In this session, Greg Osuri, founder of Akash and CEO of Overclock Labs, shares how Akash is delivering on the vision of a decentralized cloud. With a radically open-source protocol, affordable on-demand GPU access, and a growing community of contributors, Akash is proving that DePIN and AI are working alternative to traditional cloud infrastructure.
The Origins of Akash: From Centralized Kubernetes to Open-Source Supercloud
Akash didn’t begin as a GPU network. Back in 2015, it started as a way to run Kubernetes across multiple cloud providers. The goal? A truly multi-cloud architecture, deployable with a single click across 20+ regions. Over time, the limitations of centralized control planes became apparent, especially at scale. So Akash pivoted toward decentralization — splitting the control plane from the resource layer to eliminate single points of failure.
This architectural shift laid the foundation for what Akash would become: the world’s first implementation of a fully open-source “supercloud,” powered by the Cosmos SDK and featuring inter-blockchain communication (IBC).
Why DePIN Infrastructure Is Critical for AI
The Akash whitepaper, published in 2018, predicted the strain AI would place on cloud infrastructure. Today, we see the impact firsthand with GPU shortages, skyrocketing costs, and massive demand from startups and researchers.
Akash offers on-demand access to H100s — NVIDIA’s top-tier GPUs — for as little as $1.50/hour, something unheard of on traditional platforms. A Columbia University student, for example, used Akash after finding Amazon Web Services too expensive or unavailable.
Real-World Traction and Ecosystem Growth
Akash has already delivered:
- 15M+ gigabyte hours of compute
- 186K+ workloads deployed
- A growing user base across ML, Web3, and infra
Popular AI projects like Nous Research, Brev, and Venice AI train and deploy on Akash. On the Web3 side, protocols like Osmosis and BitTen also use Akash for decentralized infrastructure.
Akash is currently the only place where developers can get on-demand access to H100s at this scale.
A Radically Open Ecosystem: Built by the Community
What sets Akash apart isn’t just price or performance. It’s the culture. With 500+ contributors and over 30 on-chain funding proposals, the Akash ecosystem is truly decentralized. Anyone can propose improvements, contribute to development, or launch new features.
Multiple companies now contribute to the Akash stack, with Overclock Labs as the core team. Governance is handled by working groups and special interest groups, ensuring no single entity dominates the roadmap.
Roadmap and What’s Coming Next
Greg shared the 2024 roadmap, much of which was outlined as far back as 2019 and is now being delivered. Key upgrades include:
- Cloud parity with major providers
- User-friendly deployment tools (e.g., one-click deploys via Brev)
- Incentives for GPU providers
- Non-custodial access with full transparency
The team is now building out decentralized storage, a data marketplace, and managed backend services to round out Akash as a full-stack platform for AI builders.
Akash Proves That Decentralized AI Infrastructure Works
While most DePIN + AI projects are still making promises, Akash is delivering results. With live users, production-scale deployments, and unmatched GPU economics, Akash is defining what it means to build real-world infrastructure onchain.
AI needs compute. Akash delivers it — cheaply, reliably and without centralized gatekeepers.
