DePIN's Subsidy Test: Why Customer Revenue, Not Architecture, Decides Winners

Comparing DePIN networks as mining versus validation models misses the real divide: whether operators are paid by customers or by token issuance. On verified data, Livepeer's staking rewards were about 29 times its fees in Q1 2026, and Helium's token captured little of its consumer revenue. Moats come from real customers such as carriers and AI video apps, not from architecture.

Executive Summary

The "mining versus validation" framing groups networks that do not belong together. Firedancer is a Solana validator client written by Jump Crypto and has no token. EigenLayer is a security marketplace built on restaked ETH. Neither is a true DePIN peer of Helium, Render or Livepeer. A more useful test applies to every model: does operator income come from paying end-customers (burn, fees), or from token issuance and staking subsidy?

On verified data, Livepeer, the one network where both sides of the ledger are published, still runs on small demand-side revenue. Livepeer's Q1 2026 demand-side fees were $257.3K, against $7.4M of staking rewards in the same quarter. Helium has genuine carrier demand, but its token captured little of Helium Mobile's revenue, and Mobile has reportedly been acquired by Noble Mobile. The validation-style models have deep capital and credible security. In this evidence set they have no verified external revenue. EigenLayer's restaked TVL is roughly 31.5 times the market value of its own token, which says the market prices the security layer and the token very differently.

The thesis is that moats come from customers (carriers, AI video applications), not from the choice between a mining and a validation design. The 2026–2027 question is whether demand-linked burn and fee models can outgrow emissions. Our analysis, labelled as such where it is arithmetic or estimate, suggests that gap will not close within the next two quarters. Several data points we would want (Livepeer after Q1 2026, Helium quarterly data, Render revenue, EigenLayer AVS fees) were not found in a verified source, and we say so where they matter.

Market Context

Crypto's largest assets set the backdrop on October 27, 2026. Bitcoin traded at about $79,640 with a market cap of $1.60 trillion. ETH traded at about $2,309 (market cap about $282 billion), and SOL at about $99.70 (market cap about $58.8 billion). Prices are from DefiLlama's CoinGecko feed, timestamped October 27, 2026.

The DePIN-adjacent tokens are small by comparison. Render is the largest in this set and is still under $1 billion. Livepeer and Helium are each below $100 million. Akash, Aethir and Grass sit between roughly $110 million and $476 million. EIGEN, the token of the security marketplace, is below $200 million.

Two reference points on capital: Solana's chain TVL is $5.71 billion (DefiLlama, a DeFi measure rather than staked SOL) and EigenLayer's TVL is $5.95 billion. Both are larger than most DePIN token market caps, though TVL and market cap are different measures. For EigenLayer, capital depth is not the scarce input. The scarce input is paying demand, and that is what the rest of this report measures.

Why "Mining versus Validation" Is the Wrong Cut

The category error

Two of the five names in the suggested comparison, Firedancer and EigenLayer, are not the same kind of object as the other three. Helium, Render and Livepeer are token networks that sell a service to customers. Helium sells wireless coverage and carrier offload. Render sells rendering and compute jobs. Livepeer sells video transcoding and, increasingly, AI inference.

Firedancer is different. It is an independent Solana validator client written in C/C++ by Jump Crypto, and it has no token of its own. It addresses client diversity, meaning the concentration risk of a single codebase, not operator revenue. Firedancer is software with no token or revenue model of its own.

EigenLayer is also different. It lets stakers and operators secure additional services (AVSs) with restaked ETH. It is a marketplace for security, and its token is not the asset being restaked.

The labels also blur inside the group that looks homogeneous. Livepeer sells a compute service, yet its operator economics run through staking: participation was 52.2% in Q1 2026, and issuance responds to that participation rate. Architecture labels do not predict who pays whom.

The subsidy test

A better comparison asks three questions of any model:

Who pays the operator? Is income funded by end-customers paying fees, or by newly issued tokens and staking rewards?
Where does the payment go? Does customer revenue reach the token (through burns or fee distribution), or does it accrue to a company, an operator or a foundation?
Which line is growing faster? Is demand-side revenue closing the gap with the subsidy, or is the subsidy shrinking only because the token price fell?

