Market cap $2.20 T −3.96 % over 24 hours24h volume $73.16 BBTC dominance 58.1 %ETH 10.0 %Fear & Greed 10 Extreme fearCoins 21,490

What is DePIN in crypto? Decentralised physical infrastructure explained

What is DePIN in crypto? How tokens pay for hotspots, dashcams and GPUs, who the paying customers are in 2026 and the sums to do before buying a device.

Intermediate 31 min readUpdated Read first: What is decentralized storage? IPFS, Filecoin and Arweave explained
Connected nodes, a satellite and a chip in front of a globe
Table of contents
  1. What is DePIN in crypto: a poll, a paper and two families
    1. Physical resource networks and digital resource networks
  2. The flywheel: tokens pay for supply before any customer exists
    1. Rewards first, customers later
    2. Burn and mint: how a dollar of usage reaches the token
    3. Where the flywheel breaks
  3. Proving that physical work happened
    1. Proof of Coverage paid hotspots for existing
    2. Photos, car data and GPU jobs are easier to check than a signal
  4. Physical networks in November 2026: Helium, Bee Maps, DIMO, GEODNET
    1. Helium: from window-sill hotspots to carrier Wi-Fi offload
    2. Bee Maps and Hivemapper: drivers are now paid in USDC
    3. DIMO and GEODNET: paying for use, not for plugging in
  5. Digital networks: GPUs by the hour and shared bandwidth
    1. Akash: a marketplace that is changing chains
    2. A GPU-hour on Akash and on AWS, 17 November 2026
    3. Render, io.net, Aethir and Grass
  6. How much do customers actually pay?
  7. What failed: Helium’s 2022 and the networks that closed
  8. Buying a device: do the payback arithmetic first
  9. Securities law and tax: what the SEC, the IRS and HMRC have said
    1. SEC v. Nova Labs ended without a ruling on the tokens
    2. A staff letter, a commissioner’s view and a stalled bill
    3. Rewards are generally income when received
  10. Six free checks before believing a DePIN project

A white box the size of a paperback sits on the window sill of a café and lets a mobile carrier’s customers use the café’s Wi-Fi. A camera on a delivery van’s windscreen photographs every street the van drives down. A gaming PC rents out its graphics card while its owner sleeps. Each owner is paid in a token, by a network that keeps its accounts on a blockchain, for a service that a stranger buys.

That scene is the short answer to the question “what is DePIN” in crypto. DePIN stands for decentralised physical infrastructure networks: projects that pay people, usually in a token, to install and run real equipment. One question sorts all of them, and it has nothing to do with technology. Who is the paying customer, and how many dollars did that customer pay last month?

Two dates frame this lesson. On 2 June 2026 a proposal was filed at Helium, the best-known network of the kind, to end rewards for hotspots that were merely switched on, and the same day the sale of its phone service to another company was announced. On 17 November 2026 the Akash marketplace updated its price page with an Nvidia H100 at an average of $2.73 per GPU-hour, while Amazon’s list price for the same chip, published the same day, worked out at $6.88. The first date shows what went wrong with the model. The second shows why people keep trying.

The previous lesson of this path, decentralised storage with IPFS, Filecoin and Arweave, dealt with the oldest family of these networks. This one covers the rest: wireless, mapping, vehicle data, positioning, computing power and bandwidth, with the state of each in November 2026 and the arithmetic to do before buying any device. Nothing here is a view on any token.

What is DePIN in crypto: a poll, a paper and two families

The word is younger than the idea. Helium was paying people in tokens to run radio hotspots from August 2019, and for three years nobody agreed on what to call that. Proposals included “proof of physical work” (PoPW), “token-incentivised physical infrastructure networks” (TIPIN) and “EdgeFi”. In November 2022 the research firm Messari put those three and a fourth, DePIN, to a poll on Twitter. DePIN won with 31.6% of the vote, 136 votes, according to an account of the poll published by Cointelegraph. A name now used across a whole industry was settled by a few hundred votes.

The definition stayed loose, and marketing teams have stretched it since. The most careful attempt to pin it down is a decision tree for classifying DePIN systems, published on arXiv on 29 January 2025 by Michael Andrew and Mark Ballandies. Their test has three parts. There must be a three-sided market (people who supply hardware, people who buy the service, and the protocol in between). The suppliers must be rewarded in a token. And the hardware has to be physically placed somewhere by those suppliers. A company that owns all its servers and merely accepts crypto payments fails the test.

