io.net

What it does

io.net is a GPUGPUGraphics Processing Unit. Originally designed to render video game graphics, GPUs turned out to be exceptionally good at the massively parallel math that AI models need. Modern AI training and inference runs almost entirely on GPUs.Like a factory with 10,000 workers doing the same simple task in parallel, versus a CPU which is more like 10 workers each doing different complex tasks. AI training involves doing simple math a million times per second on a million numbers, which is exactly what the GPU factory is designed for.Read more → marketplace built on Solana. You supply GPUs, you rent GPUs, and io.net sits in the middle taking a cut. The “decentralised” part is that GPUs come from independent suppliers across 138 countries. Everything else, the orchestration, matching, billing, verification, and the node software itself, is run by io.net.

The architecture has seven layers, from a ReactJS frontend down through FastAPI and GraphQL APIs, PostgreSQL and Redis databases, RabbitMQ task queues, and a Kubernetes orchestration layer running IO-SDK (a proprietary forkForkA point at which a blockchain or its software splits into two separate paths. Forks can be temporary (two valid blocks compete and one wins), planned (a software upgrade), or contentious (a community split into two chains).Like a road that splits in two. Sometimes the split is just a temporary detour and traffic merges back. Sometimes it's a permanent fork where different drivers go different ways and never meet again.Read more → of Ray). GPU suppliers run the closed-source IO Worker binary, pass Proof-of-Work benchmarks and Proof of Time-Lock verification, stake IO tokens as collateral, and receive hourly block rewardsEmissionsNew tokens created and distributed by a blockchain protocol over time as rewards to validators, stakers, or miners. Emissions fund network security and participation at the cost of diluting existing holders.Like a company that pays employees partly in newly printed shares. Every year the total number of shares goes up, which means existing shareholders own a slightly smaller slice of the same company unless the company grows faster than the printing.Read more → plus job payments. Consumers deploy GPU clusters via IO Cloud, or access 30-plus open-source AI models through the IO Intelligence APIAPIApplication Programming Interface. A structured way for one piece of software to talk to another. In DeAI, APIs let applications request inference from a model without running the model themselves.Like a waiter in a restaurant. You don't walk into the kitchen and cook your own meal. You tell the waiter what you want, they tell the kitchen, the kitchen cooks it, and the waiter brings it back. The API is the waiter.Read more → (OpenAI-compatible endpoint).

The project launched on Binance Launchpool in June 2024 after raising $30 million in a Series A led by Hack VCVCVenture Capital. Private investors who fund projects at an early stage in exchange for equity or token allocations. VC rounds are typically pre-launch, at steep discounts to any future public price, with multi-year vesting.Like angel investors in a startup who buy shares before the company goes public. They take more risk because the company might fail, so they get a better price. Once the company IPOs they can sell, and the public market pays whatever price it thinks is fair.Read more → at a $1 billion valuation. Backers include Multicoin Capital, Anatoly Yakovenko (Solana founder), OKX Ventures, Delphi Ventures, Animoca Brands and others. The team is estimated at roughly 100 people across New York and San Francisco.

The founding story is complicated. Ahmad Shadid founded io.net in 2022, resigned as CEO two days before the tokenTokenA digital unit of value or access rights tracked on a blockchain. Tokens can represent ownership in a project, a right to use a service, a share of future revenue, or simply a tradable asset with no underlying claim.Like a physical poker chip a casino issues. The chip itself has no value. What makes it worth something is what it lets you do at the casino, what the casino has promised, and how much other people will pay you for it.Read more → launch in June 2024 amid allegations about his background in Arabic-language crypto projects and questions about inflated GPU metrics. Co-founder Tory Green (Stanford, West Point, ex-Merrill Lynch, ex-Disney strategic planning) took over as CEO, then moved to chair of the io.net Foundation in April 2025. Gaurav Sharma (ex-Binance VP of Engineering) was appointed CEO. Three CEOs in under two years.

Value proposition

Cheaper GPUs for AI workloads than AWS, GCP or Azure. io.net claims 70-90% cost savings. Whether you actually get that depends on the workload, the GPU model and availability.