We call the ratio of customer fees to token rewards the demand coverage ratio. It is only computable where both numbers are published for the same period, and in this evidence set that is true only for Livepeer. For the other models we can answer questions 1 and 2 qualitatively and must state plainly where question 3 cannot be answered.

Livepeer: Usage Is Growing, but the Subsidy Still Dwarfs Fees

Livepeer is the cleanest test case because Messari and Blockworks published both sides of the ledger in State of Livepeer Q1 2026 (published May 5, 2026). Demand-side fees were $257.3K, up 34.2% from $191.8K in Q4 2025. Staking rewards were down 44% quarter on quarter. Rewards were about 29 times fees, a direct measure of subsidy dependence. Put the other way, fees covered roughly 3.5% of rewards.

Usage outran revenue

Network usage rose 71.9% quarter on quarter, from 78.2 million minutes to 134.4 million, a quarterly high. Fees rose more slowly than minutes. Revenue per 1,000 minutes fell about 22%, from roughly $2.45 to $1.91, because throughput grew faster than fee capture.

That matters for the thesis. More usage at a lower unit price is not the same as more revenue. The network can be busier and still earn little.

The subsidy fell, but not necessarily for the right reason

Staking participation of 52.2% exceeded the 50% target, which lowered issuance. Messari attributes the 44% fall in rewards to a lower LPT price and reduced issuance. Rewards are measured in dollars, so the price-driven part of the decline is not a structural reduction in subsidy. If the LPT price recovers, dollar rewards would rise again with no change in issuance policy. By our arithmetic, the 44% drop implies Q4 rewards of roughly $13 million, which shows how large the subsidy was before the fall.

An illustration of the gap, not a forecast: if fees kept growing at the same quarterly rate and rewards stayed flat, fees would reach parity after between 11 and 12 quarters. Growth rates of that size rarely persist, and the arithmetic assumes rewards do not move.

What the AI revenue actually shows

Livepeer's AI-driven fees were $154.7K in Q1 2026, up 15.5%, and roughly 60% of total protocol revenue. The headline that "AI is 60% of Livepeer revenue" is accurate but small in absolute terms.

AI fees also grew more slowly than the total. It follows arithmetically that the quarter's growth came disproportionately from non-AI work. The Livepeer Foundation said it expects transcoding fees to decline as AI workloads take over. That implies a handover risk: the older revenue line may shrink before the newer one is large enough to replace it.

Quality of demand

The Foundation also said Q1 usage mixes production workloads with reserved capacity from ecosystem partners. It wants to grow the share driven by independent developer demand. Reserved partner capacity may not reflect organic demand, and the report does not quantify the split.

Governance also reactivated treasury rewards through LIP-101, restoring a 10% treasury cut, and launched a Protocol R&D SPE. Livepeer's own primer claims its GPU compute is 60–85% cheaper than centralized cloud. That is a project claim, not an independent benchmark, and the fall in revenue per 1,000 minutes, noted above, shows revenue has not kept pace with usage.

Valuation against revenue

LPT trades at $1.491 with a market cap of $74.11 million and 49.7 million tokens circulating. Annualizing Q1 fees is crude, but it gives about $1.03 million a year, which puts the market cap near 72 times fees. One quarter of staking rewards equals about 10% of market cap, though those rewards were measured at Q1 prices and not today's. The token is valued on expectation of growth, not on current fee income.

Helium: Real Carrier Demand, Weak Token Capture, and a Sold Consumer Arm

The model is demand-linked by design

Helium uses a Burn and Mint Equilibrium. Customers such as carriers pay in fiat, which is used to buy and burn data credits priced in HNT. On question 1 of the test, this is the right shape: operators are rewarded through HNT issuance, while customer spending is used to burn data credits, so customer demand removes tokens.

