Minerva Investment Research Note | September 2026
Executive Summary
The investment discourse surrounding artificial intelligence is entering a critical new phase.
Initially, the defining question was: Will AI succeed?
Subsequently, the debate shifted to: Can such massive AI capital expenditures eventually generate sufficient returns?
However, as AI transitions from a technological breakthrough into core infrastructure, enterprise software, and structural productivity enhancements, a deeper and more consequential question emerges:
Even if AI succeeds spectacularly, who will ultimately capture the resulting economic value?
This is not a semantic distinction; it is a fundamental inquiry that directly dictates capital returns.
Technological progress can elevate economy-wide productivity without ensuring that technology providers capture all the newly minted value indefinitely. AI costs can relentlessly decline, adoption can become ubiquitous, and consumers and businesses can reap massive efficiency gains—all while intensifying industry competition compresses excess corporate profits.
Therefore, we believe rigorous AI investing requires distinguishing among three distinct layers:
- Economic Value: The aggregate value AI creates for the broader economy.
- Economic Rent: The fraction of that value that persistently accrues to specific companies or industry segments.
- Investor Return: Whether current market valuations already fully reflect these future economic rents.
The question truly worth answering is not merely how much value AI can create.
Rather:
Who can retain this value within their capital returns, and how long can that competitive advantage be sustained?
01 | From « Will AI Succeed? » to « Who Reaps the Rewards of Success? »
The AI investment narrative over the past two years actually encompasses four distinct questions.
- The Technical Layer: Will AI capabilities continue to advance at an exponential rate?
- The Macroeconomic Layer: Will AI successfully raise total factor productivity and drive potential long-term growth?
- The Corporate Layer: Can individual enterprises leverage AI to slash costs, expand revenue, or architect novel business models?
- The Investment Layer: How much of the aforementioned value ultimately translates into durable shareholder returns?
These four layers are frequently conflated, yet they are fundamentally distinct.
A technology can achieve wild success. An entire economy can experience a productivity surge because of it. Consumers can harvest massive welfare gains. Yet, the corporate providers of that technology may still fail to permanently capture profits proportional to its profound social utility.
The reason is simple:
Economic value and economic rent are not synonymous.
02 | Higher Productivity Does Not Equal Proportional Profit Growth
Consider an intuitive scenario.
AI empowers a single knowledge worker to accomplish the output previously requiring two or three employees. Macroeconomically, this represents a major productivity leap. Corporations reduce costs. Product prices fall. Consumers enjoy enhanced services. Potential economic output expands.
The critical question immediately follows:
Who captures this newly unlocked value?
It could be corporations. It could be consumers. It could be labor. Or it could be the underlying infrastructure and software providers. More realistically, it will be shared across all of them.
If AI tools become increasingly commoditized and competition intensifies, the efficiency gains driven by AI will likely be passed downstream to consumers through lower prices.
For society at large, this is a profoundly positive outcome. From an investment perspective, however, it implies:
The social value of AI can expand continuously, while the excess profits of AI firms do not necessarily expand in tandem.
This is a vital nuance often overlooked in contemporary AI allocation models.
03 | Technological Revolutions Can Ultimately Lower the Profitability of the Technology Itself
Historical technological waves frequently follow a paradoxical life cycle:
- Scarcity Phase: Early technology is scarce, granting pioneers immense pricing power.
- Maturation Phase: As technology matures, production costs drop.
- Adoption Phase: Falling costs broaden adoption.
- Commoditization Phase: Expanding adoption breeds intense competition.
- Infrastructure Phase: Eventually, the technology morphs into ubiquitous foundational infrastructure.
From a societal viewpoint, this is the ultimate triumph of technology. From a capital perspective, however, it implies:
The utility value of the technology continues to climb, while the excess profits capturable by technology providers are progressively constrained by market competition.
Whether AI will mirror this historical trajectory remains an open debate, but it must be modelled as a baseline risk scenario.
If future compute power, foundational models, and AI infrastructure experience collapsing unit costs, total AI utilization will surge. This will magnify AI’s macroeconomic contribution. Simultaneously, however:
The economic profit generated per unit of AI service may structurally decline.