By that reading, the first network to pay strangers in tokens for running hardware was Bitcoin itself, as our guide to Bitcoin mining explains, although what miners sell is the security of the ledger and not an outside service.

Physical resource networks and digital resource networks

The usual split, which this lesson follows, sorts projects by what the hardware produces. Physical resource networks sell something tied to a place: radio coverage in a building, street imagery of a given road, a position correction valid within a few dozen kilometres of a reference station, data from one particular car. Where the device stands decides what it is worth. Digital resource networks sell something interchangeable: a gigabyte stored, an hour of a graphics processor, a slice of a home internet connection. Location matters little, price matters a lot, and the competition is Amazon.

Storage is the oldest digital family, and the previous lesson covered it in full: with Filecoin a provider locks collateral and proves at regular intervals that it still holds the data, while IPFS on its own pays nobody. This lesson does not go over that ground again.

How does this differ from DeFi? A lending protocol moves tokens between smart contracts, and everything it does happens on the chain where anyone can check it. A DePIN delivers its product off the chain, in the physical world. The blockchain only keeps the accounts: who supplied what, who paid, who is owed. It is the same division of labour as in the lesson on what Web3 is, where a dapp keeps its money logic on a chain and nearly everything else somewhere else. The machines themselves are ordinary computers talking to each other, as described in the lesson on peer-to-peer networks and in the one on what a node does, and “node” in a DePIN white paper usually means “a device someone bought”.

The flywheel: tokens pay for supply before any customer exists

Rewards first, customers later

Building a mobile network the usual way costs billions before the first subscriber pays a cent. A DePIN turns the order round. The protocol creates a token and hands it out, on a schedule fixed in advance, to whoever installs equipment. Early participants receive a lot of tokens for little work, on the bet that the token will be worth something once customers arrive. The network gets its hardware without raising the money to buy it.

The intended sequel is a loop. Coverage attracts customers, customers pay for the service, their payments support the token’s price, a higher price attracts more hardware. Promoters call it a flywheel. The weak point is the first step. Token rewards can summon hardware within months. Nothing in the design summons customers.

Burn and mint: how a dollar of usage reaches the token

Customers do not want to pay in a token whose price moves 20% in a week. Most networks solve this with a burn-and-mint design: the service is priced in dollars, the customer’s payment destroys (“burns”) tokens worth that many dollars, and the protocol creates (“mints”) new tokens for suppliers on its own schedule. If more value is burned than minted, supply shrinks. If less, holders are diluted. The whole economy can be read from those two numbers.

Helium is the textbook case. Its documentation state that “One Data Credit equals $0.00001 USD”. Data Credits (DC) can only be created by burning HNT, the amount depending on the HNT price reported by an on-chain Pyth price feed, and they cannot be transferred once created. One dollar buys 100,000 DC whatever the market does. On the older sensor network a message of up to 24 bytes costs 1 DC, so a dollar pays for 100,000 small messages. With HNT near $0.229 at 09:52 UTC on 18 November 2026 (DefiLlama price feed), burning one HNT produced about 22,900 DC.

Akash adopted the same idea for cloud computing on 23 March 2026. A tenant burns AKT at the oracle price and receives ACT, which the project’s explainer describes as a “non-transferable, USD-pegged compute credit that exists solely to pay for infrastructure resources”. Providers are paid in freshly minted AKT at the price of the day they are paid. Akash gives its own worked example: a tenant needing $1,000 of credits when AKT trades at $1.14 burns about 877 AKT; if AKT has risen to $1.50 by settlement, the protocol mints only 667 AKT for the provider, and 210 AKT are gone for good. If the price falls instead, more tokens are minted than were burned.

Others use variants. Render prices each rendering job in dollars and burns that value in RENDER, with a 5% service fee, according to the methodology DefiLlama publishes for it. GEODNET, a positioning network, uses 80% of the fees it receives for access to its stations to buy back its GEOD token on the market and remove it from circulation, the remaining 20% going to its foundation (DefiLlama methodology, read on 18 November 2026).

A flame over stacks of coins beside a gear and banknotes
Burn and mint: the customer pays in dollars, tokens of that value are destroyed, and the protocol mints new ones for the people who run the hardware.

Where the flywheel breaks

It breaks when hardware arrives and customers do not. Then the tokens handed to suppliers are funded by nothing except the people buying the token, and each new device shrinks everyone else’s share. Our lesson on crypto Ponzi schemes describes schemes where early entrants are paid with the money of later ones; a DePIN with no customers is not automatically one of those, but its rewards have the same source.

