Taiwan-based computer giant Asus is set to introduce AI servers powered by Nvidia hardware that can be installed on-prem, offering a whole new layer of intelligence and security that it hopes will help attract even more customers.
The launch comes at a time when more companies are beginning to raise concerns over certain cloud-hosted AI workloads whereby sensitive data could become exposed, putting them at risk.
The new Asus AFS Appliance is set to use the company’s own Formosa Foundation Model which has been trained in-house by its AI division, TWS, which can comprehend and generate text with traditional Chinese semantics.
Asus AFS Appliance
Peter Wu, company President for Asus Cloud and TWS, explained on an interview with Bloomberg, the firm’s goal to produce AFS for around $6,000 per month, though a higher-spec version is expected to cost around $10,000 per month.
Under the hood, Nvidia’s A100 chips take center stage. Designed with data centers and AI at their core, they promise almost three times better performance (via Nvidia) compared with 2019’s V100 chips which succeeded the A100 while balancing energy efficiency.
Taiwanese telecom companies are expected to be the first users of AFS ahead of a global rollout later this year.
TWS promises to be “reliable, energy-efficient, and easily migratable” (via Asus), while its use in AFS Appliance is set to make generative AI more secure and private as it disengages with the cloud.
With promises of a more closed circuit, healthcare and banking customers are likely to be able to trust these types of systems more, enabling them to tap into time-saving productivity enhancement like AI without the concerns of data breaches.
