Decentralized AI’s New Secret Weapon: Why Greg Osuri is Betting Big on Virtual Machines
In an interview with Ecoinimist, Greg Osuri, CEO of Akash Network, offered a candid breakdown of how decentralized cloud infrastructure is evolving—and why the next phase of innovation has less to do with flashy new virtualization tech and more to do with who gets access to compute in the first place.
Osuri, who has been building in crypto since 2014 and has contributed code to foundational virtualization tooling such as Docker and Kubernetes, framed Akash’s mission in pragmatic terms.
“It’s not compute that’s the problem,” he said during the interview. “Everything runs in the cloud already. The problem is access to the computer.”
Meeting AI Developers Where They Are
Akash launched as a container-native platform designed for advanced developers comfortable with Docker-style workflows.
But Osuri explained that the rapid rise of machine learning workloads forced the team to rethink its assumptions. Unlike traditional software engineers, many AI practitioners are researchers and scientists who prioritize flexibility over rigid production pipelines.
“In training environments, things break all the time,” Osuri said. “You want to log into the machine, tweak things on the fly, and get back up quickly.”
That reality pushed Akash to expand beyond containers and support virtual machines, even if VMs are often viewed as less modern than containerized deployments. The move wasn’t about technical nostalgia—it was about matching how users actually work.
“We had to meet the user where they are,” Osuri noted.
Blockchain as a Coordination Layer, Not a Computer
A recurring theme in the interview was Osuri’s skepticism toward the idea of fully “on-chain” computing. While blockchain plays a critical role in decentralized infrastructure, he argued it should not be confused with an execution environment for heavy workloads.
“Blockchains are security systems,” Osuri said. “They’re great for access control, permissions, and settlement—but by definition, you give up speed for security.”
Running general-purpose AI workloads directly on blockchain, he added, makes little sense due to cost and performance constraints. Instead, Akash uses blockchain as a coordination framework—handling identity, access, and payments—while compute itself runs off-chain on distributed hardware supplied by independent providers.
Why Decentralized Compute Is Gaining Momentum
Osuri’s comments land at a moment when decentralized AI infrastructure is drawing renewed attention. Akash operates a marketplace that connects underutilized GPUs and servers with developers who need compute, often at lower costs than centralized cloud providers. The network’s native token, AKT, is used for payments, staking, and governance.
Recent developments underscore that momentum. Akash has publicly discussed deprecating its Cosmos-based app chain and is evaluating alternative base networks, including Solana—a strategic shift that could reshape how the protocol integrates with the broader crypto ecosystem. At the same time, community-led initiatives such as AI agent build nights and student ambassador programs are expanding developer engagement.
A Different Way to Think About the Cloud
Ultimately, Osuri positioned Akash not as a radical reinvention of cloud technology, but as a rebalancing of power.
“Blockchain is a coordination technology for unknown parties,” he said. “If you already trust each other, you don’t need it. But if you don’t, it becomes incredibly powerful.”
As AI demand accelerates and centralized cloud capacity becomes increasingly constrained, that coordination layer—rather than new virtualization breakthroughs—may prove to be decentralized compute’s most enduring innovation.
