Decentralized infrastructure networks (DePINs) almost all start the same way: they pay their early contributors with newly minted tokens on a fixed schedule. It works while the token is climbing, but the design has a flaw baked in. The tokens keep flowing whether or not anyone is actually using the network, so supply runs ahead of real demand. When the price falls, contributors' income falls with it, they switch off their hardware, capacity shrinks, and the network spirals down.
The fix is simple in principle: stop paying on a fixed schedule and tie token emission to real usage and revenue instead. io.net's Incentive Dynamic Engine (IDE) is one of the first serious attempts to make that switch - pay suppliers in stable dollar terms, and only release or burn tokens based on what the network actually earns.
Training and running AI models takes enormous amounts of GPU power, and the best chips are expensive and hard to get hold of. The big cloud providers (AWS, Google, Microsoft) charge a premium for them and often have waitlists. At the same time, a lot of GPU capacity sits idle around the world: in independent data centres, with crypto miners, in smaller server farms. The idea behind a decentralised GPU network is to pool all that unused hardware and rent it out - cheaper, and without the queue.
io.net is one of the larger attempts to do this, built on the Solana blockchain. It pulls together GPUs from data centres, independent operators and miners across 138 countries and bundles them on demand into clusters that run AI and machine-learning workloads.
There are two sides: renters get compute through IO Cloud; suppliers plug their GPUs in through IO Worker. Beyond these two, io.net has expanded into adjacent products - most notably IO Intelligence (pre-trained models and AI agents for developers who need more than raw compute) and IO Staking (where suppliers post IO as collateral to participate in the network).
The main selling point is cost - io.net says it undercuts on-demand pricing from the big clouds such as AWS by roughly 70% by tapping that idle supply.
Like any marketplace, io.net makes money by taking a cut. IO charges a nominal margin from the renters on top of supplier price, and a 3% platform fee for users paying in Fiat. There is no platform fee while paying through IO, as a deliberate nudge towards the token.
Total Network Earnings - the gross value of compute transacted on the network - serve as a proxy for demand-side activity: when clients are purchasing more compute, TNE rises, and when the network is idle or underutilised, it falls. Daily network earnings have averaged approximately $35-36K over the period and have been trending upward since March 2026. IO price has broadly moved in the same direction over this window.
In the DePIN/Compute space its closest peers are Aethir and Akash, with Render in an adjacent niche. io.net is the cheapest from an FDV point of view. IO is what suppliers are paid in, what secures the network - operators stake IO to run a node - and a discounted way to pay.
IO is capped at 800 million: 500 million (62.5%) was allocated (with specific unlock schedules) at launch; the remaining 300 million (37.5%) is released as rewards to GPU suppliers and the stakers behind them - paid hourly, on a schedule stretching about 20 years.
Those rewards taper on a fixed schedule: the emission rate starts at roughly 12.6% in year one and declines about 12% a year - around 1% a month - reaching ~1% by the mid-2040s once the 800M cap is hit. In net terms that is about 8% supply inflation in the first year. The path is pre-set, independent of network activity.
The weakness is baked into that emissions schedule. Emissions come out at a fixed pace that ignores demand - the same flow of new tokens whether the network is busy or idle. In slow stretches, supply swells while real usage doesn't, and the price gets pushed down by design.
The real damage is to operators. They earn IO, but their bills - power, hardware, rent - are in dollars. When IO falls, their pay falls. Marginal operators unplug and leave. Capacity drops, the network weakens, demand and price slide further, and the operators still running get squeezed harder. That is the loop fixed-emission networks keep falling into - io.net’s own litepaper names it the “negative spiral.”
A burn could offset the new supply. io.net’s burns are fed by a thin slice of fees that barely registers against the emissions at today's activity.
io.net isn't a special case - most DePIN networks pay contributors on fixed schedules that ignore usage, which is why the sector's token value has leaned on emissions far more than real demand. This is evident by the relative return of DePIN tokens v/s BTC over the past one year.
io.net's answer is the Incentive Dynamic Engine, or IDE - a shift from paying suppliers on a fixed schedule to paying them based on what the network actually earns.
The central idea is to pay suppliers in steady dollar terms. Each epoch, the system tallies the total dollar income owed to all active GPU providers - roughly the number of machines times a target return each, plus running costs.