What is interesting is the revenue growth, according to figures io.net has reported publicly. Q4 2024 network revenue was $3.1 million. Q1 2025 was $5.7 million, an 82.6% quarter-on-quarter increase. Annualised, that puts the run rate north of $20 million. The company also reports 7.1 million compute hours delivered in January 2025, up from 1.4 million in September 2024. These figures come from io.net itself; no independent audit of revenue or compute hours has been published.

The confidential computeConfidential ComputeHardware-enforced computation where data and code are encrypted in memory and only the authorised application can access them. The machine's operator cannot read what the application is doing even though they own the machine.Like renting space in a bank vault. The bank owns the building and runs the security, but what you put in the vault is invisible even to the bank staff. Only you have the key.Read more → offering is a meaningful differentiator. Intel TDX and NVIDIA H100/H200/B200 GPUs provide hardware-based data protection during processing. Users can verify attestationAttestationA cryptographic proof that a piece of code is running on a specific hardware enclave in an unmodified state. Attestation lets remote users verify that a service is genuinely running what it claims to be running.Like a tamper-evident seal on a medicine bottle. The seal itself doesn't make the medicine safe, but it gives you a way to verify that nobody opened the bottle and swapped the contents before you bought it.Read more → reports. TrainingTrainingThe one-time process of teaching a neural network to perform a task by showing it massive amounts of example data and adjusting its internal weights until the outputs are good. Training builds the model; inference uses it.Like the years an apprentice spends learning a trade. You don't see any of the actual work, just thousands of repeated mistakes gradually becoming competence. By the end, the apprentice can do the job. The training was invisible, but the skill is now permanent.Read more → data, model weightsParametersThe internal numbers (weights and biases) inside a neural network that get adjusted during training. A 70-billion-parameter model has 70 billion adjustable internal numbers encoding everything it has learned.Like the synapses in a human brain. Each parameter is a tiny dial that gets nudged a little during training. With enough dials, the network can represent surprisingly complex patterns. The total parameter count is roughly how much "brain" the model has.Read more → and architectures stay encrypted. For AI workloads handling sensitive data, this matters.

IO Intelligence provides an OpenAI-compatible API with access to 30-plus open-source models (Llama, DeepSeek, Qwen and others). The Co-Staking Marketplace, launched in February 2025, lets token holders stake alongside GPU operators and share blockBlockA batch of transactions added to a blockchain at a set interval. Each block cryptographically links to the previous one, creating an append-only chain that can't be rewritten without redoing all the work since.Like a page in a ledger. Every page has a fixed number of entries, every page references the previous page, and once a page is filled and signed off it can't be edited without visibly invalidating every page that came after. The chain is just a very long series of these sealed pages.Read more → rewards without running hardware.

The counter-narrative is equally clear. Strip away the DePINDePINDecentralised Physical Infrastructure Networks. Protocols that use token incentives to coordinate real-world physical infrastructure like GPU compute, wireless networks, storage, mapping sensors, or bandwidth.Like crowd-sourced ride-sharing but for physical hardware. Uber incentivises drivers with dollars. DePIN incentivises hardware operators with tokens. The network grows because individuals choose to contribute capacity in exchange for rewards.Read more → branding and you have a company that runs a GPU marketplace with token incentives. Akash does the same thing but with open-source code, permissionless deployment, on-chain governanceDAODecentralised Autonomous Organisation. A way to coordinate decisions and manage a treasury using token-weighted voting instead of a traditional company structure. Token holders propose and vote on changes directly.Like a shareholder-run company where every shareholder can vote on every decision, the votes are public, and the company can't do anything the shareholders don't approve. The coordination is messier than a normal company but nobody has unilateral control.Read more → and consistent leadership since 2015. Akash also undercuts io.net by roughly $0.07 per GPU-hour on H100s. On the other side, centralised providers like Lambda Labs and Vast.ai offer GPU marketplaces without the token overhead.

io.net’s positioning is stuck in between: not decentralised enough to win on the sovereignty thesis, and not reliable enough to win enterprise workloads from AWS. The revenue growth suggests they are finding customers regardless, but the long-term competitive moat is unclear.