The customer base is real. As of January 2026, FalconX reported seven carriers connecting through Helium hotspots. Crypto.news later cited more than 138,900 community-operated hotspots. Helium Plus, launched in July 2025, lets businesses and public Wi-Fi providers connect existing routers by software, so new contributors need not buy specialized hardware. HNT's halving in August 2025 cut issuance.

Revenue and token demand are not the same number

FalconX, citing Blockworks, put network revenue across Helium Mobile and carrier offload at an annualized $35 million. It also noted that Helium Mobile revenue largely accrues to Nova Labs, not the token. Messari, cited by FalconX, said Helium Mobile accounted for only about 10% of data credit burns earlier in 2025 despite substantial subscriber revenue. Carrier offload therefore drives most token-level demand.

Crypto.news, citing Blockworks, reported in June 2026 that Helium burns roughly $50,000 of data credits a day. At a constant rate that is about $18 million a year. The $35 million annualized revenue and the burn rate come from different sources and dates, so we do not reconcile them or draw a conclusion from them. FalconX's own note that Mobile revenue largely accrues to Nova Labs is the supported point.

Mobile growth was strong. FalconX reported daily active users of as much as 2.5 million in late December 2025, about 10 times the prior year, and nearly 600K sign-ups at year end. Daily data transferred rose over 150% in the second half of 2025. The claim that Mobile subscriber growth proves HNT demand does not follow, because the subscriber revenue mostly bypassed the token.

The buyback episode

Nova Labs began HNT buybacks with Helium Mobile subscriber revenue in October 2025. On January 2, 2026, founder Amir Haleem announced the discretionary buybacks were paused, citing little market response. He said Helium plus Mobile generated $3.4 million in October. Data credits would continue to be burned for all carrier offload, so Helium did not end token burns. Only the discretionary, Mobile-funded purchases stopped.

The episode is informative. When the operator company tried to route consumer revenue to the token, the market response was small, by Haleem's own account. The evidence does not show why.

The sale of Helium Mobile

Noble Mobile, a US telecom startup founded by Andrew Yang, acquired Helium Mobile. Nova Labs says the Helium Network and HNT token are unaffected. Noble leases spectrum from T-Mobile and committed to using Helium Network connectivity. Crypto.news covered the deal on June 3, 2026. The terms, price and exact date come only from that coverage and are not confirmed against a primary announcement.

In our analysis, the sale turns a captive consumer arm into an arm's-length customer. That could favor the token if Noble's traffic is burned as data credits, and the commitment is a positive signal. How much volume Noble routes is undisclosed.

HNT trades at $0.3786 with a market cap of $75.15 million and 198.5 million tokens circulating. In June, crypto.news placed HNT support near $0.60. The price is now roughly 37% below that level.

What we cannot compute

This report has no verified figure for HNT emissions in dollars, so we cannot compute Helium's demand coverage ratio. Helium's quarterly data is not in our evidence set; our figures come from FalconX (January 2026), Haleem (January 2026) and crypto.news (June 2026). Its moat is visible in the carrier relationships and hotspot footprint. Whether that moat is large enough to outgrow emissions is not shown by public numbers we could verify.

Render: The Right Design, an Unverified Revenue Line

Messari's Render overview describes dynamic token issuance counterbalanced by a Burn-Mint Equilibrium, in which tokens spent on rendering and compute jobs are burned. Emissions therefore adapt to demand rather than being purely inflationary. On question 1 of the test, that is the correct structure.

The design is all we can verify. RENDER trades at $1.664 with a market cap of $862.74 million and 518.5 million tokens in circulation, the largest in this DePIN set. We found no verified source for Render's revenue, cumulative burn totals or node operator profitability.

Render's valuation therefore rests on expectations this report cannot verify. A burn-mint design is a promise that demand will matter. It does not show that demand is large.

Validation Models: Security Depth Without Verified External Revenue

Solana and Firedancer: resilience, not a revenue model

Firedancer's contribution is client diversity. The Frankendancer hybrid, which combines Firedancer's networking stack with Agave's runtime and consensus code, launched on Solana mainnet in September 2024. RPC Fast, a secondary source citing Altrady, reports that as of mid-2026 full Firedancer ran roughly 14% of staked SOL, with another 26% on Frankendancer. That is about 40% on Jump-derived code, versus 100% Agave-derived code in early 2024.