Helium in 2026 shows both the problem and one attempted repair. Helium Improvement Proposal 149, filed on 2 June 2026 and passed by token-holder vote on 2 July, records that the data carried by the network had grown about fourfold in a year, from about 24,000 GB a day in June 2025 to about 97,000 in April 2026, while the HNT price “more than halved”. Usage up, token down. The proposal’s remedy has four parts: tie hotspot owners’ pay to the price carriers pay per gigabyte, stop paying for coverage as such, create a council to oversee spending, and mint about 141 million new HNT over 36 months for operations and growth. That last figure equals about 77% of the HNT in existence when the text was written, and lifts the maximum supply from about 206 million to about 347 million.

The first full quarter under the new rules is on record. Data Credits settled on the network came to about $718,000 in the third quarter of 2026, against $3.35 million in the second, a fall of about 79%, according to the tokenholder report that Blockworks Advisory prepares for Helium. Most of that fall is the price change: before the vote, Nova Labs, the company behind Helium, cut the rate charged to carriers from $0.50 to about $0.10 per GB, saying this reflected commercial offload rates, and two of the second quarter’s three months had still cleared at the old rate. The rest is traffic. Rewarded data fell 21%, to 6,977 terabytes, after a pilot with an intermediary that routed one US carrier’s traffic ended on 31 August. That traffic had been 47% of the network’s rewarded data in July and August.

On the other side of the ledger, the supplement began minting on 6 August at 193,550 HNT a day, and HNT issuance for the quarter rose from 2.02 million to 13.58 million tokens. Less money per gigabyte coming in, far more tokens going out. The report names what to watch from here: rewarded data per day against September’s 51.0 terabytes, and the number of carriers that pay.

Proving that physical work happened

A blockchain can check a signature. It cannot check that a radio in Lisbon is really in Lisbon, or really a radio. Every DePIN therefore needs a way of turning a physical fact into something a program can pay for, and the method chosen decides how much of the reward budget goes to cheats.

Proof of Coverage paid hotspots for existing

Helium’s first answer was Proof of Coverage. Hotspots sent each other radio challenges, and those that were “witnessed” by neighbours earned tokens whether or not any customer ever used them. This rewarded exactly what the protocol could measure, which was presence. People responded by faking it: hotspots that claimed locations they were not at, or clusters of devices in one room vouching for each other, a practice known as spoofing. By September 2022 the community’s denylist of suspected cheats held more than 70,000 hotspots, Forbes reported.

HIP-149 puts the verdict in one line: Proof of Coverage “pays for existing rather than for serving subscribers.” The proposal retires it on both of Helium’s networks. Owners of mobile hotspots now share rewards in proportion to the data that actually passes through their device. That is a much easier thing to prove, because a paying carrier is on the other end of the connection and has its own records.

Photos, car data and GPU jobs are easier to check than a signal

The general rule follows from that. A delivered product can be verified cheaply; the mere availability of something cannot. A dashcam network can compare a driver’s images with those of other drivers on the same road and with satellite positioning. A compute network knows whether a job finished and whether the tenant kept paying. A positioning network can compare each reference station’s readings with those of its neighbours. A network that pays for “coverage” or “uptime” that no customer consumes has to trust the device’s own word, and devices can be made to lie.

Some projects add trusted hardware: a device sold with a signing chip, so that data arrives signed by a known machine. This raises the cost of cheating and ties the participant to one manufacturer. All of them also need outside prices to convert dollars into tokens, which brings in the price feeds described in our advanced lesson on DeFi oracles. If an oracle misreports the token price for an hour, the burn-and-mint accounts are wrong for that hour.

Physical networks in November 2026: Helium, Bee Maps, DIMO, GEODNET

Helium: from window-sill hotspots to carrier Wi-Fi offload

Helium’s first network carried tiny messages from sensors over a radio standard called LoRaWAN. In 2023 it gave up its own blockchain and moved its accounts to Solana. In 2026 the business is something else again: Wi-Fi offload. When a carrier’s subscriber walks into a hotel, a stadium or a shop equipped with a Helium access point, the phone moves its data traffic onto the venue’s Wi-Fi, and the carrier pays per gigabyte for the relief. The posts on Helium’s blog through the autumn of 2026 are about casinos, hospitals, grocery chains and a university stadium, not about boxes on window sills.

