Over my more than 20 years of building SaaS companies, one theme kept surfacing: bottlenecks follow when demand for digital services outpaces infrastructure. I saw it in the early internet era, when websites slowed to a crawl because storage and bandwidth couldn’t keep up. Now I’m seeing it with compute.
At OTOY and the Render Network Foundation, where I’ve spent the past several years, the challenge is making sure there’s enough GPU power to support the talent and ideas we have, especially because the race for AI is accelerating faster than anyone predicted.
Spending on data centers has skyrocketed nearly sevenfold since 2022, and the demand for compute is only beginning. By 2030, consultancy firm McKinsey projects that $5.2 trillion will be invested globally into data centers equipped to handle AI processing. However, there’s a catch: the power grid itself cannot scale at the same pace. The energy strain of AI workloads is already visible, and without new models for distributing compute, even trillions in new infrastructure will hit physical limits.
One promising path is decentralized GPU networks. They began as enthusiasts pooling idle graphics cards and have now matured into serious infrastructure. Today, decentralized networks are enabling digital artists to render projects in hours instead of weeks, researchers outside Big Tech to train models they once couldn’t afford, and startups to experiment without mortgaging away their futures. Analysts project the decentralized physical infrastructure (DePIN) market to reach $3.5 trillion by 2030, showing just how significant this could be.
Why This Is A Turning Point
I’ve seen firsthand what’s possible when you distribute compute differently. I’ve talked with artists whose rendering timelines dropped from weeks to hours once they tapped decentralized GPUs.
Essentially, each video frame can be rendered on a different node across the globe, enabling Hollywood-scale productions to run in parallel at a fraction of the time and cost. That same model has powered some of the largest Hollywood stages, immersive displays like the Las Vegas Sphere, award-winning music content, gaming environments, and now AI workloads, all delivered in a distributed, affordable way.
Decentralized GPU networks can expand what’s possible without completely replacing hyperscale data centers. Lower costs and open participation make computing available to more voices, regions, and industries. Value is then distributed across a much wider stack, as opposed to only flowing to a few well-capitalized players.
Consider a university lab in Nairobi training a medical AI model, a filmmaker in São Paulo rendering her vision, or an independent researcher running an experiment that sparks the next wave of discovery. These are not hypotheticals; they’re examples of what becomes possible when computing is accessible, affordable, and distributed. Around the world, millions of idle consumer GPUs are waiting to be tapped as AI architectures evolve to unlock them.
The Inversion of Power
For decades, information was funneled through publishers and broadcasters until the internet made knowledge searchable, remixable, and borderless. We’re now seeing a similar shift with raw processing power. In decentralized GPU networks, contributors can earn from idle hardware, while creators and researchers tap into compute through open marketplaces.
Pricing is shaped by demand and supply, not just enterprise budgets.
What’s at Stake
The challenge is both financial and physical. Concentrating compute in a few mega data centers puts enormous strain on local grids. Decentralized networks, by contrast, can tap into off-peak energy across regions, spreading the load in a way that’s more sustainable and resilient.
Of course, hyperscale data centers will remain critical for training the largest models and simulating complex systems. But the majority of innovation doesn’t need billion-dollar clusters; it can be achieved with affordable, distributed computing. By embracing decentralized networks alongside centralized ones, we create resilience, reduce costs, and accelerate experimentation.
If we don’t solve the bottleneck of compute constraints, we risk slowing AI’s potential and narrowing its impact. If we do, the future of AI can be shaped not just by budget committees in Silicon Valley but by creators, researchers, and entrepreneurs everywhere. That’s the opportunity decentralized GPU networks unlock: a more open, more inclusive, and more innovative internet of compute.