DePIN is the only narrative I see institutional capital actually building positions in quietly. Most retail still thinks it means "crypto buys hardware" — that's missing the entire point. Three plays priced very differently.

$RENDER — GPU compute marketplace

Render started in 3D rendering but the actual thesis post-AI is decentralized GPU inference. OctaneRender migration to a token-incentivized network is the proof-of-concept; the real prize is offloading AI inference workloads to consumer/prosumer GPUs at prices hyperscalers can't match. Watch: number of nodes serving non-rendering compute jobs. That's the leading indicator.

$HNT — wireless infrastructure

Helium pivoted from LoRaWAN to 5G and mobile data. The MOBILE subnetwork lets users earn $HNT by deploying small cells and lets carriers pay in fiat for offload capacity. Discovery Mapping (covering wireless data) is the metric. If carrier offload deals show up in protocol revenue, the bull case lands. If it stays consumer subsidies, the dilution math doesn't work.

$FIL — storage

Filecoin is the boring one, but the boring story is what makes it interesting again. The Filecoin Virtual Machine (FVM) brought programmable storage — compute deals, retrievals, paid bandwidth. Storage at scale is a commodity; the bet is that programmable retrieval becomes the durable layer between IPFS and any real consumer crypto app needing media.

What ties them together: each one is selling supply-side capacity (compute, bandwidth, storage) at prices subsidized by token issuance. The bull thesis requires real demand-side payment in fiat or stablecoins to eventually exceed dilution.

Takeaway: ignore TVL on DePIN protocols. The only metric that matters is real revenue paid by non-protocol entities, expressed as a fraction of token issuance. If that ratio is climbing on $RENDER, $HNT, or $FIL — sustainable. If not — you're subsidizing the network with your bag.