Helium’s own network snapshot page, as updated on 17 November 2026, reports more than 40,000 active access points, daily users rising from 1.4 million in the second quarter of 2025 to 3.4 million in the first quarter of 2026, and traffic growing from about 24 terabytes a day in June 2025 to about 128 in May 2026, with carrier partners in the United States and Mexico. Those charts stop before the summer. The Blockworks report gives the later picture: daily users averaged 2.26 million in the third quarter and about 1.5 million in September, once the intermediary’s traffic had left, while access points at active sites, counting those reported by partners, rose by a third to about 53,300 as hotels, hospitals and a stadium put their existing Wi-Fi on the network. Helium ended the quarter with more antennas and fewer users than in the spring, and one carrier accounted for about 63% of the quarter’s rewarded data.

The phone plan sold under the name Helium Mobile no longer belongs to the project. Noble Mobile, the low-cost carrier of the former US presidential candidate Andrew Yang, announced its acquisition on 2 June 2026. Fortune reported that the terms were not disclosed, that Helium Mobile had almost 600,000 cumulative sign-ups (a count that includes people who later left) and that it was not profitable. Nova Labs kept the offload business, and the tokens stayed under the Helium DAO.

For a hotspot owner, the rule since HIP-149 is a band. The protocol aims to pay at least half of the carrier rate for each gigabyte carried and caps pay at three times that rate; with the rate at about $0.10, that meant a floor of $0.05 and a cap of $0.30 per GB. The text says the minimum “is what the protocol aims to deliver, not a hard guarantee every epoch”, and that after a sharp fall in the HNT price payments can fall short for one to two weeks. The band has already moved once. HIP-150, filed on 18 August 2026 and approved in a vote that closed on 31 August, raises the minimum from half of the carrier rate to 80% of it, or $0.08 per GB, for one year from activation, and lets Nova Labs grant chosen venues a multiplier of up to five on both what carriers pay and what the owner earns. Its text is frank about where owners stand: the top-up that holds pay at the minimum “has fired in every epoch” since HIP-149 went live. Hotspot owners have been paid the floor, not a market price. The Blockworks report shows the effect in tokens: owners received 0.137 HNT per rewarded gigabyte in June and 0.368 HNT in September.

Bee Maps and Hivemapper: drivers are now paid in USDC

Hivemapper built a street-level map from dashcams, paying drivers in a token called HONEY on Solana, which map customers then burned to buy imagery. The company that makes the cameras and sells the data, Bee Maps, changed the deal in 2026. In a post dated 16 July 2026 it announced that from August 2026 drivers sign in with an email address, that rewards are paid in the dollar stablecoin USDC through a Circle wallet, and that the wallet connections used for HONEY would be removed from its app. The same post states: “we have terminated our services agreement with the Hivemapper Foundation.” It adds that the USDC rewards are funded by Bee Maps and offered at its sole discretion.

In plain terms, a token network became a company paying contractors in dollars. For drivers that removes the price risk of the token. It also removes the “network owned by its contributors” from the product. DefiLlama’s fee data shows the consequence on the token side: about $420,540 of HONEY spent by map customers over the year to 18 November 2026, and practically nothing over the last 30 days.

DIMO and GEODNET: paying for use, not for plugging in

DIMO collects data from cars, with the owner’s permission, and sells access to developers and businesses. Its website on 18 November 2026 presents one API covering more than 50 car brands and does not mention a token on its front page. For years it paid a weekly token allowance to every connected car. That allowance, called baseline issuance, was discontinued with effect from 28 August 2026 under governance proposal DIP-12, according to the token transparency filing for DIMO published by Blockworks. The same filing records that 150 million DIMO were burned from the treasury under that proposal and that the network’s revenue now comes from credits developers buy to use the data. Another network that stopped paying for presence.

GEODNET is the case its supporters point to. Roof-mounted reference stations listen to navigation satellites and produce correction data that lets a tractor, a drone or a surveyor’s instrument fix its position to within centimetres. Customers buy a subscription, listed at $40 a month or $400 a year on the project’s site on 18 November 2026, and most of that money buys back and destroys tokens. DefiLlama recorded $960,471 of fees over the 30 days to 18 November 2026 and $9.9 million over the previous year. Even so, the token did not follow: Messari’s State of DePIN 2025 report, as summarised by Cointelegraph on 29 January 2026, found GEOD down 41% between December 2024 and December 2025 while on-chain revenue grew about 1.7 times.

Digital networks: GPUs by the hour and shared bandwidth

Akash: a marketplace that is changing chains

Akash Network is a marketplace where a tenant posts a workload and independent providers, from data centres to individuals, bid to run it. Its first-quarter report says the network had passed $5 million of compute spending since launch by the end of March 2026. That is a small sum against the tens of billions the large clouds bill each quarter, and the report does not publish a utilisation rate. Since early 2026 a programme called Homenode has let owners of consumer cards such as the RTX 4090 and 5090 sign up to serve AI inference from home, the GPU-on-the-desk part of the opening scene.

