Planck relies on a network of global processing units worth $40 million, the protocol said.
The Planck Network, an infrastructure protocol targeting the artificial intelligence boom, has launched a layer-0 blockchain aimed at supporting decentralized AI networks, the company said Tuesday.
The blockchain is designed to serve as foundational infrastructure for AI applications, particularly those built for decentralized physical infrastructure networks (DePINs). These networks combine hardware, token incentives and distributed processing to create alternatives to centralized resources, such as cloud services.
The move aligns with a growing push within the crypto industry to bring Web3 principles of decentralization to AI development, a sector still dominated by centralized players such as OpenAI and Google.
“Currently, high-performance AI compute is heavily centralized in the hands of a few tech giants," a spokesperson for Planck told Cointelegraph.
Planck is one of several blockchain projects working on alternative approaches to AI decentralization. Bittensor, for example, focuses on decentralized machine learning, while Fetch.ai facilitates the creation of AI agents through its decentralized platform.
Planck says its blockchain will generate protocol revenue from transaction fees, SDK usage, and developer tooling. Meanwhile, GPU operators providing compute resources will be rewarded with the protocol's native token based on machine's uptime (proof-of-connectivity) and actual utilization (proof-of-delivery).
Related: AI advances audit coverage for innovative multi-language virtual machine
GPU rental and AI chip crunch
Most of Planck’s revenue comes from renting out GPUs and compute contracts. The company says its hourly, on-demand model slashes costs by up to 90% compared to traditional cloud services.
Since February, the company generated $1.5 million in revenue, mostly from renting GPU power. The company also competes with other infrastructure providers like Vast.ai, CoreWeave and Lambda, all of which are capitalizing on the ongoing AI chip shortage.
That shortage has fueled explosive growth in the GPU-as-a-service sector. According to Precedence Research, the market hit $4 billion in 2024 and is expected to grow at a 23% annual rate, reaching $32 billion by 2034.
“Centralized AI compute is prohibitively expensive. By decentralizing the GPU network, Planck can offer up to 90% cost savings,” a Planck spokesperson said.
Magazine: AI Eye: AI is good for employment says PWC — Ignore the AI doomers