The Investment Paradigm Revolution: When the Definition of an Asset Changes, How Is Investment Logic Rebuilt?

Investment is, at its core, a machine that converts "a definition of the future" into "a price today." Through the allocation of capital across time, it shapes our economic landscape and our social future. Yet when the very definition of an "asset" as we know it is overturned, and when the underlying logic that sustains investment decisions is no longer self-evident, this machine faces a fundamental rewriting—at its input end (what can be invested in), its pricing end (what it is worth), its risk end (what we fear), its output end (what we earn), and even the "rights" investors acquire. This is not a simple technological upgrade, nor a cyclical market fluctuation, but an AI-driven revolution in the investment paradigm that is reshaping the relationship between capital and value.

At the heart of this revolution lies a fundamental migration of the investment target from "labor-intensive assets" to "AI-augmented assets." For decades, talent was a company's most core asset and moat; now, the arrival of AI is—at unprecedented speed and breadth—demoting human labor from an "irreplaceable unit of execution" to "capacity that can be substituted by compute." This will pose disruptive challenges and opportunities to traditional investment valuation models, risk-identification frameworks, return structures, and even the role of capital.

Before delving into the revolution, we must first understand the six "self-evident" assumptions on which the old paradigm depended. They formed the bedrock of our investment decisions over the past several decades:

**I1 | The target is "a collection of people"; human labor is the core term in the production function and the moat.** Traditional investment thinking holds that a company is a collection of people, equipment, and organization. Whether in service outsourcing, IT services, content production, or entry-level consulting, its valuation and growth models are built on the labor leverage of "hire—train—raise prices—replicate." Labor is both a cost and an asset that can be depreciated, leveraged, and priced at a premium. 【verified · classic common knowledge + FR-12 discussion of the production function】

**I2 | Value is intrinsic to the target and can be measured by discounted future cash flows (DCF) or comparable multiples.** The classic definition of investment is "forgoing current consumption in exchange for a greater future return"; the target's value is regarded as the risk-adjusted present value of its future cash flows. Multiples such as P/E and P/S are, in essence, merely simplified shorthand for the DCF model. 【verified · classic literature (the standard framework in finance)】

**I3 | Growth requires capital, and capital mainly buys labor and operations; economies of scale come from declining marginal cost.** Software is regarded as an ideal investment target because its replication cost approaches zero and its gross margins are high, enabling economies of scale through declining marginal cost. The investor's task, therefore, is to fund "software that can scale." 【verified · standard industry understanding + the Bessemer comparison below】

**I4 | Risk is classifiable and can be compensated by a risk premium (a CAPM-style mindset).** Risk is subdivided into execution risk, market risk, technology risk, management risk, and so on; diversification across a portfolio can reduce nonsystematic risk. Higher risk corresponds to higher expected return—a widespread market belief. 【verified · classic literature (the standard framework of portfolio theory)】

**I5 | Returns follow a power law, but winners capture value over the long term by "owning assets"; exit = IPO or M&A.** The venture capital business model is built on a power-law distribution in which "a few big winners cover the many failures." Investors buy equity, share in the winners' future cash flows and valuation growth, and realize an exit through an IPO or M&A. 【verified · classic literature (the VC consensus on power laws)】

**I6 | Information asymmetry is resolved through due diligence, and the core of due diligence is team, market, and finances.** Traditional due diligence revolves around "betting on the people, the track, and the financial model." "Team background + market size + unit economics" forms the standard trio for assessing a project's feasibility and potential value. 【verified · standard industry framework】

【inference】Together, these six assumptions shaped a kind of "human-resource ontology": first there are employable, organizable people; then those people produce predictable cash flows; and finally those cash flows are discounted according to risk. The substance of investment is people, and the assets those people drive.

Not every change brought by AI amounts to a revolution. A "within-paradigm improvement" in Kuhn's sense optimizes parameters within the old definition; a "paradigm revolution" replaces the criteria for "what counts as a problem and what counts as a good answer." Applied to "investment," the specific test this report uses is:

| Question | Old-paradigm default answer | Deepening (optimization within the old definition) | Revolution (changing the definition) |

|---|---|---|---|

| What to invest in? | Enterprises that own human labor, equipment, and organizational capability | Use AI to boost the efficiency of portfolio companies | Invest in "AI-augmented production units": value = leverage structure × boundary, not its headcount and stock |

| Why is it worth something? | Discounted future cash flows + comparable multiples | Change the discount rate, change the comparables, add AI to the premium | The valuation anchor shifts to "verifiable unit economics × reuse rate × convexity (option value) × boundary executability" |

| What do we fear? | Execution risk, market risk, management risk | Add an AI due-diligence checklist | The risk subject becomes capability depreciation, model drift, compute concentration, the verification tax, and liability exposure |

| What do we earn? | A share of the target's cash flows and exit spread | Find winners faster | A share of **convexity**: near-zero marginal replication makes the upside unlimited and the downside limited, pushing the return distribution further toward a power law |

| What right is bought? | Equity, ownership, control | Better terms and preference rights | What is bought is the right to access, orchestrate, and settle on the "boundary layer" (verification/distribution/compute/data) |

| What is the role of capital? | To buy labor and growth (buy scale) | Smarter cash burn | To buy **time and certainty** (turning uncertainty into verifiability) |

【inference】The core judgment: the watershed is not "AI companies deserve higher multiples" (that is stuffing a new object into the old valuation formula—a deepening), but rather that "**when execution is no longer scarce, when revenue decouples from headcount, and when the most valuable objects can be publicly replicated, the ontology of the investment target shifts from 'a collection of people' to 'leverage and boundaries,' and the value anchor shifts from discounted cash flows toward convexity and credibility**" (that is a revolution).