Two announcements from the autumn matter for anyone following it. On 26 October 2026 the core team published its proposal to abandon Akash’s own Cosmos blockchain and rebuild the marketplace as programs on Solana, with AKT becoming a Solana token through a one-way migration. Its stated reasons: about 100 validators lock up AKT only to secure a chain the marketplace does not need, and Helium, Render, io.net and Hivemapper already settle on Solana. Nothing has moved yet. Several governance votes are required and no cutover date is set. Then on 17 November 2026 the founder presented Akash Compute Forwards, a forward market in which a month of GPU time would be bought at a price fixed today. The post is plain about its status: “It’s a design. Nothing trades yet.”

A GPU-hour on Akash and on AWS, 17 November 2026

The table below is an EXAMPLE of the price gap these networks advertise, not a ranking. The Akash column is the average price per GPU-hour shown on Akash’s GPU pricing page, marked “Prices updated November 17, 2026”. The Amazon column divides the on-demand Linux price of an eight-GPU instance in the US East (N. Virginia) region by eight, from AWS’s public on-demand price list as published on 17 November 2026 at 23:50 UTC.

GPUAkash average, per GPU-hourAWS on-demand instanceAWS price per GPU-hourGap
Nvidia H100, 80 GB$2.73p5.48xlarge, $55.04 an hour for 8 GPUs$6.88About 60% lower
Nvidia A100, 80 GB$1.67p4de.24xlarge, $27.45 an hour for 8 GPUs$3.43About 51% lower
Nvidia A100, 40 GB$0.84p4d.24xlarge, $21.96 an hour for 8 GPUs$2.74About 69% lower
Nvidia RTX 4090, 24 GB$0.34Not offeredNoneNo comparison

The gap is real and so are its limits. Amazon’s price buys a whole machine with processors, memory, fast networking, support and a service-level agreement, and large customers pay well under list through reserved or spot pricing. An Akash average hides differences between providers, some of them a single machine in someone’s home. A list price compared with a marketplace average says nothing about whether the cheaper hour will still be running at the end of a three-week training job.

Render, io.net, Aethir and Grass

Render sells graphics rendering on GPUs supplied by independent node operators; DefiLlama counted $2.5 million of user spending over the year to 18 November 2026. io.net aggregates GPUs from independent suppliers. Its tokenomics page describes a system it calls the Incentive Dynamic Engine, live since 11 June 2026 according to a research note published by CoinDesk and commissioned by io.net, under which suppliers receive “USD-targeted” rewards and, once they are paid, “at least 50% of remaining revenue is burned”. Aethir rents GPU capacity to developers and shows the largest figure in the table further down, $61.0 million of service fees over the year according to DefiLlama, of which 20% is the protocol’s cut and 80% goes to the GPU providers. For all three, the usage numbers come from the projects’ own contracts and reporting, and the count of GPUs each one claims cannot be checked from outside.

Bandwidth networks ask for less still: no device, just an application that lets a company route requests through your home internet connection. Grass, the best known, sells the web data gathered that way to firms training AI models; the site follows its current reward programme on our page on the Grass Stage 2 USDC rewards. The cost to the user is harder to see than an electricity bill. Traffic you did not generate leaves from your address, and your contract with your internet provider is in your name.

One name is missing on purpose. Bittensor pays for machine-learning output and not for hardware, and belongs to the next lesson of this path, which covers crypto and AI.

How much do customers actually pay?

Less than the token prices suggest, and more than sceptics assume. Messari’s State of DePIN 2025 report, published in January 2026 and reported by Cointelegraph, put the sector at about $10 billion of token value and counted $72 million of on-chain revenue for the whole of 2025. The same report said about $1 billion was raised across the sector in 2025, up from $698 million in 2024, that the leading projects traded at 10 to 25 times revenue, and that tokens of the 2018 to 2022 generation stood 94% to 99% below their record prices. Investors put in roughly fourteen times what customers paid.

Anyone can redo part of that count on DefiLlama’s fees pages, a public dashboard that tracks what users pay to each protocol. The readings below were taken on 18 November 2026.

