DePIN Projects: A Research Guide

State of DePIN in 2026

DePIN (DePIN = Decentralized Physical Infrastructure Networks, crypto-incentivized real-world hardware coordination across compute, wireless, mapping, and energy) sector market cap is roughly $9–10B; Bittensor, Render, Filecoin, and Akash generate the bulk of verifiable on-chain revenue. Concentration is extreme — three or four projects account for the majority of real customer spend. Last verified: 2026-05-27.

The original 2021–2022 thesis — "anyone can earn tokens by plugging in a $300 box" — is mostly dead. The DePIN sector market cap is in the $9–10B range per KuCoin, but the revenue story is highly concentrated. Bittensor generated $43M in Q1 2026 revenue from real AI subnet customers — the highest verified protocol revenue in the entire DePIN category. Akash (Akash is a decentralized GPU and CPU compute marketplace where providers list hardware and buyers deploy Docker workloads at spot-market prices) crossed $5M in Q1 2026 compute spend, a record. Helium Mobile (Helium is a decentralized wireless network that pivoted from hotspot rewards to a consumer MVNO — Helium Mobile — running on T-Mobile's wholesale network) reached 600K+ sign-ups by end of 2025. Everything else is earlier stage.

One important correction to circulating claims: the "$38M monthly revenue" figure attributed to Render is not verified protocol revenue. Render's actual 2025 annual protocol revenue was approximately $2.7M, with $38M being a forward projection based on planned integrations. The Render March 2026 monthly report's only concrete financial figure is a projected $4.3M first-year revenue from the proposed Salad subnet integration. Treat Render as a growth-stage network, not a revenue-mature one.

If you only remember one thing: DePIN is a real category, but revenue is concentrated in two or three leaders. Helium survived by pivoting from "hotspot rewards" to a consumer MVNO. Hivemapper survived by focusing on rideshare/delivery drivers and now counts Volkswagen's ADMT autonomous-vehicle division as an enterprise client. Almost everything launched in the 2022–2023 cohort either consolidated or quietly died.

Top 6 DePIN projects in 2026

Market caps as of late May 2026 (CoinGecko): Bittensor ($2.7B), Render ($1.2B), Filecoin ($795M), Akash ($260M), Helium ($139M), Hivemapper ($11M). The Bittensor/Render gap from the rest widened sharply. Last verified: 2026-05-27.

Bittensor (TAO) — deep dive

Bittensor is the only DePIN where the hardware is somebody else's problem, and the only one with $40M+ per quarter in verified AI customer revenue.

Contributors plug in their own GPUs (consumer RTX 4090s, A100s, H100 clusters) and run open-source subnet code. The protocol doesn't care which datacenter you're in; it cares whether your model wins the subnet competition. $43M Q1 2026 revenue came from 128 active subnets competing to provide AI services to real customers, with Nvidia and Polychain Capital among validators.

Best for

DePIN exposure to AI compute demand. The largest verified AI demand signal in crypto by a wide margin.

Hardware and economics

Subnets run on contributors' own hardware — primarily GPUs. No proprietary hardware required. Realistic returns: a competitive subnet-9 miner with an 8x H100 cluster (~$250K capex) can generate meaningful TAO emissions during favorable subnet rotations; a hobbyist RTX 4090 generally cannot.

The bear case

Subnet 1 (text-completion) has 100+ miners competing for a fixed reward pool — most earn near-zero after electricity. The "anyone can mine TAO" narrative is misleading for retail; in practice TAO mining at scale is professional ML-ops, not a side hustle. TAO drew down roughly 40% from its Q1 2026 high during the May 2026 AI-crypto sector correction.

Helium (HNT) — deep dive

Helium is the only DePIN that successfully pivoted from "hardware rewards" to "consumer product" — but its market cap (roughly $139M as of May 2026) is a fraction of peak estimates, even as subscriber growth continues.

Helium Mobile runs as an MVNO on T-Mobile's wholesale network with Helium-deployed CBRS small cells offloading traffic where coverage exists. 600K+ total sign-ups by end of 2025, per Messari and the Helium 2025 Year in Review. 100% of subscriber fees are converted to HNT burns, producing annualized subscriber revenue of ~$11M in Q4 2025.

Best for

Real-world consumer-facing DePIN. The consumer pitch — T-Mobile coverage at roughly half the price — is credible and generating real subscriber fees.

Hardware footprint

Hotspots ($300–500 each for IoT/LoRaWAN), small cells ($1,500–4,000 for 5G CBRS), plus T-Mobile wholesale roaming for coverage gaps. The HNT v1→v2 migration in 2022 stranded many original hotspot operators with obsolete hardware.

