ROI Math: How to Calculate the True Return on an AI Infrastructure Investment

Investing in AI infrastructure without a clear return framework is speculation. Not necessarily bad speculation — the thesis may be correct — but it is not investing. Investing requires a model of how value is created, how it accrues to your position, and what the return looks like across time under different scenarios.

The standard growth stock framework — revenue multiple applied to projected ARR — does not adequately capture the return dynamics of AI infrastructure investments because it does not account for the compounding mechanisms that make these investments potentially extraordinary.

Here is the framework that does.

Component 1 — Base Demand Growth

Every AI infrastructure investment has a base demand growth rate driven by the overall expansion of AI deployment. This is the sector tailwind — the growth that would accrue to any competent player in the space regardless of their specific execution.

AI compute demand is growing at approximately thirty to forty percent per year compounded, driven by the combination of new model training runs, expanding inference workloads, and new application categories. This is the base growth rate for a decentralized compute network like Render.

AI verification demand is earlier in its curve and harder to model precisely, but the trajectory is clear: as AI deployment in regulated industries accelerates, the demand for cryptographic verification of AI computation grows with it. This is the base growth rate for a network like Nexus.

The first element of the return model: what is the base demand growth rate for this category over the investment horizon?

Component 2 — Market Share Trajectory

Capturing the base demand growth requires the investment target to maintain or grow its market share within its category. For decentralized infrastructure protocols, this means evaluating the competitive dynamics — are there other protocols in the same category, what are their relative technical advantages, what network effects protect the leader's position?

Render Network currently processes a meaningful share of decentralized GPU compute demand and benefits from first-mover advantages: the largest network of GPU suppliers, the deepest integration with professional rendering and AI tools, and the most liquid token market for participants who need to buy and sell compute capacity. These advantages compound as the network grows.

The second element of the return model: what is the market share trajectory of this specific investment within its category, and what defends that share over time?

Component 3 — Token Economics

For crypto infrastructure investments, the return is denominated in token appreciation rather than equity appreciation. The token economics — how the token supply grows or is burned over time, how value accrues to token holders versus being extracted by other participants, and how liquidity evolves — determine what fraction of the fundamental value creation is captured by token holders.

The most important token economic variable is value capture: does the protocol have mechanisms that tie token value to the growth of the network? Burn mechanisms that reduce supply as network usage grows. Staking requirements that reduce circulating supply as participants seek to earn network rewards. Fee capture that distributes protocol revenue to token holders.

The third element of the return model: how does token value accrue from network growth, and at what rate does the supply increase dilute that accrual?

Component 4 — The Scenario Analysis

Any honest return model has multiple scenarios. For an AI infrastructure investment, three scenarios cover the relevant range.

The base case assumes the sector tailwind is real but adoption is slower than optimists project. Market share is maintained but not grown. Token economics work as designed. In this scenario, returns over a five-year horizon are driven primarily by the base demand growth rate — two to four times investment in the best protocols.

The bull case assumes the sector tailwind accelerates beyond current projections — driven by AI deployment in regulated industries creating urgent verification demand, or by a significant reduction in centralized GPU costs that accelerates adoption of decentralized alternatives. In this scenario, returns are driven by both demand acceleration and market share growth — potentially ten to thirty times investment in well-positioned protocols.

The bear case assumes regulatory headwinds, technical execution failures, or competitive displacement reduce adoption below base case projections. In this scenario, returns are negative — the risk in any venture-stage investment.

Applying the Framework

For Render Network: base demand growth of thirty to forty percent per year in GPU compute. Market share maintained through first-mover advantages and network effects. Token burn mechanism that reduces supply as network fees grow. Base case suggests two to four times over five years. Bull case — if AI inference workloads migrate meaningfully to decentralized compute — suggests much higher. Bear case — if centralized cloud reduces GPU costs dramatically — suggests flat or negative.

For Nexus: base demand growth dependent on AI deployment in regulated industries, which is earlier in its curve. Market share potentially very high as there are few comparable ZK proof networks. Token economics still developing. The investment is longer-duration with higher variance — appropriate for a smaller position with a longer time horizon.

The framework does not eliminate uncertainty. Nothing does. But it distinguishes between informed positioning and speculation, and it provides a basis for comparing AI infrastructure investments against each other and against alternative uses of capital.

Invest with a model, not a hope.

The math matters.

Run it before you size the position.