In Brief

When you think of machine learning at scale, your mind probably goes to hyperscale data centers, billion-dollar CapEx races, and a handful of centralized players dominating AI infrastructure. But what if the future of AI isn’t built in server rooms, but across a permissionless network of everyday devices?

DePINed Podcast is your front-row seat to the builders, thinkers, and founders reshaping our physical world through decentralized infrastructure. Each episode explores how crypto incentives and community-powered networks are transforming industries — from connectivity and storage to robotics and mobility.

In Episode 47 of the DePINed Podcast, Tom Trowbridge sits down with Ben Fielding and Harry Grieve, the co-founders of Gensyn, to talk about a bold vision for decentralized machine intelligence.

This isn’t just about cheaper compute — it’s about unlocking the next era of scalable AI infrastructure.

From Marketplace to Protocol: What Makes Gensyn Different

There are already several DePIN projects that let users rent GPUs. But Gensyn isn’t trying to outcompete on price. It’s focused on building the infrastructure to make machine learning a global, decentralized utility.

Gensyn focuses on three key technical pillars:

• Consistent Execution

Ensures identical outputs on different devices (GPU, CPU, TPU), verified down to the bit. Deterministic execution allows outputs to be hashed and compared, creating a cryptographic layer of trust.

• Communication

Enables multiple devices to perform computations together, allowing distributed training or inference across heterogeneous compute.

• Verification

Makes trust programmatic. You don’t need legal contracts or centralized intermediaries. The correctness of ML operations is verified cryptographically, with no human in the loop.

This trinity of execution, communication, and verification is what enables a new class of decentralized machine learning networks to emerge — ones that are actually usable.

The Shift from CapEx to Protocol-Scale

Centralized players are racing to build the biggest GPU clusters. But Gensyn sees this as a bottleneck, not a moat.

Ben and Harry argue that once you’ve built your centralized data center, that’s your ceiling. Protocols like Gensyn, on the other hand, allow you to stitch together any compute from anywhere — including unused devices sitting idle across the world.

It’s not about owning the infrastructure. It’s about orchestrating it.

And it’s not just about spinning up idle compute. By reducing the cost of establishing trust between machines, Gensyn removes the hidden costs of centralized coordination — legal, operational, human.

“The trust layer we rely on today is incredibly expensive. We just don’t notice it because it’s embedded everywhere — in contracts, in buildings, in institutions. Gensyn replaces that with software and code.”

Why ML Ops Are the New Economic Unit

Most people think the race in AI is about who has the biggest model.

Gensyn flips that narrative. The founders believe the real value will accrue not to the model layer, but to the ops layer — the execution of machine learning operations themselves.

The logic is clear: models can be copied, distilled, or replicated. But compute demand is persistent. Every operation has a cost, and that cost compounds as AI becomes more deeply embedded in every system.

“We think of ML ops as the atomic unit of the next economy — just like network packets in the 90s or ad impressions in the early 2010s.”

If compute becomes the base layer of economic activity in an AI world, then whoever can orchestrate, verify, and monetize those operations — at global scale — controls the new infrastructure.

Verification Is the Deep End of the Problem

Fluence focuses on decentralized CPU compute. Gensyn focuses on GPU-heavy machine learning workloads. But both have learned that verification is the bottleneck in trustless computing.

That’s why Gensyn was designed around the verification challenge from the start. By building a protocol that guarantees correctness — even when devices are untrusted — Gensyn removes the need for centralized validators, audits, or KYCed infrastructure.

“We could’ve just built this as a centralized company. But then we’d be limited by the cost of managing every new node. Verification makes scaling decentralized compute actually viable.”

Open Networks and the End of Centralized AI

Gensyn’s long-term vision is to support a world of billions of models, each with their own biases, specializations, and contexts.

Instead of relying on one monolithic foundation model, future applications will route across many models. And instead of humans trying to verify model quality — models will verify each other.

That’s not a flaw. That’s the point.

“Truth in machine systems, just like in human systems, comes from consensus. And consensus doesn’t require central authority — just agreement across nodes.”

The implication is huge: it’s not just that compute is decentralized — truth itself becomes decentralized.

How Gensyn Fits into the DePIN Landscape

If most DePIN projects struggle to attract demand, Gensyn has the opposite problem: demand for ML compute is growing exponentially. The challenge is unlocking supply and making it verifiable, trustless, and programmable.

Ben and Harry believe that Gensyn will become the default backend for ML workloads — a global protocol like TCP/IP, but for intelligence instead of data.

• Anyone with a GPU can join

• Anyone with a model can train

• No centralized gatekeepers required

It’s a powerful vision — and one with major implications for AI accessibility, infrastructure economics, and global competitiveness.

Incentives, Alignment, and Avoiding DePIN Traps

Gensyn’s approach to incentives is unusually grounded.

Instead of flooding the market with subsidized supply, they’ve built for real product-market fit. That means matching demand and supply based on actual compute needs — not token rewards.

“A lot of protocols get captured by their own supply side. We don’t want that. We want real usage.”

The Gensyn token — once live — will facilitate payments between compute buyers and compute sellers. But the focus isn’t on speculation. It’s on creating a functioning ML economy.

The Founders’ Background: From ML Pain to Protocol Design

Ben started with a PhD in deep learning, working on neural architecture search and swarm optimization a decade ago. He learned first-hand how compute-intensive real ML workflows are — even in academia.

Harry came from the applied side, modeling catastrophic insurance events with ML systems and building data-intensive pipelines. He saw how traditional companies couldn’t keep up with the infrastructure demands of modern ML.

They met at Entrepreneur First in London, bonded over a shared interest in decentralization, and built Gensyn to solve the real pain points they had both experienced.

“We weren’t crypto-native at first. But crypto gave us the primitives to solve trust in a decentralized way.”

Stay Connected with the DePIN Ecosystem

DePINed Podcast is a front-row seat to the builders, thinkers, and founders reshaping our physical world through decentralized infrastructure. Hosted by Tom Trowbridge, every episode explores the real-world use cases, economic models, and visionaries behind the DePIN movement.

Want to go deeper?

Join us at DePIN Day — a global event series designed to spotlight the latest trends, challenges, and breakthroughs in the decentralized infrastructure space. From bandwidth to compute, from storage to mobility, DePIN Day brings together founders, researchers, developers, and ecosystem leaders to shape the future — together.

🛰️ Let’s build the future of infrastructure:
• https://x.com/depinday
• https://www.youtube.com/@fluencelabs