Trade-offs

Hotspot earnings collapsed from 2022 hype peaks. Most IoT hotspots now earn roughly $5–30/month (vs $1,000+/month at the 2021–2022 peak, when HNT briefly traded above $50). Only profitable in dense urban areas with good radio placement and meaningful device traffic.

The bear case

Helium's market cap compressed from over $400M (mid-2024 estimates) to roughly $139M by May 2026 — even as subscriber counts and HNT burn rates grew. The subscriber base is growing, but converting sign-ups to paying subscribers at scale, and retaining them against cheaper MVNOs, is the execution risk. If T-Mobile renegotiates the wholesale agreement, unit economics deteriorate.

Render (RENDER) — deep dive

Render's actual 2025 protocol revenue was roughly $2.7M annually — far below the $38M/month figure that circulated widely. The $38M was a forward projection; treat it as a target, not a fact.

The network processed 68M+ frames through March 2026, with AI tasks growing to 35–40% of job volume and token burns up 278.9% year-over-year. A March 2026 governance proposal (RNP-023) would integrate Salad's 60,000 consumer GPUs as a subnet — Salad estimates $4.3M in first-year revenue from that integration alone, which would materially move actual revenue if approved.

Best for

Anyone with idle high-end GPUs (RTX 4090, A6000, workstation cards) who wants to earn rendering fees. Also the best DePIN for actual workload buyers needing GPU rendering — Render has integrations with Octane, Cinema 4D, Blender, and Houdini that artists already use.

Hardware footprint

Consumer or workstation GPUs. NVIDIA partnership announcements in 2024 drove significant price action; the partnership itself is real but the price reaction was outsized relative to direct revenue impact. Current RENDER price (~$2.07–2.38 in late May 2026) is roughly 82% below all-time high — reflecting both the gap between growth narrative and current revenue, and broader market beta.

Revenue reality check

Network revenue is driven by the Burn-Mint Equilibrium: clients pay in RENDER, 5% is collected as a protocol fee, the rest is burned, and new tokens are minted to reward node operators. The measurable metric is token burns (1M+ RENDER burned in 2025) and frame volume — not USD revenue directly. The ~$2.7M annual revenue figure is the best available estimate as of late 2025.

Filecoin (FIL) — deep dive

Filecoin is the largest decentralised storage network by raw capacity, with storage utilization hitting ~36% in Q3 2025 — a meaningful improvement from the sub-5% real utilization widely cited in 2022–2023, and driven by verified deals rather than self-sealing.

Filecoin Onchain Cloud launched on mainnet in March 2026, adding verifiable on-chain proofs, smart-contract payments, two-copy replication, and S3-compatible APIs designed for AI agents and pipelines. Since testnet, Onchain Cloud attracted over 100 teams with 49.41 TiB across 478 datasets and 81 connected payment wallets — early numbers, but these are paid on-chain deals. Early investor vesting ends in October 2026, potentially shifting FIL supply dynamics.

Best for

Storage providers with cheap electricity and existing hard drive capacity. Customers needing very large cold storage at low cost. Filecoin Onchain Cloud for AI dataset archiving and verifiable data pipelines.

Hardware footprint

Commodity hard drives plus a sealing GPU for proof generation. Realistic provider economics: a $5K storage rig generates $50–200/month depending on deal flow and token price, with 12–60 month payback range.

The bear case

The 2026 strategy explicitly acknowledges a shift "away from growing supply and towards scaling demand" — confirming the prior problem: too much capacity, too few paid customers. The ~36% utilization in Q3 2025 represents a verified improvement, but the majority of capacity still earns from block rewards rather than storage fees. If emissions decline faster than paid deal growth, provider economics deteriorate.

Akash (AKT) — deep dive

Akash is the spot market for generic GPU compute. It crossed $5M in total Q1 2026 compute spend — an all-time record — driven by the Burn-Mint Equilibrium upgrade (live March 23, 2026) that directly links compute demand to AKT token burns.

H100 GPU pricing on Akash runs $1.20–1.80/hr compared to $4.50–5.50/hr on AWS — a 60–75% discount. The Homenode program (2026) extended supply beyond datacenters to individual GPU owners with RTX 4090s, RTX 5090s, and Quadro RTX 6000 Ada. AkashML processed 1.7 billion tokens per day on OpenRouter, outpacing Cloudflare in daily AI token throughput.

Best for

Developers running AI inference workloads who want major cost savings vs AWS/GCP/Azure GPU instances. The BME upgrade (Proposal 318, March 2026) makes every compute dollar a deflationary event for AKT — 53,520 AKT burned in the first 8 days post-launch.