Tokenomics

IO is an SPL token on Solana. Max supply: 800 million. Circulating supplyCirculating SupplyThe number of tokens currently in circulation and tradeable on the open market. Differs from total supply (which includes locked or unvested tokens) and max supply (the upper limit, if there is one).Like the number of cars on the road today versus the number ever produced. Some are in showrooms, some in junkyards, some still at the factory. Only the ones on the road count toward what people are actually driving.Read more →: approximately 301 million (37.7%) as of March 2026. The remaining tokens unlock through emissions and vestingVestingA schedule that locks up tokens allocated to insiders, investors, and team members, releasing them gradually over months or years. Vesting prevents insiders from dumping on public buyers immediately after launch.Like a new employee's stock options at a startup. You don't get all the shares on day one. They unlock over four years so you stick around and do the work rather than cashing out and leaving.Read more → schedules running to 2028.

Distribution:

• Emissions (GPU supplier and staker rewards): 37.5% (300M IO, originally distributed hourly over 20 years, disinflationary starting at 8% annual rate, decreasing roughly 1% per month, now transitioning to monthly emission epochs)
• R&D and Ecosystem: 16.0% (30% unlocked at TGETGEToken Generation Event. The moment a project's token first becomes tradeable. TGE is when vesting clocks usually start, when liquidity hits exchanges, and when public price discovery begins.Like the IPO day for a startup. Everything that happened before TGE was private valuations and paper agreements. Everything after is the public market deciding what the thing is worth in real time.Read more →, linear vest over 35 months from January 2025)
• Seed investors: 12.5% (13-month cliffCliffA waiting period at the start of a token vesting schedule during which no tokens unlock at all. After the cliff ends, tokens begin releasing according to the vesting schedule.Like a probationary period at a new job. You don't get your stock options on day one. You wait 12 months to prove you'll stick around, then everything starts unlocking normally.Read more →, then linear vest over 23 months; unlocks began July 2025)
• Core contributors (team): 11.3% (13-month cliff, then linear vest over 35 months; unlocks began July 2025)
• Community (airdrops, incentives): 10.0% (45% at TGE, linear vest over 41 months)
• Binance Launchpool: 2.5% (fully unlocked at launch)

Insider allocation (seed plus team plus Series A) totals roughly 34%. The Binance Launchpool provided broad initial distribution, and 50% of total supply is earmarked for community over time, but the “community” emissions primarily benefit GPU suppliers, not passive token holders.

IDE tokenomics overhaul (announced December 2025)

io.net announced the Incentive Dynamic Engine (IDE), a significant restructure of how emissions and revenue flow through the token. The core changes:

• Demand-driven emissions: Replaces the fixed disinflationary schedule with a system that adjusts rewards based on actual network demand. Suppliers earn more when the network is busy, less when it isn’t. This is a meaningful shift from “pay GPUs to exist” to “pay GPUs to work.”
• Dual-vault mechanism: A Reward Vault funds supplier incentives. A Fee Vault collects platform revenue. The separation creates clearer accounting between emission subsidies and earned revenue.
• 50% buyback-and-burn of revenue: Half of all platform revenue collected in the Fee Vault is used to buy IO on the open market and permanently burnBurnPermanently removing tokens from circulation by sending them to an address that no one controls. Burns reduce total supply, which (all else equal) makes each remaining token worth more of the network's value.Like a company buying back its own shares and shredding them. The company's total value stays the same, but each remaining share now represents a slightly bigger slice of that value.Read more → it. This replaces the previous, vaguer burn mechanism with a structured, auditable flow.
• Target 50% supply reduction: The IDE targets reducing the remaining 300 million emission pool by half, effectively cutting long-term dilution from 300 million tokens to 150 million. If implemented, this materially changes the supply trajectory.