NetworkWhat customers buyFees, last 30 daysFees, last 12 monthsHow DefiLlama classes it
AethirGPU time for developers$5,505,327$61,038,767Developer tools
GEODNETPositioning corrections$960,471$9,905,650Oracle
RenderRendering and AI jobs$593,835$2,531,816DePIN
AkashCloud and GPU leases$241,772$2,115,804Chain
FilecoinNetwork fees$113,516$1,939,837Chain
HivemapperMap imagery paid in HONEYAbout $0$420,540DePIN

Four warnings go with the table. The categories are inconsistent: of six networks that everyone calls DePIN, the dashboard files two under that label. “Fees” means what users paid into the protocol’s contracts, which is not audited company revenue and can include payments from related parties. And coverage has holes. DefiLlama’s Helium entry counts only the HNT that Helium Mobile bought back with subscriber revenue, a programme paused on 2 January 2026 according to the dashboard’s own note, so it shows zero for a network whose Data Credit burn was $3.35 million in the second quarter alone. Aethir’s line, for its part, showed no fees at all for the last seven days at that reading, a sign that some figures arrive in batches. A zero on a dashboard is a reason to read the project’s own reports before drawing a conclusion.

What failed: Helium’s 2022 and the networks that closed

The 2021 and 2022 boom is the case every later project gets compared with. Hotspots sold in large numbers on the strength of token rewards, long before the network had customers to speak of. In September 2022 a Forbes investigation, republished by the magazine’s Argentine edition, reported that 30 wallets linked to employees, their friends and relatives and early investors had mined 3.5 million HNT, almost half of all tokens issued in the three months after the August 2019 launch, and that data moving over the network had brought in about $92,000 in the year or so to August 2022. Amir Haleem, Helium’s chief executive, told Forbes that none of those numbers felt unreasonable to him.

The customers were thinner than advertised, too. On 29 July 2022 Mashable reported that Lime, the scooter company Helium had presented for years as one of its biggest users, had done no more than a test of a month or two in 2019. “Helium has been making this claim for years and it is a false claim,” Russell Murphy, Lime’s senior director for corporate communications, told the site. Helium took the Lime and Salesforce logos off its website after the article appeared. The episode resurfaced in court: on 17 January 2025 the SEC sued Nova Labs, alleging among other things that it had told prospective investors that Lime, Nestlé and Salesforce were using the network “when those companies were not in fact using the network”, in the words of the agency’s litigation release. The outcome is in the legal section below.

Some networks simply closed. Wayru, which was building a Wi-Fi network from community-owned hotspots, told its community in December 2025 that Wayru Inc. and the Wayru Foundation would cease operations by the end of the year. The crypto news site Our Crypto Talk reported 30 December as the last day and said the team planned to publish its code and to repurpose or donate its equipment. The hotspots did not become worthless as objects, but whatever rewards they were bought for ended on a date chosen by someone else. That is the risk specific to DePIN hardware: the device outlives the reward programme.

Buying a device: do the payback arithmetic first

Vendors quote earnings in tokens, at the best recent price. The useful sum runs the other way: start from what a customer pays the network, work down to one device, and subtract the costs.

Here is a worked EXAMPLE using only Helium’s published rules and figures. It is an illustration, not a forecast and not what any owner earns. With the carrier rate at about $0.10 per GB, a hotspot owner’s pay per gigabyte carried had a $0.05 floor under HIP-149, has an $0.08 floor under HIP-150 for its first year, and is capped at $0.30. For traffic, take September 2026, when the network carried 51.0 terabytes of rewarded data a day according to the Blockworks report. Spread over the 53,300 access points counted at active sites, that is about 1 GB a day each. Spread over the roughly 13,000 “network authenticators” the report counts as carrying traffic on an average day of the quarter, it is about 4 GB. A stadium’s 2,296 access points sit behind a handful of authenticators, so neither average describes a box in a café.

Traffic carried by one devicePer year at $0.05 (HIP-149 floor)Per year at $0.08 (HIP-150 floor)Per year at the $0.30 cap
1 GB a day (about the average per access point)$18.25$29.20$109.50
4 GB a day (about the average per authenticator)$73$116.80$438
10 GB a day (a busy venue)$182.50$292$1,095

Now the costs. A device drawing 10 watts around the clock uses 87.6 kWh a year; at an example price of 17 cents per kWh that is about $15. Add the purchase price shown in the vendor’s shop that day, call it P, and any antenna, cabling or subscription. At the $0.08 floor, a device carrying 4 GB a day clears about $102 a year after electricity, so the payback time in years is P divided by 102: about two years for a device priced at $200, a figure chosen only to finish the sum. At 1 GB a day the same device clears about $14 a year, and the sum gives fourteen years. At the old $0.05 floor the two cases become about three and a half years and about sixty. Helium changed its reward rules twice in the summer of 2026 alone.