Hardware footprint

Heterogeneous: H100 clusters in tier-3 datacenters down to single RTX 4090s via Homenode. Treat it like spot instances — cheap, interruptible; redundancy is mandatory for production workloads.

The bear case

$5M Q1 compute spend is significant for a decentralized network but tiny vs AWS. Revenue per day remains in the tens of thousands of dollars — orders of magnitude below centralized cloud. If AWS, Azure, and GCP cut H100 pricing by 40%+ as GPU supply normalises (possible by 2027), Akash's price advantage narrows materially.

Hivemapper (HONEY) — deep dive

Hivemapper has mapped 6M+ unique kilometres across 90+ countries (34% of world roads) and landed Volkswagen's ADMT autonomous-vehicle division as an enterprise client — but trailing 12-month revenue was ~$644K as of late 2025, making it the earliest-stage project in this roundup.

Bee Maps (formerly Hivemapper, Inc.) raised $32M and replaced the $589 upfront dashcam cost with a $19/month Bee Membership subscription (hardware + LTE + fleet software). The Hivemapper Foundation committed 10M extra HONEY in incentives for contributors driving uncovered roads through June 30, 2026. 75% of HONEY spent on map data is permanently burned, creating a deflationary mechanism tied to enterprise demand.

Best for

Mapping/automotive contributors who already drive 200+ miles/day for rideshare or delivery — the economics work best for drivers in uncovered road segments. Enterprise map buyers needing fresh street-level imagery at lower cost than HERE or TomTom.

The bear case

Revenue concentration risk: ~$644K trailing 12-month revenue is thin for a network with thousands of contributors. The Volkswagen ADMT deal and the $32M raise are forward-looking signals. Until enterprise deal flow scales, HONEY earnings depend heavily on token price, not underlying demand.

How to actually use DePIN

Two roles: contributor (earn tokens by deploying hardware) or consumer (buy services at below-market rates). Most retail confuses the two.

If you're a contributor:

1. Run the realistic ROI model first. Use current token price minus 50%, factor in electricity at your local rate, include hardware depreciation over 24 months. If the answer isn't clearly positive, skip it.
2. Pick a network where you have geographic or workload advantage. Helium hotspot in a dense urban area you can place on a rooftop — possibly. Helium hotspot in a suburban garage — almost certainly not. Hivemapper Bee dashcam if you drive 200+ miles/day for rideshare or delivery in an uncovered segment — yes. If you drive 20 miles/day for commute — no.
3. Treat token rewards as crypto exposure. Don't hold all rewards in the network token; the historical correlation between hardware deployment and token price has been brutal during bear cycles.

If you're a consumer:

1. Akash for cost-sensitive batch inference. H100-hour pricing on Akash clears at $1.20–1.80; AWS lists at $4.50–5.50. The savings cover operational overhead easily for non-production workloads. BME means your compute spend also burns AKT.
2. Render for GPU rendering. If you use Octane, Cinema 4D, or Blender professionally, the cost savings vs traditional rendering services are real, though project-level revenue is still modest.
3. Filecoin via Onchain Cloud for AI-compatible cold storage with verifiable proofs. Onchain Cloud's S3-compatible API lowers onboarding friction versus raw Filecoin deals.
4. Helium Mobile if you live in a T-Mobile coverage area and want a low-cost MVNO. The service is competitive at its price point; your subscription fee is converted to HNT burns.

Best DePIN by use case

AI exposure: Bittensor. Consumer product: Helium Mobile. GPU supply: Render. Storage/AI data: Filecoin Onchain Cloud. Cheap GPU compute (buyer): Akash. Mapping contributor: Hivemapper. Last verified: 2026-05-27.

• Best DePIN for AI compute demand exposure — Bittensor (TAO). Only one with $40M+/quarter verified AI customer revenue.
• Best DePIN for consumer-facing product — Helium Mobile (real MVNO, T-Mobile wholesale, 600K+ sign-ups).
• Best DePIN for GPU rendering supply — Render (RENDER). Growing frame volume; 68M+ total frames; AI inference growing to 35–40% of jobs.
• Best DePIN for storage with AI compatibility — Filecoin Onchain Cloud (S3 API, smart-contract payments, verifiable proofs; launched March 2026).
• Best DePIN for mapping/automotive contributors — Hivemapper (dashcam earnings, Volkswagen ADMT partnership, Bee subscription at $19/mo).
• Best DePIN for energy/grid — Daylight, Powerledger (early stage, treat as speculative).
• Best DePIN for wireless coverage — Helium (5G CBRS + LoRaWAN).
• Best DePIN for cheap GPU compute (buyer) — Akash (60–75% below AWS H100 pricing; BME-deflationary).
• Worst DePIN bet — Any project requiring proprietary hardware purchase plus token-incentive-only economics, with no measurable end-user demand. The 2022 graveyard is full of these.

