Research index
Explanatory briefSources checked October 20, 2026

Decentralized compute & DePIN: usage versus token incentives

How three networks document the link between paid work and token flows, and how a reader can separate customer-funded usage from incentive-driven activity without relying on a single headline number.

Not a JARCH investment position or recommendation. No prices, metrics, or performance claims are made about any network.

00Analytical interpretation

Summary

Decentralized compute and DePIN networks pair two flows: payments for work and token rewards for supplying capacity. Both can produce on-chain activity. Only one of them is a customer choosing to pay for a service. This brief reads three official mechanism descriptions and builds a framework for telling the two apart. It does not measure any network, and it does not identify winners.

01Primary-source fact

What the documentation says

Each item below restates official documentation as checked on October 20, 2026. Nothing here is extrapolated beyond the source.

Akash: lease reconciliation

Akash's provider guide describes comparing active on-chain leases with the workloads running in the provider's Kubernetes cluster, to detect leases that are live while their workloads are unhealthy, or leases that are closed while resources remain allocated.[1]

Render: burn-mint equilibrium

Render's documentation states that jobs can be paid in RENDER or fiat, with fiat converted to RENDER for settlement. The RENDER is burned on completion, while emissions reward suppliers on a schedule.[2]

Helium: Data Credits

Helium's documentation states that Data Credits pay for data transfer and also for certain onboarding and location-assertion actions; Data Credits are generated by converting HNT and are nontransferable.[3]

02Analytical interpretation

Why a mechanism is not evidence of adoption

These examples illustrate why service-payment records and supply incentives should be examined separately. A documented settlement mechanism can create a trail a reader can follow, but that trail still needs context before it supports an adoption claim.

The Akash guide is instructive precisely because it is operational. If providers are told to check whether a live lease actually has a healthy workload behind it,[1] then an on-chain lease and useful completed work are, by the documentation's own logic, two different things that can diverge. A count of leases is a count of agreements, not of delivered computation.

Render's model separates payments from emissions.[2] That separation lets a reader ask how much supplier income comes from completed customer jobs versus scheduled emissions, but the existence of a burn does not by itself say who paid, whether they returned, or whether the economics are sustainable.

Helium's Data Credits cover both data transfer and some onboarding and location actions.[3] A single Data Credit total can therefore blend recurring traffic with one-time network-building activity. The mechanism is documented; the mix within any given period is a separate empirical question.

These documents describe how systems are designed to work. They are not independently audited statements of adoption, revenue, or profitability, and the rules they describe can change.

03Analytical interpretation

Four distinctions worth holding

The following is an original analytical framework, not a measured finding about any network.

Installed supply vs. completed useful work
Registered GPUs, hotspots, or providers describe capacity. Completed, healthy, paid jobs describe use.
Customer-funded vs. subsidized payments
Separate independent customer payments from grants, rebates, credits, and related-party spend.
Retained fees vs. gross spend and rewards
Gross customer spend, operator rewards, and what the protocol nets are three different numbers.
Repeat customers vs. addresses
Many addresses can be one payer; one payer returning over time says more than many first-time wallets.

Burns are a settlement event. On their own they do not establish customer retention, operator profitability, or how much value accrues to a token.

04Analytical interpretation

Comparisons to avoid

  • GPU hours across unlike hardware or unlike workloads. An hour of one accelerator on inference is not an hour of another on rendering.
  • Wireless onboarding or location actions alongside recurring data traffic, as if both were the same kind of demand.
  • Burn totals as a proxy for revenue, margin, or customer count.
  • Address counts as a proxy for independent customers.
05Analytical interpretation

A practical diligence framework

A reader evaluating any network in this category can work through these steps in order. Each step asks for evidence; none presumes an answer.

  1. 01

    Fix the period and units

    State the time window and the unit (jobs, leases, credits, hours) before comparing anything.

  2. 02

    Reconcile without double counting

    Match job records to invoices or on-chain settlement once. A job paid in fiat and settled on-chain is one job, not two.

  3. 03

    Ask about failures and retries

    How are failed, retried, or unhealthy jobs counted? Are they excluded from usage figures?

  4. 04

    Test payer independence

    Are the largest payers unrelated to the team, foundation, or major token holders?

  5. 05

    Trace subsidy funding

    Who funds grants, rebates, or credits, and are those payments shown separately?

  6. 06

    Watch behaviour after incentive changes

    Does usage repeat after a reward schedule or rebate program changes?

  7. 07

    Compare service quality

    Benchmark quality, latency, and reliability against conventional alternatives for the same workload.

Where a project does not publish evidence for a step, the answer is unknown. Missing public evidence should be recorded as unknown, not as zero, and not as an indication of fraud.

06Analytical interpretation

Limitations and disclaimer

This is an explanatory brief. It restates mechanism descriptions from three official documentation pages, checked on October 20, 2026, and proposes a framework for reading them. It reports no network metrics, prices, or results, and has not independently verified adoption, revenue, or profitability for any project. Protocol rules and documentation can change after the date checked.

Nothing here is a JARCH investment position, a recommendation to buy, sell, or hold any asset, or financial, legal, or tax advice. Mention of a network is illustrative of a mechanism, not an endorsement.

Primary sources

  1. [1]Akash Network documentation: Lease Management (provider operations)

    https://akash.network/docs/providers/operations/lease-management

    Provider operations guide. Used only for the described practice of reconciling on-chain leases against provider Kubernetes workloads. Checked October 20, 2026.

  2. [2]Render Network knowledge base: Burn-Mint Equilibrium

    https://know.rendernetwork.com/basics/burn-mint-equilibrium

    Describes how job payments and supplier emissions are handled. Used only for the mechanism description. Checked October 20, 2026.

  3. [3]Helium documentation: Data Credit

    https://docs.helium.com/tokens/data-credit

    Describes what Data Credits pay for and how they are created. Used only for the mechanism description. Checked October 20, 2026.

Back to research index