Implementation is planned for Q2 2026. As of March 2026, the IDE remains an announcement, not a deployed system. The documentation has been updated to reflect the monthly emission epochEpochA fixed-length period in a Proof of Stake blockchain during which the validator set is stable and rewards are calculated. Epochs are the natural unit for staking rewards and network state changes.Like a payroll period at a job. Within the period, your role is fixed and your pay is calculated based on hours worked. At the end, the period closes, paychecks are issued, and a new period begins with potentially different conditions.Read more → transition, but the dual-vault and demand-driven mechanisms aren’t yet live. The design is sound on paper: demand-driven emissions plus structured burns plus supply reduction is exactly what the token needed.

But io.net has a track record of bold announcements followed by delayed or modified execution. Score this based on what’s deployed, not what’s promised.

Existing burn mechanism and price action

The existing burn mechanism (at least 50% of platform revenue permanently burned) continues to operate. At io.net’s reported annualised revenue of roughly $20 million, that implies approximately $10 million of annual buy pressure against insider unlocks — assuming those revenue figures are accurate. If the IDE’s enhanced burn mechanism and supply reduction go live, the maths shifts considerably in holders’ favour.

IO is 98% below the all-time high of $6.43, set on 12 June 2024, the day after listing. The all-time low was hit in February 2026. Listed on Binance, Coinbase, Gate, Bybit, KuCoin, WhiteBIT and others.

The token was valued at $1 billion at the Series A. Investors who bought at that valuation are deeply underwater.

How to participate

Supply GPUs. Install the closed-source IO Worker software on NVIDIA GPUs (minimum 4GB VRAM, 100 Mbps down / 75 Mbps up). Stake a minimum of 200 IO per chip (adjusted by GPU model multiplier). Pass a 12-hour initial stress test and maintain minimum 5-hour daily uptime. Earn hourly block rewards plus compute job payments. Higher-end GPUs (A100, H100, RTX 4090) earn significantly more.

Co-stake IO tokens. The Co-Staking Marketplace (February 2025) lets token holders stake alongside GPU operators and share block rewards without running hardware. Average APR is roughly 8.7%. Fourteen-day unstaking cooldown during which tokens earn no rewards.

Rent compute. Deploy GPU clusters via IO Cloud for AI/MLMLMachine Learning. The branch of AI where systems learn patterns from data instead of being explicitly programmed with rules. Modern AI (LLMs, image generation, recommendation systems) is almost entirely machine learning.Like teaching a child to recognise dogs by showing them thousands of pictures of dogs, instead of writing down a precise rulebook for what makes a dog. The child learns the pattern from examples rather than from instructions.Read more → workloads. Access the IO Intelligence API (OpenAI-compatible) for inferenceInferenceRunning a trained AI model to produce an answer. Inference is what happens when you type a prompt into ChatGPT and get a response. The model takes your input, computes a best guess, and returns it.Like asking an expert for their opinion. The training was the decades they spent becoming an expert. The inference is the 30 seconds it takes them to answer your specific question.Read more → on 30-plus open-source models. Confidential compute VMs available for sensitive workloads. Pay in IO or fiat.

Build. IO Intelligence APIs, IO-SDK (Ray fork) for distributed computing, Python library. Documentation is adequate but lacks depth compared to major cloud providers.

io.net earnings: what GPU operators actually make

Before buying hardware for io.net, understand the honest picture. io.net earnings depend on three variables that the marketing tends to collapse into one: the hourly block reward rate (set by emissions), the frequency of actual paid compute jobs (set by demand), and the network-wide utilisation rate (which is far lower than headline GPU counts suggest). The first two are controllable by io.net and the market. The third is the one that kills most would-be suppliers’ return calculations.

