Targon — subnet 4

What it does

Targon rents out GPU compute that has been hardware-sealed so that not even the person renting it out can peek inside while a job is running.

Targon is an AI infrastructure platform built on trusted execution environments and confidential compute: hardware-enforced memory encryption, secure boot, GPU-level isolated execution, and remote attestation, meant to make a rented GPU's workload verifiably private even from whoever owns the machine. [source] [source]

This subnet sits at the bottom of the tracked set on current earnings: its emission share is 0.0% @ 8871732, rank 29 @ 8871732. That is a striking pairing with its holder base — 8116 @ 8871732 wallets hold its alpha token, the second-largest holder count of any subnet this course tracks. 7 @ 8871732 miners are currently active, out of 256 @ 8871732 registered keys against a cap of 256 @ 8871732.

How the incentive mechanism works

Targon miners do not compete on a single shared score; they bid GPU capacity into per-hardware-type auctions. Each auction — one per GPU model, such as an H200 or B200 configuration — publishes a target number of cards, a target price per card, a payment cap per card, and a minimum cluster size a miner must meet to participate at all, all readable from a public API. [source] [source] [source] While an auction sits below its target card count, every live, verified card earns up to the payment cap; as more cards come online the pool spreads across them and the per-card rate settles down toward the target price. [source]

The validator's own weight-computation code turns this directly into a payout: for each auction it sums each bidding miner's card count against the pool, pays each capped at the auction's maximum price, and — if the combined asks from every auction would exceed the whole emission pool — scales every payout down by the same fraction rather than letting any one auction crowd out the rest. [source] [source] [source] Any part of the pool not claimed by a verified, participating card is not redistributed to other miners — it is burned, split across designated burn keys. [source] [source]

What miners are actually optimising

The stated goal is secure, verifiable, confidential compute. [source] What the weight-setting code actually pays for is narrower and more mechanical: verified card count inside a specific hardware auction, up to that auction's price cap, nothing about the workload actually run on the card. [source] A miner here is optimising for clearing the minimum cluster size, staying verified, and filling an auction that is still under its card target — not for any measure of what a renter's confidential workload actually needed.

With only 7 @ 8871732 active miners against 256 @ 8871732 registered keys, and an emission share of 0.0% @ 8871732, whatever this auction structure is currently paying for is being distributed across a very small field relative to how many wallets still hold the token.

Why it earns

Mechanism evidence

Cross-subnet emission is set by one formula, identical for every subnet: each one's share is its own moving-average alpha price, discounted by a miner-burn penalty, against every other subnet's. [source] [source] Mechanism quality has no other route into that number — it can only matter by first moving demand for the token. Targon's own auction-and-burn design is real, checkable code: unmatched auction demand is burned rather than redistributed, which is a genuine anti-dilution choice. [source] Whatever that design is worth, it is not what explains Targon's current standing — its measured emission share is 0.0% @ 8871732, which this dossier's evidence can only trace back to price, never to mechanism quality directly.

Stake evidence

13.9% @ 8871732 of the judging power behind Targon's committee traces to the root network rather than to stake placed directly in its own alpha. [source] A validator permit is awarded purely by stake weight, the top accounts by money behind them, with no separate judgment of scoring quality involved. [source] Targon's committee is one of the smallest in this batch — the exact count is in the technical section below — which means whatever judgment does still happen here rests on very few accounts, whatever emission is or is not currently flowing through it.

Price evidence

Unverified Counting directly across this snapshot's full subnet list — a computation this dossier performed itself rather than one available as a single citable figure — only a minority of the network's subnets carry any nonzero emission share at block 8871732; Targon is not among them.

Targon's own measured share is 0.0% @ 8871732. [source] [source] That is a measured zero, not a missing measurement — Targon's alpha still trades, at 0.055218354 @ 8871732 TAO, down -2.270048436174165 @ 8871732% over the trailing week but up 2.158621250422166 @ 8871732% over the trailing month, with 8116 @ 8871732 wallets still holding the token. Because cross-subnet emission runs entirely on relative price against every other subnet, a large legacy holder base does nothing on its own to restore a share once the price side of that ratio falls far enough behind the rest of the field.

Failure modes and attack surface

The auction design's own comment is explicit about a specific risk it accepts on purpose: if every auction's combined demand exceeds the emission pool, every payout is scaled down by the same fraction rather than letting the chain's own weight-normalisation silently favour whichever auction happened to be processed differently — the code scales first so that stored scores and actual payouts agree. [source] That is a real safeguard against one specific failure, not a defence against every one: the code's own comment elsewhere calls part of the burn-key distribution logic "kinda cursed," a sign this is intricate, hand-tuned arithmetic rather than a simple formula. [source]

More generally, every subnet's weight submissions collapse into a stake-weighted median that cannot tell a handful of validators who independently agree apart from one operator running several keys — and Targon's committee is small enough that this matters more than on a larger subnet. [source]

What to copy, what to avoid

What to copy: burning unclaimed auction demand instead of redistributing it to whoever happens to be present is a real defence against a thin field quietly capturing a full pool it did not actually earn — exactly the failure mode several other, smaller subnets in this batch show no documented protection against. [source]

What to avoid: a large legacy holder base is not a reserve of future emission. Targon holds the second-highest holder count this course tracks while earning a measured 0.0% @ 8871732 share — a reminder that under the current price-driven formula, past popularity buys nothing once relative price has moved on.