Decentralized compute is the busiest corner of DePIN: networks renting out GPUs and CPUs for AI training, inference and rendering. This post takes four of them — Akash, io.net, Aethir and Nosana — and shows exactly what we measure and where each number comes from. Every figure below carries a source and a verify_url, and the market and supply values are a point-in-time snapshot as of 2026-08-09 17:15 UTC. Open any source link and compare.

A note on what "metric depth" means: the counts below are how many distinct live metrics we track for each network across all tiers — first-party provider telemetry, market data and protocol fees combined. The specific dollar and supply figures we cite here are the market and revenue metrics, sourced from CoinGecko and DefiLlama with their own verify_urls. How all of this is sourced and anchored is covered on the Provenance page.

Akash (AKT) — 41 metrics tracked

Akash is the deepest compute network in our coverage, with 41 distinct live metrics spanning market data and first-party provider and lease telemetry read from the Akash console.

The provider and lease side comes straight from Akash's own console (akash_console), which is why AKT carries far more than the market fields alone.

io.net (IO) — 20 metrics tracked

io.net aggregates GPU supply for AI workloads. We track 20 live metrics for it.

With circulating supply just under half of the 800M cap, the gap between market cap and fully-diluted valuation is a number an agent can compute directly from these fields — no repackaging required.

Aethir (ATH) — 13 metrics tracked

Aethir is enterprise-grade GPU-as-a-service, and it is one of the few compute networks in this set with usable fee data. We track 13 live metrics for it, including protocol fees from DefiLlama.

The fee figures come from DefiLlama and carry their own verify_url. Market and revenue live in different tiers, and we keep those labels all the way to your query so you always know which is which.

Nosana (NOS) — 13 metrics tracked

Nosana runs a Solana-based GPU grid aimed at AI inference. We track 13 live metrics for it, and its supply schedule is unusually clean.

Circulating supply sits within a rounding error of the 100M cap, so for Nosana market cap and fully-diluted valuation are effectively the same number — again, derivable straight from the fields.

Why the provenance matters here

Compute DePIN is noisy. Supply figures drift, fee endpoints go quiet, and "GPU count" means different things on different networks. Our answer is not to smooth it into a confident-looking dashboard — it is to hand you every number with the source it came from and the timestamp it was true.

And every one of these values is part of the daily batch that gets Merkle-rooted and timestamped to Bitcoin, so the snapshot you read today is provably the snapshot we published today.

See the full method on the Provenance page, or explore coverage and depth across the DePIN data product. API access to query these networks directly is available on approval.