Akash Network, a decentralized cloud compute marketplace, announced that it is offering GPU support in its sixth mainnet upgrade.
This comes at a time when the demand for GPU resources exceeds supply as artificial intelligence startups develop new innovations to capitalize on the AI gold rush.
“For those looking for on-demand compute, it’s not always easy to find. Users of CSPs are forced to accept permissioned access and the risk of vendor lock-in, and this is only if the desired GPU units are even available,” Akash Network explained in a blog post.
The platform added that top-tier GPUs are typically reserved for exclusive use to major clients due to high demand.
Describing itself as the “Supercloud for AI,” Akash Network said that it aims to address these issues by providing permissionless access to computing resources. This includes GPUs from various providers, ranging from independent to hyperscale.
New Features and Enhancements in Mainnet Upgrade
Akash’s sixth Mainnet upgrade introduces GPU support, stable payments, and Take Rates. Network providers can now offer GPU resources to global deployers, and both parties will be able to settle payments in USDC.
The main feature of the upgrade is Akash’s open-source marketplace for high-density GPUs. The company intends to modify Akash’s current cloud infrastructure, enabling network providers to support GPUs.
Akash said that it will focus its initial support and testing on NVIDIA GPUs given the AI industry’s preference for them. In future, the network plans to expand support and testing for other chip manufacturers including AMD and others.
According to the network, over 1,300 people signed up to participate in the Akash GPU Testnet in August. It hosted NVIDIA H100s, and A100s, along with consumer-grade GPUs from the 30-series and 40-series.
“Access to consumer-grade GPUs is one way the Akash Supercloud stands apart from other cloud providers. These models are typically overlooked, even though they can often run inference on less memory-intensive AI models,” the company said.
The next phase of the testnet will involve deploy various AI models on each GPU type to determine the most suitable GPU type for specific deployments.
Those interested in accessing high-performance NVIDIA GPUs can view available models at this link.
Source: mPost
