TensorWave and MK1 Join Forces to Boost AI Inference on AMD Cloud

Published 6 months ago

In a move aimed at optimizing AI inference, TensorWave has formed a strategic partnership with MK1, as announced on May 22, 2024. This partnership aims to leverage AMD hardware to rival the dominant player in the market, NVIDIA.

MK1 and TensorWave Commit to Collaboration

MK1, established by early Neuralink engineers, is known for handling large language models and other intensive AI tasks. The company’s co-founder, Paul Merolla, and TensorWave’s CEO, Darrick Horton, have voiced their dedication to this partnership. They plan to improve the efficiency of AI inference tasks, which currently make up approximately 70% of AI workloads.

The Role of Each Partner

MK1 contributes its expertise in software development, specially designed for AMD’s hardware, to this alliance. The partnership is expected to utilize TensorWave’s cloud infrastructure to enhance efficiency and reduce costs for AI operations.

Paul Merolla underlined the potential of AMD’s technology to rival and possibly surpass existing solutions. “The potential of AMD hardware to catch up to and even surpass existing solutions was a game-changer for us. This partnership with TensorWave aligns perfectly with our goals,” commented Merolla.

Echoing these sentiments, Darrick Horton emphasized their objective to make high-performance AI inference more accessible and affordable. “We are getting closer to outperforming existing solutions every day. Our goal is to democratize access to high-performance AI inference, making it available and affordable for everyone,” stated Horton.

Potential Impact on the AI Industry

The partnership between TensorWave and MK1 aims to not only enhance performance but also provide a viable alternative in a market currently dominated by NVIDIA. This could significantly alter the AI hardware landscape by introducing competitive solutions, giving consumers more options.

MK1’s software currently operates on over a thousand GPUs and processes over 15 trillion tokens each month, highlighting the scale at which these solutions are deployed. This partnership promises to effectively tackle real-world AI workloads, with the aim of delivering robust and user-friendly services.