This gives rise to a central investment paradox:
The more successfully AI becomes a general-purpose infrastructure, the greater the social wealth it creates; but this does not automatically guarantee that the AI industry can permanently secure an equivalent proportion of economic rents.
04 | What Becomes Truly Scarce Might Not Be AI Itself
This structural reality explains why future AI research cannot remain anchored solely to the question of « who owns the best model. » A far more critical inquiry is:
What remains scarce once AI is ubiquitous?
If model capabilities rapidly commoditize, the economic rents derived from models will erode. If compute power scales exponentially and unit costs plummet, standalone hardware advantages will succumb to market forces. If AI tools become frictionless to acquire, merely « possessing AI » ceases to constitute a durable moat.
True long-term scarcity will inevitably migrate elsewhere:
- Proprietary, high-grade datasets
- Deep distribution channels and locked-in user ecosystems
- Complex enterprise workflows and mission-critical integrations
- Ecosystem network effects
- Power generation and critical physical infrastructure
- Bottlenecked hardware resources
- Customer switching costs
- Brand equity and systemic trust
- The operational capability to embed AI into core business processes
Consequently, a pivotal strategic question arises:
Will AI reinforce existing corporate moats, or erode them?
Different answers carry vastly divergent implications for long-term capital compounding.
05 | The Same AI Growth Can Yield Vastly Different Investment Outcomes
We can frame these dynamics through four distinct structural scenarios:
Scenario A | Concentrated Economic Rents
AI creates immense value, and a select oligopoly possessing critical technology, infrastructure, and ecosystems continuously captures a dominant share of it.
- Productivity ↑ | Corporate Profit ↑↑ | Economic Rent Concentrated
- Outcome: AI success directly translates into multi-decade excess returns for a handful of enterprise titans.
Scenario B | Platform Competition
AI creates massive value, but technological deflation and intense competition continuously compress pricing power. Utilization surges. Enterprise revenues grow, but operating margins face relentless structural pressure.
- Productivity ↑↑ | AI Usage ↑↑ | Margins →
- Outcome: Society reaps immense economic dividends, but capital returns trail volume growth.
Scenario C | Consumers Capture the Bulk of Gains
AI drives the marginal cost of knowledge, software, and services near zero. Inter-firm competition ultimately transfers all efficiency gains into lower end-user prices. Consumers capture the primary economic surplus.
- Outcome: A golden age for human welfare, coupled with highly challenging return profiles for AI asset investors.
Scenario D | Economic Value Expands, but Rents Constantly Reallocate
Intra-industry technological disruption remains unrelenting. New model architectures, novel chips, and disruptive application-layer startups continuously displace incumbents. Today’s market leader cannot guarantee its position a decade hence.
- Outcome: Aggregate economic value climbs steadily, while individual firms’ economic rents are perpetually reshuffled. This represents perhaps the most probable and demanding scenario for long-term allocators.
06 | Therefore, We Must Focus on « Rent Duration »
Traditional financial valuation typically asks: What will future earnings look like?
In a hyper-dynamic ecosystem like AI, we argue that an indispensable second question must be added:
How long can those earnings be sustained?
This is fundamentally an investigation into a company’s Economic Rent Duration.
A business generating excess returns for a five-year horizon commands a profoundly different intrinsic valuation than one sustaining them for fifteen. Identical profit margins backed by a competitive advantage lasting three years versus fifteen imply completely different terminal value assumptions.
Therefore:
A core variable in AI investing is not merely earnings growth velocity, but the duration of excess capital returns.
This is why evaluating price-to-earnings (P/E) multiples at a single static point in time fails to capture the long-term risk profile of AI assets. What truly demands rigorous modeling is:
$$\text{ROIC} \rightarrow \text{Reinvestment} \rightarrow \text{Competitive Advantage} \rightarrow \text{Duration}$$
07 | What Do Microsoft, Google, and Other Platforms Truly Need to Prove?
For major technology incumbents, the guiding question is no longer simply: Can AI drive top-line revenue?
Instead, it is:
Can AI structurally reinforce these companies’ economic rents, and extend their duration beyond current market consensus expectations?
We evaluate this across several key vectors:
- Does AI deepen customer switching costs? If AI embeds deeply into mission-critical enterprise workflows, ecosystem lock-in strengthens.
