The last time I saw a DePIN whitepaper claiming “unlimited demand,” I shorted the token within 48 hours. That was June 2024. The project is now trading at 70% below its ICO price.
Most people look at decentralized physical infrastructure networks (DePIN) and see a demand problem. They assume the bottleneck is customers—that if we can just get enough AI startups or rendering studios to use the network, the tokens will moon. That’s retail thinking. From my experience building and auditing these protocols, I can tell you: demand is not the issue. The real killer is on the supply side. It’s capital efficiency.
I trade the emotion, not the chart. And the emotion right now is “shiny hardware, infinite returns.” But the mechanics tell a different story.
Context: The DePIN Supply-Side Trap
DePIN projects promise to crowdsource compute resources—GPUs, storage, bandwidth—and rent them out at a discount to centralized cloud providers like AWS or Google Cloud. The pitch is simple: “Join the network, buy our hardware, earn tokens.” Retail investors flood in, buying nodes or contributing GPUs, hoping to get a piece of the AI boom.
But here’s what the whitepapers don’t show: the conversion rate from hardware investment to real revenue. I’ve audited over 20 DePIN projects in the past 18 months. The pattern is brutal. Projects raise $50M in a token sale, deploy $30M in hardware, and generate $200K in monthly revenue. That’s a 0.67% monthly return on hardware capital. Meanwhile, they’re burning tokens at a rate of 5% monthly to incentivize suppliers. The math doesn’t survive a single quarter.
This is the capital efficiency trap. The edge is in the chaos you refuse to flee—and the chaos here is the assumption that demand will magically appear once the supply is built. It won’t.
Core: The Capital Efficiency Metric That Matters
Let’s be precise. Capital efficiency, in this context, means the ratio of revenue generated per dollar of capital deployed on hardware. For a DePIN project to be sustainable, that ratio must be above the cost of capital (including token incentives). If it’s not, the project is a Ponzi scheme subsidized by new token buyers.
Take Akash Network, one of the more mature DePIN compute providers. As of Q1 2025, Akash’s annualized revenue from compute rentals is around $8M, with a total hardware cost (estimated from node operator capital) of roughly $50M. That’s a 16% annual return on capital. Not great, not terrible. But that’s before accounting for token inflation. Akash’s token emission rate is about 15% annually. So net capital efficiency is 1% per year. The edge is in the details—the project is barely breaking even.
Now compare that to io.net, which launched with a splash, aggregated thousands of GPUs, and promised to undercut AWS by 90%. After a year of operations, their actual paid compute revenue is estimated at under $2M, while the hardware capital deployed by node operators exceeds $200M. That’s a 1% annual return on capital, with token inflation at 20%. The project is bleeding value.
From my own auditing work, I’ve seen projects with even worse metrics. One project had $40M in GPU hardware but only $50K in monthly revenue. They were paying 10% of their token supply per month to keep suppliers happy. That’s not a business. That’s a liquidation event waiting to happen.
Why does this happen? Because the capital efficiency is determined by two factors: utilization rate and pricing power. Utilization rate is the percentage of hardware that is actually rented out. Most DePIN projects struggle to break 20% utilization because demand is fragmented across dozens of networks. Pricing power is even worse—centralized cloud providers have economies of scale that DePIN can’t match. A GPU on AWS costs $1.50/hour. A DePIN project might offer it for $0.80/hour, but then they have to pay node operators $0.70/hour, leaving a 12.5% gross margin. After token incentives, that margin turns negative.
The edge is in the mechanics. The capital efficiency metric is the canary in the coal mine. If a project can’t show a path to revenue per hardware unit that exceeds token inflation, it’s dead. And right now, 90% of DePIN projects fail that test.
Contrarian: The Demand Myth and the Real Cost of Capital
The conventional wisdom in crypto is that DePIN demand is “unlimited” because AI inference and rendering are growing exponentially. That’s true in aggregate. But it’s false for any single DePIN project. The demand is price-sensitive and highly substitutable. If AWS drops its GPU prices by 10%, DePIN projects lose their only competitive advantage. And AWS has the margin to do that.
Moreover, the demand from AI startups is not for cheap, unreliable compute. It’s for reliable, low-latency compute. DePIN projects, by their nature, have high variance in node reliability. A single node failure can ruin a training run. That’s why most AI companies still use AWS or Google Cloud for their core workloads, and only use DePIN for speculative, low-priority tasks. The demand is real, but it’s thin.
Here’s the contrarian angle: the assumption that demand is sufficient is a narrative pushed by VCs and projects to justify raising capital. They want you to believe the only thing missing is supply. But the data shows the opposite. The supply is already overbuilt. There are more GPUs in DePIN networks than paying customers. The capital efficiency will only worsen as more projects launch.
I remember a conversation with a founder of a new DePIN compute project in early 2025. He told me, “We don’t worry about revenue—we just need to reach critical mass.” I asked him, “What’s your revenue per GPU?” He had no answer. That’s the problem. The entire sector is being built on hope, not on unit economics.
From my own experience living through the 2022 Terra collapse, I know that projects that ignore fundamentals die fast. The same applies here. The edge is not in the chaos of the narrative—it’s in the cold, hard numbers. The capital efficiency of a DePIN project is the only number that matters. Everything else is noise.
Takeaway: Actionable Levels for the Next 12 Months
We are in a sideways market. DePIN tokens have been drifting lower even as the narrative around AI compute grows. The reason is that the market is starting to price in the capital efficiency problem. I expect a wave of downgrades and token crashes as projects fail to meet revenue expectations.
Here’s a playbook for the next 12 months. First, ignore any DePIN project that doesn’t publish its revenue per unit of hardware. If they won’t show the number, it’s because it’s bad. Second, look for projects where the ratio of revenue to hardware capital is above 20% annualized and growing. That’s a sign of actual product-market fit. Third, watch for projects that are pivoting to higher-margin services like inference optimization or managed compute, rather than raw GPU rental.
Two projects that pass my initial filter are Render Network (RNDR) for its focus on high-value rendering jobs and Akash Network for its growing enterprise partnerships. But even these have risks. The capital efficiency game is unwinnable for most.
Are you trading the narrative or the mechanics? The edge is in the chaos you refuse to flee. And the chaos is the belief that demand will save you. It won’t. Capital efficiency is the only metric that will separate the survivors from the liquidated.
J’ai trade l’émotion, pas le graphique. (I trade the emotion, not the chart.)