Andrew Sobko is a serial entrepreneur with a background in creating transformative marketplaces.
He founded one of the fastest-growing companies in America, an achievement recognized by the Financial Times and honored with Goldman Sachs' Builders & Innovators award. Andrew has raised over $200 million from leading global investors, including Sequoia Capital, Brookfield, and others.
We recently spoke with Andrew Sobko, CEO of Argentum AI, about the inefficiencies in the compute market and how his new venture plans to build a decentralized “Airbnb for GPUs.”
Andrew Sobko is a serial entrepreneur with a background in building multiple logistics and resource marketplaces. After identifying key inefficiencies in the compute market, namely massive AI demand versus a global surplus of underutilized GPUs, he applied his marketplace principles to the sector to found Argentum AI.
Argentum AI is a decentralized, AI-optimized compute marketplace that matches idle global GPU and data center capacity with AI developers and enterprises. Through secure, real-time auctions and on-chain verification, it provides infrastructure liquidity, enabling providers to monetize unused compute and giving users access to transparent, affordable, and borderless power for their most demanding workloads.
Read more about Argentum’s strategy to unlock global compute liquidity in the interview below.
Why focus on compute infrastructure now?
Having built multiple logistics and resource marketplaces, I realized similar inefficiencies in compute: high demand for AI training and inference on one side, and a global surplus of underutilized GPUs on the other. With AI adoption soaring and centralized hyperscalers hitting power and scalability limits, now is the time to rethink compute access.
Argentum applies marketplace principles to unlock liquidity in a sector where delays and high costs hinder innovation. This strategy goes beyond a purely technical opportunity, creating economic and strategic benefits that will influence how AI is developed and scaled globally.
What is Argentum AI in one sentence?
Argentum AI is a decentralized marketplace that uses secure auctions to match idle GPU providers with AI developers, creating liquidity in the infrastructure market and providing affordable, on-demand compute power.
What are your key milestones after the Kraken-led funding?
The pre-seed round provides us with the resources to expand rapidly. We’re focused on completing our AI-powered auction engine, launching a fully auditable compute network, and onboarding top-tier GPU providers, ranging from data centers to individual operators.
Key deliverables include expanding cross-border compute capacity, integrating zero-knowledge trust and telemetry layers, and launching pilot AI training jobs with early enterprise partners. We’re also investing in liquidity programs for second-life GPUs to create a secure, fair, and scalable ecosystem.
You’ve described the concept as the “Airbnb of compute.” Can you explain that?
We use the Airbnb analogy because, like homes, GPUs often sit idle and can be monetized in an open market. But to truly unlock compute liquidity, you need more than just a listing. You need real-time discovery, dynamic pricing, performance guarantees, and trust in execution.
Argentum addresses this with a decentralized auction system, cryptographic proofs, and verifiable benchmarks. We make sure jobs are matched to the right hardware with transparent results, enabling compute to move freely across borders, just like how Airbnb transformed spare rooms into a global lodging supply.
How do GPU entrepreneurs get started on Argentum?
Anyone with a compatible GPU can join Argentum by installing our node software and participating in the bidding system. Our onboarding process is secure but simple: providers undergo lightweight verification and build credibility by successfully completing tasks. As they finish jobs, they are evaluated for speed, accuracy, and dependability.
High performers move up in visibility and earnings. Whether you’re a solo enthusiast or a mid-sized facility, Argentum allows you to turn hardware into income without complex contracts, middlemen, or hidden fees, offering transparent rewards worldwide.
What’s the job flow for users?
Users begin by posting compute tasks with details such as model type, hardware requirements, duration, and budget. The system then initiates a real-time auction, and qualified providers submit bids. Once a provider is chosen, funds are secured in escrow through a smart contract.
Execution occurs through encrypted telemetry streaming to the network, enabling users to monitor progress. Once finished, the results are validated against the SLA, and payment is released. This entire process (matching, execution, settlement) is automated, traceable, and designed to provide cloud-grade security with greater flexibility and control.
How do you enforce SLAs and performance?
We combine real-time telemetry, cryptographic proofs, and economic incentives to maintain consistent quality. Each job has specified performance metrics (latency, throughput, accuracy), and our smart contracts hold providers accountable. Poor execution results in penalties or stake slashing, while excellent performance fosters trust and higher rankings.
Our “Living Benchmark” system continuously assesses all nodes to ensure effective and intelligent job matching. Redundant tasks and on-chain reporting ensure everything is auditable. This creates a self-regulating network where reliability is rewarded, and bad actors are economically discouraged from the start.
Who are your main competitors, and what is Argentum’s edge?
We respect other players in decentralized compute like Akash, Render, and Golem, but most of them focus on general or CPU-based workloads. Argentum is specifically designed for high-performance GPU tasks, including AI training, inference, and rendering. What sets us apart is our AI-native approach: real-time bidding, verifiable execution, adaptive benchmarking, and a strong focus on zero-knowledge privacy for cross-border operations. While others provide static infrastructure, we’ve created a dynamic, liquid, and fully transparent marketplace that adapts in real-time based on job complexity, provider reliability, and energy availability.
What is the five-year vision for Argentum?
In five years, we envision Argentum as the decentralized backbone of global AI compute, a self-optimizing, borderless grid where anyone can access high-end GPUs like B300s and H100s with just a few clicks. Our goal is to empower innovation everywhere, from Silicon Valley startups to African AI labs, by making compute accessible, secure, and affordable.
We want to see thousands of GPU entrepreneurs earning income, millions of jobs running daily, and Argentum serving as a global settlement layer. Simply put, we aim to democratize infrastructure the same way cloud did a decade ago, but this time, it’s decentralized.