The rise of AI does not merely improve efficiency; through seven mechanisms it is fundamentally shaking the foundations of the old paradigm:

**M1 | The cost of execution collapses, and people are demoted from an "irreplaceable unit of execution" to "capacity substitutable by compute."**

The marginal cost of execution for tasks such as content generation, code writing, customer-service responses, first-draft writing, data wrangling, and basic analysis is falling sharply. The old-paradigm equation "more people = more capacity = more cash flow" is broken: AI means the term "people" no longer automatically equals a moat. This is precisely the projection onto the investment side of FR-10's conclusion (talent = defining the problem + wielding tools + taking responsibility for verification). 【inference, based on FR-10 + FR-12】

**M2 | Revenue decouples from headcount, producing investable samples of "tiny unit, extremely high output."**

AI-driven, extraordinarily high revenue per employee is challenging traditional labor-intensive valuation models. According to Epoch AI's estimates, as relayed by media outlets such as KuCoin/AIMPACT, Anthropic generates roughly US$9 million in revenue per employee, OpenAI about US$5.6 million, and Nvidia about US$5.1 million ([KuCoin/AIMPACT relaying Epoch AI, 2026-05](https://www.kucoin.com/news/flash/anthropic-generates-9-million-in-revenue-per-employee-surpassing-openai-and-tech-giants)). 【verified · title checked/secondhand relay · pending primary-source confirmation】Midjourney has been estimated to reach roughly US$500 million in annualized revenue with about 40 employees and zero outside funding ([Value Add VC, 2026-07](https://valueaddvc.com/blog/how-does-midjourney-make-money-500m-arr-zero-funding-and-the-ai-image-business-model-explained)). 【secondhand estimate · to be verified (the company does not publish financials)】This means that for the first time "output per person" can be high enough to render "valuation by headcount" ineffective, and the target's value must find another anchor. This is consistent with FR-13's judgment (what is truly accumulable is process capital and trust capital). 【inference, based on M2 + FR-13】

**M3 | The value anchor migrates from "cash flow/multiples" to "capability × reuse rate × verifiability × boundary."**

When generation is cheap and replication costs nearly nothing, scarcity migrates from the production side to verification, attribution, distribution, and trust. The object of investment due diligence changes accordingly: not "how many people, how much revenue," but "whether unit economics are verifiable, how high the reuse rate is, and whether the boundary is executable." One verifiable signal: at the end of January 2025, the median forward ARR multiple for AI public software companies was about twice that of non-AI software companies (AI 11.2x vs. software 5.5x), AI companies accounted for about 70% of B2B Series A rounds (about 40% in early 2024), and AI Series A valuations averaged roughly 40% higher ([Tomasz Tunguz, 2025-02-03](https://tomasztunguz.com/private_valuations_2025/)). 【verified · primary-source blog (the author is a VC; the data is self-reported)】This may be "revolutionary pricing," or it may be "a premium within the old paradigm"—the distinguishing point being whether the premium buys the cash flows of labor substitution (deepening) or buys convexity and boundaries (revolution). 【inference】

**M4 | The source of risk shifts from "execution/labor" to "capability depreciation/model drift/compute concentration/the verification tax/liability."**

An AI-augmented asset needs continuous operation, monitoring, and re-verification to preserve its value. Once a model is upgraded, a data distribution drifts, or an interface changes, yesterday's "asset" may become worthless today (following FR-13 M5/M7). FR-12 has already argued that "cost disease migrates to the verification end," so investment risk is rewritten as: the rate of capability depreciation, the drift rate, concentration on a single model/compute vendor, the verification tax, and liability exposure when things go wrong. In the old paradigm, "people" are a stable, trainable, replaceable asset; in the new paradigm, "models" are assets that depreciate rapidly, can be replicated by open source, and may be held accountable by regulators. 【inference, based on FR-12/FR-13】

**M5 | The return structure shifts from "sharing cash flows" to "sharing convexity."**

The extreme property of non-rival digital products is that a success has an almost unlimited upside while failure carries a limited marginal cost. When AI lets a "successful product" be replicated at near-zero cost, the return function looks more like an option than a bond. This explains why capital is willing to pay for "high-multiple, low-margin, high-uncertainty" AI targets: it is not buying current cash flows but convexity. At the same time, market data show capital betting earlier and more concentratedly—"fewer bets, more capital" ([TechCrunch citing Carta data, 2026-03-20](https://techcrunch.com/2026/03/20/ai-startups-are-eating-the-venture-industry-and-the-returns-so-far-are-good/); for exact figures 【verified · title checked/secondhand relay】). 【inference, based on M5】