- Does AI amplify platform network effects? If AI deepens user dependency on existing platforms, competitive moats widen.
- Is AI merely bloating capital expenditure? If revenue growth derived from AI requires perpetually escalating CapEx, the capital efficiency of incremental profits deteriorates.
- Has AI become a commoditized utility? If all major players eventually access parity AI capabilities, AI per se ceases to act as a long-term differentiator.
- Are entirely new profit pools emerging? The ultimate question remains: Is AI simply redistributing existing corporate profits, or creating a genuinely new, expandable profit pool?
08 | Scrutinizing the « Certainty Premium »
That markets are willing to pay a premium for growth certainty is entirely rational. What requires rigorous scrutiny is:
Where does that certainty actually originate?
If certainty is anchored in pre-existing cash flows, unassailable customer ecosystems, verifiable competitive moats, and consistent high capital efficiency, it rests on solid empirical ground.
Conversely, if market certainty increasingly relies on speculative assumptions—such as perpetual hyper-growth in AI adoption, frictionless commercialization, pristine long-term profit margins, or permanent incumbency—investors are exposed to a severe underlying risk:
An unhedged reliance on the duration of future economic rents.
This does not imply the market is fundamentally mistaken. It dictates that:
We must remain explicitly aware of precisely what type of certainty our capital is pricing.
09 | The Minerva Monitoring Framework
Accordingly, Minerva’s ongoing monitoring of the AI cycle will not be tethered solely to the absolute magnitude of sector-wide CapEx. We track the dynamic relationships across a core set of variables:
| Variable | Our Core Analytical Question |
| AI CapEx | Are deployment outlays continuing to accelerate or plateauing? |
| AI Revenue | Are monetizable revenue streams genuinely materializing at scale? |
| Free Cash Flow | Does top-line AI growth successfully convert into tangible cash? |
| ROIC | Are returns on incremental invested capital compressing? |
| AI Pricing | Are compute and model unit costs continuing their deflationary path? |
| Enterprise Productivity | Is AI demonstrably enhancing operational margins on the ground? |
| Profit Margins | Is competitive pressure transferring productivity gains directly to buyers? |
| Market Share | Are economic rents remaining structurally concentrated? |
| Competitive Entry | Is the influx of disruptive new entrants accelerating? |
| Rent Duration | How long can excess capital returns realistically be defended? |
We place maximum emphasis on this final metric.
Because:
If the aggregate value created by AI expands, but the duration of excess profits shrinks, AI may go down in history as a magnificent technological revolution while failing to deliver equally magnificent long-term investment returns.
Conversely, if market leaders successfully translate AI-driven productivity advantages into structurally higher ROIC—and defend those advantages over time—current lofty market expectations may gradually find validation in fundamental earnings power.
10 | Minerva Perspective
We do not believe investors need to resolve whether AI is a structural bubble, nor do they need to polarize between blind technological optimism and absolute skepticism.
We believe the more intellectually honest and rewarding inquiry is:
How will the economic value created by AI be partitioned among capital, corporations, labor, and consumers?
Technological breakthrough is merely step one. Productivity expansion is step two. The factor that definitively dictates long-term investment performance is step three:
Who captures the economic rents, and how long can those rents be preserved?
This is why a corporation can be exceptionally managed, and an industry can possess a dazzling long-term growth outlook, yet professional allocators must still independently assess:
What scale of economic rent does today’s market price actually imply?
And:
Has the market priced in multi-year rent persistence assumptions that may ultimately prove overly optimistic?
We cannot precisely forecast the exact aggregate value AI will create over the coming decades. Nor is such precision necessary.
What is required is continuous, rigorous observation of economic reality:
- Is productivity visibly rising?
- Is tangible value forming?
- Who ultimately captures that value?
- Do accounting profits convert into free cash flow?
- Can excess returns be defended over time?
Ultimately, Minerva’s core focus is not on whether AI will change the world—that premise is already settled.
What we must study is:
When AI reshapes the world, where will the economic rents ultimately settle?
Because for an investor,
Creating value is one thing.
Owning value is another.
And owning a business capable of retaining that value over the long term at a rational price is entirely a third.