If you’re watching decentralized physical infrastructure (DePIN) — token incentives paying people around the world to contribute GPU compute, storage, and network gear, stitched into something cheaper than AWS — this one lesson lets you tell truth from fiction in about ten seconds.

This vertical has an unusually clean credibility test, and it’s a single framework: subsidizing supply vs. demand sustaining supply.

• Subsidizing supply (early stage): token subsidies attract resource contributors to build up the supply side first. Any “boom” here is printed by the token, not paid for by a customer.
• Demand sustaining supply (mature stage): real, paying customer demand covers, or exceeds, what’s being paid out to contributors. Revenue here is real cash from real customers.

So judging whether a DePIN project is real comes down to one question: is the token incentive a contributor receives ultimately covered by real, paying downstream demand, or is it being backfilled by newly arriving speculative capital? Crossing from “subsidizing supply” to “demand sustaining supply” is the one genuinely important inflection point in this entire vertical. A project that’s crossed it deserves serious analysis; one that hasn’t is, more often than not, just token speculation wearing an infrastructure costume.

Akash Network is the sample with the most solid data behind it — because the downstream demand it’s bound to (AI inference) is real, high-growth, and paid for. Its key numbers all point toward real customer payment, not subsidy-built scale: a record $5 million in compute consumption in a single quarter, its AkashML platform processing 1.7 billion tokens a day, GPU utilization above 80% and up sharply year-over-year, real inference requests running on the platform. What matters about these numbers isn’t their absolute size — it’s that they were generated by actual downstream usage, not printed by the token.

But the honest caveat: across the whole industry, monthly real on-chain revenue only just crossed $150 million in early 2026 — relative to years of narrative excitement around this vertical, that absolute scale still isn’t large. A large share of DePIN projects are still stuck at “subsidizing supply” while getting priced as if they were already mature “demand sustaining supply” assets.

This framework travels well beyond DePIN — it works on any network that used incentives to bootstrap cold-start supply (restaking, AI agent networks included): don’t look at how many nodes it has or how big its TVL is. Look at how much of its revenue is customer money, and how much is the network printing its own fuel. The former is a business. The latter is subsidy life support.

When you look at a DePIN project, do you check node count and TVL first — or do you go straight to what percentage of revenue is actually customer-paid?

— Adapted from Crypto Sector Leaders, Chapter 15: DePIN — Decentralized Physical Infrastructure

#DePIN #AICompute #DecentralizedInfrastructure