The three revenue streams

GPU suppliers on io.net earn from:

• Hourly block rewards paid in IO for keeping the worker node online and passing uptime checks. Block rewards continue whether the GPU is doing actual work or sitting idle. They come from the 300 million IO supplier pool under a disinflationary schedule decreasing roughly 1% per month. The Incentive Dynamic Engine restructure (announced December 2025, implementation planned Q2 2026) will shift this to demand-driven rewards, but that system is not yet live as of April 2026.
• Compute job payments paid per rental when a customer actually books the GPU for AI training, inference, or rendering. Higher-end GPUs (A100, H100, H200, RTX 4090) receive proportionally more job payments because they match more rental queries. Consumer cards still earn block rewards but rarely pick up paid jobs.
• Co-staking participation. The Co-Staking Marketplace (February 2025) lets IO token holders stake alongside operators and share block rewards. Operators accept a slightly lower effective rate in exchange for a larger bonded stake. Co-stakers earn approximately 8.7% APR on average. Operators use this to meet stake requirements without putting up all the IO themselves.

Participation requirements for GPU suppliers

The utilisation reality that determines actual earnings

This is the part the io.net landing page buries. In Q1 2025, io.net reported 327,000 registered GPUs but the daily average of verified, active GPUs was 6,720. That is 2% utilisation of the registered base. Only 5,350 GPUs are cluster-ready on any given day. Verified GPU count declined 11.1% quarter-on-quarter in Q1 2025 even as company-reported revenue grew.

Practically, this matters enormously for earnings. Block rewards are paid to every operator passing uptime checks. Compute job payments only flow when a renter actually books your specific GPU. If the network has surplus supply (which the utilisation figures indicate), block rewards can dominate take-home earnings while paid jobs remain sparse. Break-even on hardware purchased specifically for io.net supply is therefore driven more by the IO token price and the emission schedule than by rental demand. For consumer GPUs, that break-even is often longer than the hardware’s useful life. For data-centre tier hardware (A100, H100), the math is more favourable but still depends on winning enough paid jobs to justify the data-centre economics.

Current rates and the live dashboard

Because the emission rate and the per-tier GPU multipliers change month to month, the most reliable earnings figure for any specific GPU is the one shown in io.net’s live supplier dashboard at worker.io.net/worker/earnings-and-rewards (authentication required). Model earnings over at least a 30-day horizon before committing capital, and factor in restructure risk from the planned IDE rollout in Q2 2026, which will shift the block-reward component toward demand-driven payouts.

For a direct comparison of io.net’s revenue model against Render and Akash, see Render vs Akash vs io.net: the revenue question.

Honest assessment

What works

Revenue is reportedly real and growing. io.net reports $5.7 million in Q1 2025 with 82.6% quarter-on-quarter growth. Seven million compute hours delivered in January 2025, by the company’s own figures. The confidential compute offering with hardware-based attestation (Intel TDX plus NVIDIA) is ahead of most DePIN competitors. IO Intelligence provides a practical API for AI modelModelA trained neural network that takes inputs (text, images, audio) and produces outputs (more text, classifications, generated content). In DeAI the model is the thing that actually does the work.Like a very experienced apprentice who has spent years watching thousands of masters make furniture. They can't explain how they know when a joint is right, but they can make a chair that looks and functions like a Chippendale. The training is invisible. The output is what matters.Read more → access. Leadership under Gaurav Sharma (ex-Binance VP Engineering) brings operational credibility after a turbulent 2024. Eighty-five-plus partnerships announced by January 2025, including integrations with Filecoin, Render Network, ChainGPT, Injective and Nillion.

What doesn’t work

GPU numbers are the headline problem. io.net markets 327,000 “registered” GPUs. In Q1 2025, the daily average of verified, active GPUs was 6,720. 2% utilisation of the registered base. Verified GPU count declined 11.1% quarter-on-quarter in Q1 2025 even as revenue grew. How that gap between headline metrics and operational reality got this wide is a story from April 2024: approximately 1.8 million fake GPUs attempted to connect to the network to farm airdropAirdropDistributing tokens for free to eligible wallets, usually to reward early users, bootstrap a community, or decentralise token ownership away from a small group of insiders at launch.Like a supermarket handing out free samples to people who already shop there. The samples cost the supermarket nothing to print. The goal is to convert casual shoppers into loyal customers by giving them something tangible to talk about.Read more → rewards. Attackers exploited vulnerabilities that allowed them to mimic genuine GPU signals and spoof metadata. An RTX 4090 was reportedly split into infinite virtual GPUs. Martin Shkreli publicly highlighted the discrepancies. According to community reporting at the time, an io.net executive acknowledged the failure publicly. Only 5,350 GPUs are cluster-ready on any given day.