**M6 | The return of capital intensity: from "asset-light software" back to "power + chips + data centers."**

Contrary to the intuition that "AI makes everything lighter," AI infrastructure is capital-heavy. TrendForce has revised upward its forecast for the combined capital expenditure of the world's top nine cloud providers in 2026 to about US$830 billion, a year-on-year increase of roughly 79% (including about US$190 billion for Microsoft, over US$230 billion for AWS, US$180–190 billion for Google, and US$125–145 billion for Meta) ([Evertiq relaying TrendForce, 2026-05-06](https://evertiq.com/news/2026-05-06-ai-boom-pushes-hyperscaler-capex-towards-usd-830-billion-in-2026)). 【verified · body checked/secondhand relay (TrendForce via Evertiq)】This means the investment paradigm is not a one-way "decapitalization": the application layer becomes extremely light while the infrastructure layer becomes extremely heavy, and the logic of capital allocation between the two layers diverges. 【inference】

**M7 | The unit of account migrates: from seat/hour/EBITDA to outcome/credit/result.**

The seat-based SaaS model is being eroded by agents—the more capable the agent, the fewer seats the customer needs, and the more successful the product, the more it cannibalizes its own revenue. After analyzing more than 30 SaaS vendors, Bain found that 35% merely bundled AI into a higher seat price, 65% adopted a "seat + usage" hybrid, and none had fully shifted to outcome-based pricing ([UC Today citing Bain, 2026-05-12](https://uctoday.com/is-per-seat-saas-pricing-dead-monday-coms-consumption-based-pricing-pivot)). 【verified · body checked/secondhand relay】monday.com has already rolled out a "seat + credits" model. This means the measurement basis of revenue is being renegotiated, and investment valuation must ultimately follow the unit of account. 【inference】

These mechanisms are not idle speculation; they show up in the actual data of capital flows and point to a structural reallocation of investment logic:

These data clearly show that capital is undergoing a structural reallocation toward AI at unprecedented speed and scale, heralding a fundamental shift in the rules of investment.

Faced with the shaking of the old paradigm and the emergence of new evidence, we need to establish a new investment logic. The following are six candidate definitions for the new paradigm:

**D1 | The ontology of the investment target (candidate):**

The old definition asks "how many people, how many assets, how much cash flow does this company have"; the new definition asks "how many times can this unit amplify AI capability, how strong an exclusive and verifiable claim on results can it establish, and who bears responsibility." 【inference, based on M1–M4】This definition directly inherits FR-13's view of assets: models and compute are **rented**, while proprietary flows and trust boundaries are what is **owned**; what investment should buy is the latter, yet it often mistakenly buys the former.

**D2 | The valuation benchmark (candidate):**

**D3 | The definition of risk (candidate):**

**D4 | The return structure (candidate):**

**D5 | Investor rights (candidate):**

**D6 | The role of capital (candidate):**

**The final determination table for deepening vs. revolution:**

| Action | Classification |

|---|---|

| Using AI to improve portfolio-company efficiency, giving AI companies higher multiples | Deepening (within the old definition) |

| Improving DCF parameters, writing AI into the due-diligence checklist | Deepening |

| Revaluing software companies by "revenue per employee" | Deepening (new numbers for an old metric) |

| The target shifting from "a collection of people" to "a production unit of leverage + boundaries" | **Revolution** |

| The valuation anchor shifting from discounted cash flows to convexity + credibility | **Revolution** |

| The risk subject becoming capability depreciation/drift/concentration/verification | **Revolution** |

| Returns shifting from sharing cash flows to sharing convexity | **Revolution** |

| Investor rights shifting from ownership to access/verification/settlement rights | **Revolution** |

The deep impact of this paradigm revolution is that the core of economic scarcity is shifting from "labor" and "capital" toward "trust" and "verification." When execution capability becomes nearly infinite and cheap, what becomes truly valuable is who can reliably define the problem, who can efficiently wield the tools, and who can accurately verify results and take responsibility for them (following FR-10).

This will lead to hyper-concentration of capital, because only a few "boundary nodes" with strong moats in technology, data, and brand can capture convexity returns in a world of near-zero marginal-cost replication. Other links further down the value chain will face fierce price competition and margin compression.

At the same time, our traditional economic statistics and growth models will also fail. Indicators such as GDP and output per capita may be unable to accurately reflect the true output and wealth creation of the new economy. A new concept of "trust capital" and a "verification tax" will become crucial—invisible yet decisive costs that sustain the healthy operation of the AI economy.

Faced with this investment paradigm revolution, different players need to adopt different action strategies:

**1. For entrepreneurs: Redefine your "assets" and "moat."**

**2. For investors: Rebuild your "screening checklist" and "valuation framework."**

**3. For researchers and policymakers: Define new metrics and guide a new order.**

The investment paradigm revolution is not a distant concept; it is unfolding before our eyes at astonishing speed. When the definition of an "asset" is re-encoded, when the anchor of "value" shifts, and when the nature of "risk" changes completely, we must rebuild the underlying logic of investment. This is not an elective question; it is a mandatory one posed by our era.