exaBITS has launched the world’s first truly accelerated decentralized computing network, set to revolutionize affordable AI. By pooling global GPU resources and optimizing their usage, they dramatically reduce the cost of training and deploying large AI models.
Already, exaBITS has onboarded over 65,000 GPUs worldwide during its private beta phase, attracting a wide range of clients from budding startups to prominent unicorns in the AI industry and saving them hundreds of thousands of dollars. The company’s unique access to supply and boosting consumer-grade GPUs has led to an ever-growing active customer base. This base benefits from unparalleled cost efficiency, accelerated performance, and access to the essential machines for their projects, positioning exaBITS as the leading choice in the global AI sector.
Recommended AI News: Riding on the Generative AI Hype, CDP Needs a New Definition in 2024
One example is with one of their large enterprise customers who used an institutional cloud compute provider early in their journey. They quickly realized that the high operational costs of this compute became a major resource drain with costs amounting to over $150K monthly. When they turned to exaBITS, their compute cost dropped by 71% and their performance increased to 130%.
Recommended AI News: OPENLANE Launches Visual Boost AI to Pinpoint Vehicle Damage
The challenge in AI is sourcing powerful GPUs like Nvidia’s H100, A100, and AMD’s MI300 and MI200, which have become highly sought after and as a result, scarce. exaBITS addresses this by utilizing not just these advanced GPUs but also transforming consumer-grade GPUs like the 3090/4090 to match their performance. This capability is crucial, considering many companies face years-long waiting lists to acquire enterprise-grade GPUs or experience bottlenecks to secure instances at other cloud compute providers.
H100 / A100 Supply & Consumer GPUs
At the heart of the exaBITS network are T3 data centers, mining facilities and massive distributed consumer-level GPUs, encompasing over 65,000 GPUs, including numerous H100s and A100s. A key to providing affordable AI at scale is their ability to enhance the performance of weaker GPUs through the blockchain. They’ve achieved up to 15x cost reduction and 90% lower costs, paving the way for a robust decentralized network that effectively utilizes consumer GPUs. This reduces dependence on enterprise grade machines, increasing resource availability across the network.
The Roadmap Ahead
exaBITS has proven its ability to meet the technical demands of the market holistically and the roadmap shows a clear path to the future. Part of the roadmap includes an ML DevOps Platform that intelligently manages and accelerates training and inference, ensuring the production efficiency and service quality of AI models. This makes exaBITS not only disrupt, but surpass, incumbents like AWS and Google. In addition to acceleration, exaBITS will deploy its Anti-Fault-Technology. This solves the high failure rate of GPU clusters, which is the hidden killer of AI projects. This technology improves system stability by more than 50%, significantly enhancing the efficiency of resource usage. Their researchers have built an auto-recovery feature that immediately identifies the moment a GPU fails live and allows training to resume at the failure point as opposed to having to restart months-long training, reducing the risk of prolonged training and increased costs when workloads are interrupted.
The team, with rich backgrounds in companies like Bitmain, Meta, and (mention here something related to data centers or distributed compute) and various mobile operators, brings a wealth of experience in brokering deals for data centers and mining facilities.
exaBITS team is comprised of scientists with deep expertise in cloud computing, AI, and blockchain, backed by a business team well-versed in Web3 business models. Together, they bring over 50 years of collective experience in these fields.
Recommended AI News: WiMi Developed RPSSC Technology With Multiple Advantages in Hyperspectral Image Processing
exaBITS is tapping into a vast resource of underutilized GPUs with a technology stack and team capable of harnessing them for powerful AI inference and training. With access to top-tier GPUs like the H100 and A100, and the ability to transform consumer GPU’s to perform like enterprise level GPUs, exaBITS positions itself as the premier decentralized cloud compute provider, democratizing AI development and reducing cost barriers for AI projects and developers globally.
[To share your insights with us, please write to sghosh@martechseries.com]
Comments are closed.