Both the “decentralised” and “open source” branding are misleading. io.net describes itself as “The Open Source AI Infrastructure Platform” on its website. The core platform code (backend, IO-SDK, GPU orchestration, matching engine, billing, IO Worker node software) isn’t open source. Only 9 public GitHub repos exist, containing setup scripts, binaries, documentation and demo apps. 131 GitHub stars total. Akash publishes everything.

There’s no governance. No DAO. No on-chain voting. No governance forum. No proposals mechanism. All decisions are made by the io.net team and foundation. Token holders have zero governance rights. The complete absence of governance is notable for a project of this size.

io.net achieved SOC 2 compliance, which covers organisational security controls and is relevant for enterprise customers. However, no smart contractSmart ContractA program stored on a blockchain that runs automatically when its conditions are met. Smart contracts are how blockchains do anything beyond just transferring tokens — DeFi, NFTs, DAOs, and DeAI infrastructure all run on smart contracts.Like a vending machine. You put in the right input and it produces the expected output, no human operator required. The rules are fixed in the machine itself, anyone can use it, and nobody can stop a transaction in the middle.Read more → audit or platform security audit from a recognised blockchain auditing firm (CertiK, Trail of Bits, Halborn, etc.) has been published, despite $30 million in Series A funding and a platform handling GPU compute workloads and a stakingStakingLocking up a cryptocurrency to help secure a blockchain network, usually in exchange for rewards. The locked tokens act as a security deposit that can be taken away if the staker misbehaves.Like putting down a large rental deposit for an apartment. You get the money back if you behave, you earn interest while it's locked, and the landlord takes it if you trash the place.Read more → system.

The risk

The founder risk has already materialised. Ahmad Shadid departed under a cloud and his subsequent project O.XYZ also faced misrepresentation allegations. Three CEOs in under two years creates leadership instability. Tory Green’s move to Foundation chair suggests organisational restructuring is ongoing.

The token is down 98% from ATHATHAll-Time High. The highest price a token has ever reached. ATH is usually quoted as a reference point for how far the current price has fallen (or risen) since the peak.Like the record lap time on a racetrack. It tells you what the car has been capable of at its absolute best, not what it will do today. Whether that record gets broken again depends on conditions that may or may not come back.Read more →. Insider unlocks began in July 2025 and continue through 2028. Burns provide some counterbalance, but roughly $10 million of annual buy pressure from burns against continuous insider unlocking isn’t enough to reverse the trend without substantial revenue acceleration. The IDE tokenomics overhaul announced in December 2025 addresses these structural problems directly: demand-driven emissions, 50% buyback-and-burn, and a target 50% reduction of the remaining emission pool. If implemented as described in Q2 2026, it would materially improve the supply dynamics. But “if implemented” is doing heavy lifting in that sentence.

Competitively, the position is squeezed. Akash is more decentralised, more transparent, fully open source, and cheaper on H100s. Centralised providers are more reliable and come with SLAs. io.net’s differentiation (the confidential compute, the IO Intelligence API, the DePIN supply aggregation) is real but isn’t yet a moat.

My position

I don’t hold IO. The closed-source core, absence of governance, founder scandal, inflated GPU metrics, and 98% price decline from ATH put this firmly in the “show me” category. The revenue growth is worth monitoring, but the decentralisation credentials are too weak for the sovereignty thesis. If you want decentralised GPU compute, Akash is the more honest option. If you want cheap GPUs without caring about decentralisation, Lambda Labs and Vast.ai don’t require a token.

Freedom Score: 38/100

io.net scores 38/100 (F grade). Full methodology at Freedom Score Methodology.