The income is paid in HNT, so its dollar value moves with the token between the day it is earned and the day it is sold. HIP-149 noted that the floor was “binding” with HNT near $0.27 in June 2026, meaning owners were at the bottom of the band, and on 18 November 2026 the token traded lower still, near $0.229. The higher floor is also temporary by design: HIP-150 sets it for one year, with a second year authorised in advance. And an average hides the distribution: a stadium carries terabytes, a flat on a quiet street may carry nothing.

An hourglass and a plug beside a falling line and a warning sign
In the worked example, a $200 device at the $0.08 floor pays for itself in about two years at 4 GB a day and in fourteen years at 1 GB a day.

The same sum works for any device network, and the same risks apply to all of them.

  • Rewards are diluted as devices join: the reward smart contract holding reserves of two or more tokens, supplied by users, that a decentralized exchange uses to execute trades automatically at a price set by a formula.">pool is split among more participants, while customer demand in one neighbourhood does not grow because a tenth hotspot arrived.
  • The rules change by vote. Helium ended Proof of Coverage, DIMO ended its weekly allowance and Bee Maps replaced a token with discretionary USDC payments, all within 2026, and a device owner has little say unless they also hold a large amount of the governance token.
  • Most devices work with one network only, sold by one or two approved manufacturers, so the resale value depends on the reward programme and not on the electronics.
  • A halving of the token price or of the reward rate halves the income, and both can happen in the same quarter.
  • A dashcam or a car dongle sends a company a continuous record of where you drive, and a bandwidth app sends other people’s traffic through your connection. Read what is collected and who can buy it before installing.
  • Projects with no network behind them sell “DePIN miner” boxes or rent “cloud nodes” with a promised daily yield. A fixed return quoted in advance is the mark of the schemes described in our guide to avoiding crypto scams.

Securities law and tax: what the SEC, the IRS and HMRC have said

SEC v. Nova Labs ended without a ruling on the tokens

The SEC’s January 2025 complaint had two parts: that selling hotspots which mined tokens amounted to an unregistered securities offering, and that Nova Labs misled equity investors about its customers. On 23 April 2025 the court entered a final judgment by consent. According to SEC Litigation Release No. 26291, it is limited to the misrepresentation claim about a private placement of preferred stock, and Nova Labs agreed to pay a $200,000 civil penalty without admitting or denying the allegations. The agency dismissed the other claims with prejudice, stating that the decision rested on its wish to reform its approach to crypto, “not any assessment of the merits of the claims”.

So no court has ruled that hotspot rewards are, or are not, securities. The registration claims against Nova Labs cannot be brought again. That binds the SEC in that case and nobody else.

A staff letter, a commissioner’s view and a stalled bill

On 29 September 2025 the SEC’s Division of Corporation Finance sent a no-action letter to the DoubleZero Foundation, whose 2Z token pays for and rewards services on its network, saying staff would not recommend enforcement over the programmatic distribution of its 2Z token to participants as described in the request. The letter applies to that project on those facts and does not cover earlier sales of the token. Commissioner Hester Peirce published a statement the same day arguing that such rewards fall outside securities law: “These tokens are neither shares of stock in a company, nor promises of profits from the managerial efforts of others.” It is the view of one commissioner, not a rule.

A statute would settle the matter, and there is none. The CLARITY Act, a market-structure bill, passed the House of Representatives by 294 votes to 134 on 17 July 2025. On 15 September 2026 the Senate voted 49 to 50 on ending debate on the motion to take the bill up, short of the 60 votes needed. In the European Union, tokens offered to the public fall under MiCA’s rules on white papers and marketing whatever hardware stands behind them, as our lesson on the MiCA regulation sets out.

Rewards are generally income when received

Tax authorities have not written rules for DePIN rewards by name, but the general ones point the same way. In the United States, the IRS’s frequently asked questions on virtual currency say that someone who receives virtual currency in exchange for performing services has ordinary income, and that “The amount of income you must recognize is the fair market value of the virtual currency, in U.S. dollars, when received.” A later sale then produces a capital gain or loss measured from that value. The details are in our guide to crypto taxes in the United States.

In the United Kingdom, HMRC’s Cryptoassets Manual draws the line at whether something was done in return. Tokens received “in return for, or in expectation of, a service” are subject to Income Tax, and a later disposal can bring Capital Gains Tax. That page is written about airdrops, so which category fits a device reward is a question for the taxpayer’s own facts; our guide to crypto taxes in the United Kingdom covers the framework. In both countries the practical point is the same: keep a dated record of every reward and its market value that day, because a token that later falls 80% does not cancel the income declared when it arrived.