DePIN hardware ROI: 2026 reality check

The 2022 payback math no longer works. Realistic 2026 ranges: Helium urban hotspot $10–50/mo, Hivemapper Bee dashcam varies with mileage, RTX 4090 on Render/Akash $50–300/mo. Always model 50% lower token price plus electricity. Last verified: 2026-05-27.

The 2022 hotspot/dashcam payback math no longer works. Current realistic ranges:

Assume 50% lower token price in your model. Add electricity at $0.12–0.20/kWh (US average). Add roughly 20% hardware depreciation per year. If the resulting payback exceeds 24 months, you're betting on token-price appreciation, not on the underlying service economics.

Risk summary

Five risks: token emissions and insider unlocks, hardware obsolescence, regulatory pressure on spectrum and imagery, AI-narrative beta correlation, and concentration in 3–4 winners. Last verified: 2026-05-27.

• Token-emission risk. Most DePIN tokens have aggressive emission schedules to bootstrap supply. Insider unlocks crash price independent of network growth. Filecoin's early-backer vesting ends in October 2026 — a supply event to watch. Always check the unlock schedule on Token Unlocks or DropsTab before sizing; a 25%+ supply unlock in the next 12 months is near-guaranteed sell pressure.
• Hardware obsolescence. Helium's HNT v1→v2 migration in 2022 stranded contributors who had paid $500+ for original hotspots that became incompatible with the new architecture. GPU generations turn over fast — an RTX 3090 that paid back in 2022 doesn't pay back in 2026 because newer cards compete it down. Buy hardware with at least one generation of headroom.
• Regulatory risk. Wireless DePIN faces spectrum regulation (FCC in the US, Ofcom in the UK). Hivemapper-style street imagery faces privacy regulation under GDPR — there has already been one EU enforcement query against a comparable dashcam-mapping operator. Energy DePIN faces grid-operator licensing requirements that vary by US state.
• Sector beta correlation. DePIN trades correlated with the AI crypto narrative. Both sold off sharply in May 2026 — TAO drew down roughly 40% from its Q1 high, RENDER roughly 35%, FIL roughly 45%, almost entirely in coordinated weeks. The "real revenue" argument helps over multi-year horizons but doesn't insulate from short-term beta selloffs.
• Concentration risk. Two or three projects generate the majority of measurable DePIN revenue. New entrants face a structural disadvantage — incumbents have hardware deployed, customer relationships, and token-network effects that compound. The 2024–2025 cohort has been a graveyard outside a handful of notable exceptions.

Looking ahead to 2027

A few specific signals worth watching:

• Helium Mobile paying-subscriber conversion. 600K+ sign-ups is not the same as 600K paying subscribers. The key metric to watch is annualized subscriber revenue vs total sign-up count. If paying-subscriber conversion converts the HNT burn rate upward materially in 2026, the consumer-DePIN thesis moves from "promising" to "proven."
• Bittensor subnet specialisation. The most interesting frontier is subnets doing things centralised AI labs can't — private model inference, regulated-domain ML, financial signal generation. If those mature in 2026–2027, TAO's revenue ceiling moves up. The leading indicator is subnet 8 (financial signals) and subnet 9 (fine-tuning) revenue share.
• Filecoin paid-storage ratio. Utilization hit ~36% in Q3 2025. If it reaches 50%+ through Onchain Cloud adoption, Filecoin's economics shift from primarily emission-driven to genuinely demand-driven. The Onchain Cloud's 100-team testnet cohort and 81 payment wallets are the early indicator. If it stalls under 40%, expect continued emission-driven supply growth without matching demand.
• Akash compute spend trajectory. $5M in Q1 2026 compute spend is the base. If BME-driven AKT burns create genuine deflationary pressure and compute spend doubles by Q4 2026, the pricing model is working. If hyperscalers (AWS, Azure, GCP) cut H100 hourly rates 40%+ as supply normalises, Akash's cost advantage narrows materially.
• Hivemapper enterprise conversion. The Volkswagen ADMT deal and $32M raise are signals; the outcome metric is whether trailing 12-month revenue crosses $5M by end of 2026. The Bee subscription model lowers contributor friction — whether it expands coverage fast enough to win enterprise map contracts at scale is the unanswered question.

---

Related: Best AI Crypto Tokens 2026 · What Is DeFi