Company: io.net
Founder: Gaurav Sharma
Website: io.net
About io.net
While headlines in 2024 and 2025 chronicled AI industry excesses -OpenAI missing revenue targets despite massive funding, Meta surveilling employees for AI training data, and hyperscalers locking developers into exploitative contracts – io.net launched in June 2024 with a different vision. By pooling more than 30,000 GPUs from data centers, enterprises, universities and individual suppliers across multiple continents, io.net built the world’s largest decentralised physical infrastructure network (DePIN) for AI compute. The mission: prove that AI infrastructure can be democratised, cost-effective and ethically sound.
io.net entered the market precisely when the compute crunch reached crisis levels. Data-center GPU lead times climbed to a full year. Nvidia Blackwell chip rental prices jumped 48 percent in two months, from 2.75 dollars per hour to 4.08 dollars per hour. Hyperscalers began requiring minimum 50 million dollar multi-year contracts and commitments to rent at least 1,000 Blackwell GPU units for one year. The AI bubble was pricing out everyone except Fortune 500 companies and well-funded startups, creating a two-tier system where access to compute determined who could participate in the AI revolution.
The challenges io.net overcame were substantial. On the technical side, the team built an industry-leading orchestration layer capable of managing heterogeneous infrastructure at scale. This means mixing enterprise-grade and consumer GPUs in a single network while delivering performance that matches centralised providers in latency, concurrency, and throughput. The orchestration layer abstracts the complexity of provisioning, networking, and workload distribution, allowing AI teams to spin up GPU clusters on-demand without extensive coordination or custom coding.
Economically, io.net had to prove that a decentralised model could compete on cost while remaining sustainable. The numbers speak for themselves: H100 GPUs cost 2.19 dollars per hour on io.net versus 12.29 dollars per hour on AWS, delivering 82 percent cost savings. Most startups save more than 10,000 dollars monthly. The company has paid more than 24 million dollars to compute suppliers since launch, enabling thousands of individuals and organisations to monetise underutilised hardware.
io.net also addresses one of the AI industry’s most absurd paradoxes: 95 percent of enterprise GPU capacity sits unused, according to the 2026 State of Kubernetes Optimisation Report by Cast AI, while AI startups struggle to access compute. Billions of dollars worth of hardware gathers dust in corporate data centers while developers cannot afford hyperscaler pricing. By activating this stranded capacity, io.net turns idle hardware into productive infrastructure and challenges the narrative that compute scarcity justifies predatory pricing.
The impact extends beyond cost savings. Multi-million dollar enterprise contracts demonstrate that decentralised infrastructure can serve demanding production workloads, not just experimentation. The network serves developers in more than 50 countries, bringing AI compute access to regions in Africa, Latin America and Southeast Asia previously priced out by hyperscaler economics. While big tech consolidates power through exclusive compute deals and opaque pricing, io.net is proving that decentralised infrastructure can deliver performance parity, eliminate single-point-of-failure risk, and democratise access.
The broader DePIN sector has grown from 5.2 billion dollars to more than 19 billion dollars in market capitalisation within one year, with projections reaching 3.5 trillion dollars by 2028. io.net is positioned as a category leader in this emerging infrastructure model. The June 11, 2026 anniversary milestone marks a major step forward, an 8 million dollar enterprise deal and new integrations expanding the network’s reach. In an industry plagued by ethical scandals and extractive business models, io.net represents what AI infrastructure should be: accessible, transparent, and built to serve developers rather than lock them in.
Want to find out what TechRound can do for your business?
