are Its Valuations
AI is real, adoption is accelerating, and productivity effects are likely to be significant and compounding. However, valuations are more uncertain because they depend on:
• sustainable monetization models
• competitive pressures
• timing of adoption
• capital intensity
• margin durability
The final picture may look like this:
• Economic value creation – near certain
• Productivity gains – already underway
• Investment returns – unevenly distributed
• Valuations – fragile for some segments, justified for others
Some layers of the ecosystem will become commoditized, others will extract premium value, and some investors will lose money despite AI’s positive impact on society.
The debate going forward will not be about whether AI is transformative – that part is settled – but rather which companies will actually capture the profits needed to justify their valuations.