The shares are not settled. A Spark research note dated August 23, 2026 lists Firedancer-family stake share crossing 33% as a threshold to track, which implies the share was below 33% then. The 14% and 26% shares are approximate and come from a secondary source, so we do not state a precise current share. Spark also lists Alpenglow's mainnet activation, and we did not confirm its status.

None of this is revenue evidence. SOL trades at about $99.70, with a market cap of about $58.8 billion. Solana's chain TVL is $5.71 billion, and the Jito protocol's TVL is $1.04 billion. Our evidence set contains no verified figure for Solana fees, MEV and tips, vote costs or validator profitability. Solana cannot be scored on questions 1 and 3 of the subsidy test here, and we do not guess.

EigenLayer: a security marketplace with an unanswered fee question

EigenLayer activated mainnet slashing on April 17, 2025. CoinDesk reported that slashing, a core part of shared security, had been missing since the 2024 mainnet launch. It said EigenLayer held more than $7 billion in restaked assets at the time.

Eigen Labs' slashing release gives operators opt-in participation, Operator Sets and Unique Stake Allocation, which isolate slashing risk per AVS. Those are real security features. EigenLayer's documentation says AVS rewards go to stakers and operators, and that Rewards v2 (operator-directed rewards) implements EigenLayer Improvement Proposal-001 (ELIP-001).

The documentation we cite does not say whether AVS rewards are funded by end-customer fees or by token incentives. That is the missing line in the subsidy test.

The capital figures are large and the token figures are small. EIGEN trades at $0.1934 with a market cap of $188.81 million and 976 million tokens circulating. Restaked capital, measured by the DefiLlama TVL cited above, is therefore about 31.5 times the token's market value. For scale, the ETH being restaked has a market cap of about $282 billion, so EigenLayer's TVL is about 2% of it.

That TVL compares with the more than $7 billion CoinDesk cited in April 2025, though the two numbers may differ in method. We use only the live DefiLlama TVL figure.

What the gap does show is that, as with Helium, security or activity and token value capture are separate questions. The restaked asset is ETH, not EIGEN. A large pool of staked capital does not show that anyone pays for the security it provides.

Data: The Subsidy Test Applied Across Models

The table scores each model on the three questions. "Not verified" means we found no source in the evidence set. Livepeer's coverage ratio is derived from Messari's Q1 2026 report. All other figures are dated in the text above.

Network	Type	Who pays operators (verified)	Verified end-customer revenue	Token capture	Demand coverage
Livepeer (LPT)	Compute service, staking-based	Staking rewards plus fees	Fees of $257.3K in Q1 2026	Fees and rewards on protocol; 10% treasury cut	Fees about 3.5% of rewards
Helium (HNT)	Wireless network, BME	Issuance plus burned data credits	~$35M annualized across Mobile and offload (Blockworks via FalconX)	Mobile revenue mostly to Nova Labs; offload burned	Not computable, emissions in USD not verified
Render (RENDER)	Compute marketplace, BME	Dynamic issuance offset by burns	Not verified	Burns tied to spent jobs by design	Not computable
Solana (SOL)	L1 with validators	Not verified	Not verified	Not applicable to Firedancer (no token)	Not computable
EigenLayer (EIGEN)	Restaking marketplace	AVS rewards to stakers and operators	Not verified	Restaked asset is ETH, not EIGEN	Not computable; funding source of AVS rewards not stated

The market-cap table gives the scale of the tokens at issue. Figures are from DefiLlama's CoinGecko feed on October 27, 2026 and are listed separately, not summed.

Token	Price	Market cap	Circulating supply
RENDER	$1.664	$862.74M	518.5M
GRASS	$0.6800	$475.67M	699.5M
AKT	$0.6712	$201.50M	300.2M
EIGEN	$0.1934	$188.81M	976.1M
ATH	$0.005501	$110.70M	20.12B
HNT	$0.3786	$75.15M	198.5M
LPT	$1.491	$74.11M	49.7M

Akash, Aethir and Grass have no demand-side data in the evidence set. They appear here to show scale only.

