How Does io.net Work?

DeFi|Risk C+|6 mechanisms|5 interactions

io.net is a Solana-based decentralized GPU computing platform that aggregates underutilized GPUs worldwide for AI and machine learning workloads. Backed by $30M from Hack VC and Multicoin Capital at a $1B token valuation, the network faced a major credibility crisis in April 2024 when 1.8 million fake GPUs were found attempting to connect, revealing that reported node counts were massively inflated. Its C+ grade reflects this track record incident, the ongoing challenge of verifying decentralized GPU supply, and the early-stage nature of compute demand, partially offset by the team's transparent postmortem response and introduction of new verification mechanisms including the Incentive Dynamic Engine.

TVL

Sector

DeFi

Risk Grade

C+

Value Grade

D

Core Mechanisms

5.3 Compute Layer

Decentralized GPU compute marketplace aggregating underutilized GPUs for AI/ML workloads on Solana

Standard decentralized compute marketplace pattern. Similar to Render Network, Akash, and Golem. Users pay IO to rent GPU compute, suppliers earn IO for providing resources.

7.2 Node Operator Incentives

Novel

IO token emissions to GPU suppliers with Incentive Dynamic Engine (IDE) replacing fixed inflation

Novel: IDE automatically adjusts emissions based on real compute demand, performs buybacks, and burns revenue-derived tokens to reduce inflationary pressure. This demand-driven tokenomics model is a departure from standard fixed-emission DePIN designs.

1.3 Fee / Burn Mechanisms

IDE burn mechanism: material portion of compute revenue used for IO buybacks and burns

Standard buyback-and-burn mechanism. Revenue from compute jobs funds IO purchases and burns to reduce circulating supply.

7.3 Identity & Sybil Resistance

Proof of Work GPU verification to prevent fake node registration

Standard PoW-based verification. Introduced after the April 2024 Sybil attack to authenticate genuine GPU hardware. Nodes must complete computational challenges to prove real GPU capacity.

3.3 Staking / Lockup

IO staking for GPU suppliers and network participants

Standard staking mechanism. GPU suppliers and token holders stake IO to participate in the network and earn rewards.

1.2.1 Linear vesting with cliff

1-year cliff with 2-year vesting for investors, 3-year vesting for team

Standard vesting structure. Seed (12.5%) and Series A (10.15%) investors have 1-year cliff then 2-year vesting. Core contributors (11.34%) have 1-year cliff then 3-year vesting.

How the Pieces Interact

GPU node registration and verificationIO token emission rewardsHigh

Node operators are incentivized by IO emissions to register GPUs. The April 2024 Sybil attack demonstrated that fake nodes can spoof GPU metadata to farm token rewards. Despite new PoW verification, the economic incentive to game node registration persists as long as emissions reward supply-side growth.

IDE demand-driven emissionsActual compute utilization ratesHigh

IDE adjusts emissions based on compute demand, but if demand metrics can be gamed (self-referencing jobs, wash compute), the dynamic adjustment mechanism could be exploited to maintain inflated emissions without genuine organic demand.

IO token buyback-and-burnCompute revenue generationMedium

The burn mechanism depends on real compute revenue. If organic demand remains low relative to emissions, buyback-and-burn cannot offset inflationary pressure from supplier rewards, leading to persistent net inflation.

Decentralized GPU supplyCentralized cloud compute competitionMedium

GPU owners can redirect hardware to competing platforms (AWS, Lambda Cloud, Render Network) if io.net rewards become uncompetitive. Simultaneously, centralized providers offer reliability guarantees and SLAs that decentralized networks struggle to match.

Investor vesting unlocks (July 2025+)IO token market liquidityMedium

Seed (12.5%) and Series A (10.15%) investors began unlocking in July 2025 with 2-year vesting. Combined with core contributor unlocks (11.34%), approximately 34% of total supply enters vesting release over 2025-2028, creating sustained sell pressure during critical growth phase.

What Could Go Wrong

  1. Major Sybil attack in April 2024: approximately 1.8 million fake GPUs attempted to connect to the network, exploiting vulnerabilities in node verification. The true network size was found to be approximately 120,000 verified nodes, far below the initially reported 500,000+, severely undermining credibility of reported metrics.
  2. GPU supply verification remains a fundamental challenge for decentralized compute networks. Despite security patches and a new Proof of Work verification mechanism introduced after the Sybil attack, the difficulty of remotely verifying physical GPU capacity creates persistent gaming risk.
  3. Early-stage network with limited production AI workload adoption. While the network spans 7,000+ GPUs across 50+ countries, real compute utilization and revenue generation metrics are opaque, creating uncertainty about organic demand.
  4. Significant token price decline: IO has fallen substantially from its launch price, with market cap dropping from approximately $860M FDV to $155M circulating market cap, reflecting market skepticism about the project's metrics and adoption trajectory.

Metric Credibility Crisis

Moderate

Trigger: Independent analysis reveals that reported GPU count or compute utilization metrics are inflated by more than 50%, similar to the April 2024 discovery, triggering a credibility crisis.

  1. 1.Independent researchers or journalists discover that io.net's reported metrics remain significantly inflated despite PoW verification Market confidence collapses as the April 2024 Sybil attack history amplifies skepticism about the network's real capacity
  2. 2.Enterprise and institutional compute clients avoid or exit the platform due to unreliable capacity guarantees Compute revenue drops, undermining the IDE buyback-and-burn mechanism and tipping the network toward net inflation
  3. 3.IO token price decline accelerates as credibility loss compounds with vesting unlock sell pressure GPU suppliers exit the network as IO rewards become worthless, reducing real capacity and further degrading the platform
  4. 4.Competing DePIN networks (Render, Akash) absorb displaced compute supply and demand io.net loses network effects and competitive positioning, potentially becoming unviable

Risk Profile at a Glance

Mechanism Novelty3/15
Interaction Severity6/20
Oracle Surface2/10
Documentation Gaps4/10
Track Record6/15
Scale Exposure7/10
Regulatory Risk4/10
Vitality Risk8/10
C+

Overall: C+ (40/100)

Lower score = safer

More on io.net

Related DeFi Explainers