Six free checks before believing a DePIN project

None of these needs a wallet, an account or a purchase.

  1. Find the paying customer by name. A carrier, an insurer, a mapping firm or an AI lab that confirms the relationship on its own website counts. A logo on the project’s site does not, as Lime showed in 2022.
  2. Look for the burn. If the service is paid through a burn-and-mint design, the tokens destroyed per month are public on the project’s explorer or dashboard, and they measure paid usage far better than a count of devices.
  3. Compare burn with issuance. Tokens minted for suppliers and for the team, minus tokens burned by customers, is the dilution holders absorb. Helium’s third quarter of 2026 is a worked case.
  4. Check the same network on DefiLlama’s fees pages, and read the methodology note to see what is and is not counted.
  5. Read the tokenomics and the schedule of each token unlock for investors and the team: a network can have real customers and still see its token sold steadily by early backers.
  6. Read the last three governance votes. They show who decides the reward rules and how often those rules change.

Two things will test the model again. Akash’s community has yet to vote on leaving its own chain for Solana. And the fixed-rate phase of Helium’s 141 million HNT supplement is written to last about twelve months from its start in the third quarter of 2026, after which the mint tapers over two more years: by late 2027 the reports will show whether carrier payments have grown into the supply created to chase them.

The next lesson of this path covers crypto and AI: agents that hold wallets, markets for data, and what the compute networks described here are used for.

Sources

  1. Helium HIP-149: Helium Utility and Emissions Realignment (GitHub) github.com
  2. Helium HIP-150: Mobile Deployer Prioritization (GitHub) github.com
  3. Helium documentation: Data Credits docs.helium.com
  4. Helium network snapshot (17 November 2026) helium.com
  5. SEC Litigation Release No. 26291, SEC v. Nova Labs, Inc. (24 April 2025) sec.gov
  6. SEC, Commissioner Hester Peirce: statement on the DoubleZero no-action letter (29 September 2025) sec.gov
  7. Akash Network: What Burn-Mint Equilibrium means for Akash akash.network
  8. Akash Network: GPU pricing (prices updated 17 November 2026, read 18 November 2026) akash.network
  9. Bee Maps: USDC rewards become the default in August 2026 (16 July 2026) beemaps.com
  10. Cointelegraph on Messari's State of DePIN 2025 (29 January 2026) cointelegraph.com
  11. Helium tokenholder report, Q3 2026 (Blockworks Advisory) blockworks.com
  12. Helium blog: HIP-150, higher earnings for deployers (4 September 2026) helium.com
  13. Fortune: Andrew Yang's Noble Mobile acquires Helium Mobile (2 June 2026) fortune.com
  14. SEC Litigation Release No. 26229, SEC v. Nova Labs, Inc. (18 January 2025) sec.gov
  15. SEC Division of Corporation Finance: no-action letter to DoubleZero (29 September 2025) sec.gov
  16. Clerk of the US House: Roll Call 199, H.R. 3633 CLARITY Act (17 July 2025) clerk.house.gov
  17. Decrypt: Clarity Act stalls in the Senate (15 September 2026) decrypt.co
  18. Forbes Argentina: Forbes investigation into Helium's early token distribution (26 September 2022) forbesargentina.com
  19. Mashable: Lime says it is not a Helium client (29 July 2022) mashable.com
  20. Akash Network: Why we selected Solana for Akash (26 October 2026) akash.network
  21. AWS: EC2 on-demand pricing (price list published 17 November 2026) aws.amazon.com
  22. DefiLlama: fees by protocol (read 18 November 2026) defillama.com
  23. Blockworks: DIMO token transparency filing blockworks.com
  24. Andrew and Ballandies: Are you a DePIN? A decision tree (arXiv, 29 January 2025) arxiv.org
  25. IRS: Frequently asked questions on virtual currency transactions irs.gov
  26. HMRC Cryptoassets Manual, CRYPTO21250 gov.uk
Coindaro Editorial Team

Who wrote this article

Editorial team

Guides, reviews and glossary checked against primary sources

  • Crypto education
  • Exchange reviews
  • Wallets
  • Glossary

The Coindaro editorial team researches and writes every article, review and guide published on Coindaro. Each fact is checked against primary sources such as official documents, regulators and project documentation, and every page lists its sources. We cover crypto markets, technology and regulation for readers in the United States and Europe, independently and without paid placements.

Keep learning