What Would Change the Thesis

Several observable events would weaken the conclusion that demand-side revenue is still small next to subsidy. Each is specific and checkable:

Livepeer Q2 and Q3 2026 reports showing fees growing faster than usage and the coverage ratio moving materially above 3.5%. We did not find these reports in our evidence set.
A published split of Livepeer usage between reserved partner capacity and independent developers, which the Foundation has said it wants to grow.
Helium quarterly data, including the volume Noble Mobile routes as data credits and HNT emissions in dollars. That would let the coverage ratio be computed.
Primary-source Render revenue and burn data, verified against Render's documents or on-chain records.
Disclosure of what funds EigenLayer AVS rewards, split between customer fees and token incentives.
Firedancer-family stake share confirmed by validator dashboards. This would settle the client diversity question but would not answer the revenue one.

Risks

Subsidy dependence persists (High). Triggered if fees stagnate while token prices fall. Rewards measured in dollars would then shrink, operator income would weaken, and security or service quality could suffer.
Data quality (High). Our evidence set lacks Livepeer data after Q1 2026, Helium quarterly data, Render revenue and Solana validator economics. Conclusions rest on a partial view, and aggregator claims we excluded may be right or wrong.
Token capture failure (Medium). Triggered if customer revenue keeps accruing to companies (as with Helium Mobile and Nova Labs) instead of reaching burns or fees. The buyback pause followed what Haleem described as little market response.
Demand quality at Livepeer (Medium). Triggered if reserved ecosystem-partner capacity lapses, or if transcoding fees fall faster than AI fees grow.
Unit price erosion (Medium). Livepeer's revenue per 1,000 minutes fell about 22% in a quarter. If that continues, usage growth will not translate into fee growth.
Dependence on Noble Mobile at Helium (Medium). Terms of the commitment are undisclosed. Triggered if Noble routes less traffic through Helium than expected.
Restaking revenue unproven (Medium). Triggered if AVS rewards prove to be funded mainly by token incentives, which would make restaking another subsidy model.
Validator client transition (Low). Solana's client mix and Alpenglow status are unclear in our sources. Risk rises if a bug in a minority client causes an outage, though nothing in the evidence points to one.

Outlook

Our three to six month view runs to roughly April 2027. The probabilities below are our estimates, not facts, and rest on one argument: demand-side revenue is growing from a very small base, while subsidies were large at last measurement.

Base case (60%). Customer revenue keeps growing at the leading networks, but no network with verified data shows fees close to token rewards. Livepeer's coverage ratio improves but stays in single digits. Helium's carrier-driven burns continue without a visible step-up. Render, EigenLayer and Solana still lack verified external revenue. Tokens trade on expectations.

Upside case (15%). New quarterly reports show demand coverage improving sharply. For Livepeer that means fees above roughly 10% of rewards, or a published independent-developer share that rises. Helium discloses meaningful Noble volume. A primary-source Render revenue figure supports its valuation. We rate this lowest because closing the gap requires fee growth far above the quarterly rate seen in Q1 2026.

Downside case (25%). Transcoding fees fall faster than AI fees grow, unit prices keep dropping, and token prices weaken further. Rewards in dollars fall, but customer revenue does not rise to replace them, so operator income declines. The risk is higher than the upside because two of the data lines (price per minute, HNT versus June support) already moved the wrong way.

Takeaways

Traders: Do not treat DePIN tokens as demand proxies. Use the burn rate, quarterly fee data and the coverage ratio. Livepeer and Helium each have market caps near $75 million, so liquidity and reporting gaps can move prices.
Builders: Customers are the moat. Publish customer revenue next to token rewards each quarter, report how much usage is reserved versus independent, and design for revenue to reach the token.
Investors: Replace the architecture question with the subsidy test. For EigenLayer and Solana, ask what customers pay and who receives it before treating capital depth as a revenue signal. Treat any DePIN comparison that includes Firedancer as a category error.