Planck Network aims to make decentralised GPU computing power available to AI developers

Anyone looking to train AI models will sooner or later run into the same problem: computing power is expensive, scarce and concentrated in the hands of a small number of large cloud providers. Planck Network, a deeptech company from Harderwijk, wants to organise things differently by pooling GPU capacity in a decentralised network on which developers and businesses can run workloads directly.

The company was founded in 2023 by Diam Hamstra and Rohan Talwadia and combines AI infrastructure with crypto technology. In early 2025 it raised €200 million in a Series B round, signalling investor confidence in the approach, even though the network is still at an early stage of rollout.

Planck positions itself as an infrastructure layer for other networks and applications that require computing power for AI workloads. That makes the company less visible to end users, but potentially relevant to a broad segment of the AI market.

The problem they are addressing

Demand for GPU computing power for AI is growing faster than supply from traditional cloud providers. Companies such as Google Cloud, AWS and Microsoft Azure dominate the market, with the corresponding prices and dependencies that entails. For startups and smaller AI labs, that means high costs and little room to negotiate.

Planck Network targets this shortage by bringing together existing, underutilised GPU capacity, spread across different locations and owners, into a single network. The idea fits within the broader movement around decentralised physical infrastructure networks, also known as DePIN, in which hardware such as solar panels, mobile masts or, in this case, GPUs is offered collectively via a blockchain protocol.

The premise is that GPU hardware owners make their capacity available to the network and are rewarded for doing so, while buyers gain access at lower rates than those charged by centralised providers. Whether that plays out in practice depends on the scale the network manages to achieve and the reliability it can deliver.

The architecture has two layers

Technically, the platform consists of two separate layers, each with its own role. The first layer, referred to as Planck₀, functions as a so-called Layer 0 protocol. It coordinates computing power, security and message exchange between the various components of the network. This is where the orchestration of everything happening elsewhere in the system takes place.

The second layer, Planck₁, is a GPU-optimised smart contract platform. AI workloads can be executed directly on it, meaning developers can run their models and applications on the infrastructure without the involvement of a central party. Planck describes this layer as a Layer 1 platform, built on the foundation laid by Planck₀.

The protocol also offers the ability to launch so-called sovereign AI chains, separate blockchain networks tailored to specific AI applications that make use of the underlying Planck infrastructure. How this works in practice and which developers are already using it is not yet fully detailed on the company's website.

The founders come from different backgrounds

Diam Hamstra, CEO and co-founder, has a background as a serial entrepreneur in the Web3 world. He previously built EstateX, a platform for real estate investment via blockchain, and was involved in scaling multiple Web3 communities. At Planck, Hamstra focuses on the intersections of AI, DePIN and crypto infrastructure.

CTO Rohan Talwadia brings technical experience from Google Cloud, where he worked on early AI development tools. He was also a co-founder of BNB Tech and experimented with NFT experiences through the Decentraland platform. His background in cloud infrastructure and decentralised applications maps directly onto what Planck is looking to build.

The fact that the company operates from Harderwijk rather than Amsterdam or another major city is notable but not unique in the Dutch tech scene. The geographical location says little about the ambitions, which are clearly internationally oriented given the scale of the funding.

Where the network is headed

With €200 million in Series B capital, Planck has the resources to further develop its infrastructure and attract partners. The relevant question at this point is whether the network can connect sufficient GPU supply and demand to become a self-reinforcing system. Decentralised networks succeed or fail on their ability to reach a critical mass of participants.

Planck explicitly targets developers who want to build AI chains on top of the existing infrastructure. That makes the company a potential supplier for a next layer of AI applications, comparable in some ways to how Ethereum functions as infrastructure for thousands of applications built on top of it. The analogy does not hold entirely, but it gives an indication of the role Planck envisions for itself.

For the Dutch and European AI scene, Planck is an example of a company seeking to build infrastructure at a moment when Europe is actively exploring ways to reduce its dependence on American cloud providers. Whether decentralised networks can play a structural role in that effort is a question the market will answer more concretely in the years ahead.