Infrastructure decentralisation (10/20): GPU supply comes from independent operators across 138-plus countries, which is architecturally distributed. Permissionless GPU onboarding is a genuine positive. But the core platform (orchestration, matching, billing, API, verification) is entirely centralised and operated by io.net. The IO Worker software is closed-source binaries. The 2024 Sybil attack demonstrated that verification systems were insufficient. Only 5,350 cluster-ready GPUs on any given day despite 327,000 registered. The platform can censor or deny service to any provider or consumer.

Governance decentralisation (2/20): No on-chain governance. No DAO. No governance forum. No proposals. All decisions made by the io.net team and foundation board. The io.net Foundation (chaired by co-founder Tory Green) controls strategic direction. Token holders have no governance rights whatsoever. This is fully centralised governance with a “foundation” label.

Token distribution fairness (5/15): Insider allocation is roughly 34% (seed 12.5%, team 11.3%, Series A 10.15%). Vesting schedules exist (two to three years with 13-month cliff). The Binance Launchpool (2.5%) was accessible but required Binance KYC. Community airdrop (10%) had reasonable distribution. Not a fair launchFair LaunchA token launch where everyone has the same access from day one. No private sale, no insider allocation, no VC discount. Tokens are distributed by mining, staking, or open public sale at a single price.Like a 100m sprint where everyone starts behind the same line at the same time. Some runners are faster, but nobody gets to start 10 metres ahead because they paid extra. The race is decided by the run, not by who bought the best position.Read more →. VC-backed at $1 billion valuation. Token declined 98% from ATH while insiders began unlocking in July 2025.

Censorship resistance (5/15): io.net can censor any user or provider. The platform controls account access, job allocation and the matching engine. KYC requirements for some participation. Terms of Service can restrict usage. The IO Worker binary is closed-source, so providers can’t verify what code they run. The underlying Solana token is censorship-resistant for transfers. Confidential compute VMs protect data during processing. But the platform layer is a centralised censorship vector.

Data sovereignty (8/15): Confidential compute VMs with Intel TDX and NVIDIA H100/H200/B200 provide genuine hardware-based data protection during processing. Users can verify attestation reports. Supports encrypted training data, model weights and architectures. Federated learning with dataset isolation is supported. SOC 2 compliance adds enterprise-grade organisational security controls. This is a real strength. However, standard (non-confidential) workloads lack these protections, and metadata (usage patterns, billing) is held centrally by io.net.

Open source and transparency (8/15): Only 9 public GitHub repos containing setup scripts, binaries, docs and demo apps. Core platform code is not open source. IO-SDK (Ray fork), GPU orchestration engine, matching engine, billing system and IO Worker node software are all closed-source. No reproducible builds. The “Open Source AI Infrastructure Platform” branding on the website is misleading given the closed-source core. However, SOC 2 compliance demonstrates audited organisational security processes, Messari publishes quarterly reports with verifiable network metrics, and token economics are well documented with verifiable on-chain data (contract: BZLbGTNCSFfoth2GYDtwr7e4imWzpR5jqcUuGEwr646K on Solana).

Path to improvement

Three changes would materially increase io.net’s score:

1. Open-source the core platform. The IO-SDK (Ray fork), GPU orchestration engine, and IO Worker node software are all closed-source. Publishing this code under a permissive licence would address the most fundamental credibility gap. If the code is good, open-sourcing it costs nothing. If it isn’t good enough to open-source, that tells you something. Akash publishes everything and it hasn’t hurt their competitive position.
2. Implement governance. Token holders have zero governance rights. No DAO, no proposals, no voting. Even a basic governance framework with proposal submission, quorum thresholds and on-chain voting would demonstrate that IO is more than a payment rail for a centralised company. The io.net Foundation controls all strategic decisions without community inputPromptThe text you give an AI model to tell it what to generate. A prompt can be a simple question, a long instruction, a chunk of context plus a task, or a conversation history the model uses to produce its response.Like a brief you give to a junior designer. A vague brief gets a vague result. A detailed brief with context, constraints, and examples gets something usable. The quality of the output depends heavily on the quality of the brief.Read more →.
3. Publish a real-time network dashboard with verified metrics. The gap between 327,000 “registered” GPUs and 6,720 daily active verified GPUs, combined with the 2024 Sybil attack history, has destroyed metric credibility. A real-time, independently verifiable dashboard showing actual active GPUs, compute hours delivered, revenue and burn amounts would rebuild trust. Self-reported quarterly figures filtered through Messari reports are not enough.

