theCUBE sets July 23 AI infrastructure series focused beyond GPUs

SiliconANGLE Media’s theCUBE will run a July 23 video series on enterprise AI infrastructure, with interviews planned with executives and experts from Supermicro, AMD, Vast Data, TensorWave and Crusoe. No financing, valuation or product pricing is attached to the program; the point is market positioning around the systems layer for AI inference, where CPUs, storage, networking and deployment software are getting more attention as GPU supply stops being the only bottleneck.

The series, called “Beyond The GPU: The Full-Stack Fight for Enterprise AI,” is being presented by theCUBE, SiliconANGLE’s livestreaming video studio. SiliconANGLE said the discussions will center on infrastructure needed for agentic deployments and enterprise production readiness, including storage, networking and open ecosystems.

The framing follows comments from AMD Chief Executive Lisa Su on the company’s May earnings call. Su told analysts that the current ratio of 4.5 GPUs for each CPU is expected to move closer to parity as AI agents and inference workloads demand more CPU capacity. That is a useful signal for infrastructure buyers: inference systems are being sold less as standalone accelerator clusters and more as full data-center designs with compute balance, data access and networking treated as first-order constraints.

Dave Vellante, chief analyst at theCUBE Research, has argued that organizations moving faster on AI adoption will be able to support more AI projects and build a compounding productivity loop. SiliconANGLE quoted him as saying data strategy will be a core part of that shift and that the transition will affect departments and operating models across the organization.

Supermicro’s role in the stack discussion

Supermicro is one of the companies expected to feature in the program, and its recent announcements show the kind of hardware consolidation the series is likely to emphasize. Last month, Supermicro said it had introduced a DCBBS Blueprint for high-performance computing based on Nvidia’s Vera Rubin NVL4 platform.

According to Supermicro, that design can support configurations with as many as 1,152 Nvidia Rubin GPUs and 576 Nvidia Vera CPUs. The company said the platform includes liquid cooling and Nvidia Quantum-X800 InfiniBand networking, with power scaling from 3.2 megawatts to 1 gigawatt. Those figures put the offering in the category of data-center-scale AI infrastructure rather than departmental AI servers.

Supermicro has also been adding products for smaller and edge deployments. The company announced an edge computing lineup using Intel Core processors and Intel Arc Pro B-series GPUs. Earlier this month, it also announced Kubernetes Edge AI appliances built with Red Hat and Everpure.

For enterprise buyers, the common thread is that inference workloads are pushing vendors to sell more than accelerator density. Faster retrieval, larger working data sets and persistent context for AI systems make storage and networking part of the performance equation. The vendors participating in theCUBE’s program cover several parts of that chain, including chips, systems, cloud infrastructure and data platforms.

How to watch

TheCUBE said the July 23 interviews will be available through its event page and YouTube channel, with coverage also available on SiliconANGLE. The company also distributes related programming through theCUBE Pod and its Breaking Analysis show on Apple Podcasts, Spotify and YouTube.

SiliconANGLE disclosed that theCUBE is a paid media partner for the video series. It also said Supermicro is the sponsor of theCUBE’s event coverage and that Supermicro and other sponsors do not have editorial control over content on theCUBE or SiliconANGLE.

This story draws on original reporting from SiliconANGLE.