It then checks IO's live price and releases just enough tokens to cover that figure: when the price is low it takes more tokens, when high, fewer. The point is that an operator's pay - though settled in IO tokens - is sized to hit a fixed dollar target regardless of where the token trades, which is what keeps them from walking away in a downturn.
Two reserves sit behind it: A reward vault (Block Rewards), refilled by emissions, is drawn from first; a fee vault (Client Payouts), filled by the dollars clients pay, backs it up when the first runs thin. Together they're designed to let the network dip into savings during a rough patch instead of dumping new tokens on the market.
The engine is built around one test: is the network earning more than it owes its suppliers? (io.net calls this the sustainability ratio) Block rewards are emitted regardless, and once supplier payouts are covered, at least half of the surplus - block rewards plus any client earnings above the payout - is used to buy back and permanently destroy IO, so the busier the network gets, the more is burned. io.net's stated aim is to retire at least 115M of the ~231M of rewards still to be emitted this way over time. When earnings fall short, the engine instead leans on the reserves and, if needed, releases extra tokens to keep supplier pay whole.
Done right, this would flip the old dynamic: instead of inflating regardless of usage, the token tightens as the network gets busier.
The Incentive Dynamic Engine is designed to achieve a single objective: maintain stable supplier compensation in dollar terms while allowing IO emissions to adjust dynamically around that target. During each epoch, the engine works out the dollar payout owed to all active suppliers, converts it to its IO-equivalent at the live price, and compares that with the Block Rewards available. If Block Rewards fall short, it draws on the Reserve Vault; if that is still short it taps client payouts. At a sustainability ratio of 1, at least 50% of block reward emissions are burned; above 1, both the block rewards and any client surplus are available to burn, with at least 50% of the combined pool permanently destroyed.
Looking to stress test the model, there are three core scenarios:
Between healthy surplus and ordinary deficit, the outcome turns on the sustainability ratio - network earnings divided by what is owed to suppliers. The burn mechanism is already active at launch: at a sustainability ratio of 1, block rewards continue to be emitted, but the engine burns at least 50% of them - so even at a ratio of 1., the IDE is already partially offsetting its own emissions to suppliers. Above 1, a profit margin opens up on top of the block reward emission, and both are available to burn.
As that margin grows, so does the burn: once the profit margin equals the value of block rewards emitted, the total surplus available to burn is 2× the emission value - and burning even 50% of that fully offsets emissions. Once the profit margin exceeds block rewards, the engine becomes net-deflationary.
The chart below assumes a payout of ~$17.5M - the level implied by today's conditions, roughly current network earnings (~$13M) plus the value of block rewards emitted over the same period (~$4.5M) - and sweeps IO price along the x-axis. The sustainability ratio sets the direction. At a ratio below 1 the engine is a net issuer; at 1 it still issues, because block rewards keep flowing and only about half are offset by burns; and it turns net-deflationary only once the ratio rises enough that the surplus above the payout matches the block-reward emission.
Price sets the magnitude: the dollar surplus or deficit is fixed by demand, but the number of IO tokens needed to settle it varies inversely with price, so every line steepens toward the cheap-IO left and flattens to the right.
io.net has identified a real, sector-wide problem - fixed emissions that leave most of DePIN structurally inflationary - and built one of the more complete answers to it. The IDE is coherent economic design: it stabilises supplier income in dollars, makes net token supply respond to network earnings - issuing into shortfalls, burning into surpluses - and so, in principle, converts IO from an inflationary reward into a token whose scarcity tracks usage. That's a genuine advance on the fixed-schedule model it sits on top of.
The engine is already burning tokens -and io.net expects to remove at least 12 million $IO from circulation over the first year. At a sustainability ratio of 1, at least 50% of block reward emissions are burned, partially offsetting new supply. Net deflation requires the profit margin above a ratio of 1 to grow until burning 50% of the combined surplus - surplus client earnings plus block rewards - exceeds the full emission rate.
The fastest route there is on the demand side: io.net has already closed an $8M enterprise contract contributing roughly $650K in monthly network earnings, with further deals reported in advanced stages. Contracts of that scale lift earnings, widen the margin above the payout, and bring the deflationary flip forward.
The IDE is now live. The signal to watch is simple - whether the sustainability ratio holds at or above 1 and the burn counter keeps moving. io.net has built the engine; what turns it on is demand.