Returns Score: 54/100

IO scores 54/100 (D grade). Full methodology at Returns Score Methodology.

Token utility (12/20): IO functions as payment for GPU compute and as staking collateral for providers, which gives it more utility than most DePIN tokens. The Co-Staking Marketplace adds a passive participation layer. But the platform itself is entirely centralised. io.net controls orchestration, matching, and billing. The token is a payment rail for a company’s services, not a protocol-native necessity. If io.net decided to accept USDC-only tomorrow, the token’s utility would vanish overnight.

Value accrual (9/20): The IDE announcement materially improves the outlook here. The dual-vault mechanism (Reward Vault for supplier incentives, Fee Vault for platform revenue) creates a structured separation between emission subsidies and earned revenue. The 50% buyback-and-burn of Fee Vault revenue is a clear, mechanical link between platform usage and token value, exactly the flywheel that was missing. Previously, io.net took platform fees but the flow back to token holders was poorly defined. The IDE design fixes this on paper. Score reflects the credible design offset by it not yet being deployed. If implemented in Q2 2026 as planned, this dimension would score higher.

Supply dynamics (10/20): The structural problems remain: 34% insider allocation, vesting unlocks through 2028, and inflationary emissions. But the IDE’s target 50% reduction of the remaining 300 million emission pool is a significant commitment. Cutting long-term dilution from 300 million to 150 million tokens, combined with demand-driven emissions that reduce rewards when the network is underutilised, would fundamentally change the supply trajectory. The transition from hourly to monthly emission epochs is already in the documentation. Score bumped from 8 to 10 to reflect the credible plan, but not higher. The supply reduction is a target, not a deployed mechanism. The token is still down 98% from ATH, and insider unlocks continue regardless of IDE.

Revenue sustainability (15/25): This is io.net’s apparent bright spot. The company reports $20 million-plus annualised revenue with 82.6% quarter-on-quarter growth in Q1 2025, and 7.1 million compute hours in January 2025 alone. No independent audit of these figures has been published, so they are treated as self-reported. The figures suggest actual customers paying for actual GPU time, but the question of whether the growth rate holds and whether margins survive competition from Akash and centralised providers remains open.

LiquidityLiquidityHow easily a token can be bought or sold without moving the price. High liquidity means you can enter or exit large positions quickly at the quoted price. Low liquidity means even small trades can swing the market.Like the difference between selling a house and selling a share of Apple stock. The house might be worth more on paper, but finding a buyer at that price takes weeks. The Apple share converts to cash in one click.Read more → and access (8/15): Binance listing provides baseline accessibility, and the token trades on major exchanges including Coinbase, Gate, Bybit, and KuCoin. Volume has recovered somewhat from its post-launch trough but remains thin relative to market cap. The 98% drawdown from ATH means most holders are deeply underwater. Liquidity is adequate for retail but not for institutional-scale entry or exit.

Path to improvement

Three changes would materially increase io.net’s returns score:

1. Ship the IDE. The dual-vault, demand-driven emission, and 50% supply reduction plan is exactly what the tokenomics needed. But it’s an announcement, not a product. Delivering IDE on schedule in Q2 2026 and publishing transparent vault balances, burn amounts, and emission rates would prove the team can execute on tokenomics as well as they execute on revenue growth.
2. Demonstrate net deflationary months. Once IDE is live, publishing months where burns plus supply reduction exceed new emissions would be the single most powerful signal for token holders. The maths needs to work publicly, not just on a whiteboard.
3. Accelerate the insider unlock schedule or commit to no further raises. Two to three more years of continuous insider selling at current prices will suppress any organic price recovery. A compressed unlock with a commitment to no further dilutive funding rounds would signal confidence in the revenue trajectory.

Score change log