Sandisk: Building a Foundation with Enterprise SSD Data Layers to Empower NVIDIA MGX AI Factories
Sandisk is providing data-layer support for the NVIDIA MGX AI factory architecture with enterprise storage products such as the SANDISK SN861 NVMe SSD, reducing data transfer bottlenecks to GPUs through PCIe Gen 5 connections and standard NVMe interfaces. Sandisk says its eSSDs can maintain stable performance in high-density, thermally constrained, sustained-load environments, and it brought the E1.S form factor to modular AI systems relatively early; in the future it will evolve toward E3.S form factors that support liquid cooling and higher power.
Body · AI translation
By Jeff Fochtman, Senior Director, Flash Products Marketing and GTM, Sandisk
AI infrastructure is undergoing a fundamental transformation. The large-scale, centralized cluster architectures originally built for model training are gradually evolving into more dynamic systems to enable continuous operation of large-scale inference. For example, the NVIDIA MGX AI factory ecosystem was born in response to this trend, and data centers are moving toward being purpose-built for generating and delivering AI outcomes to handle increasingly rich workload demands.
Why data is both the bottleneck and the breakthrough opportunity
Today's AI workloads—from model training and fine-tuning to, especially, real-time inference—all depend on continuous access to massive datasets. Models must retrieve weights, context, and intermediate data such as the key-value cache (KV Cache) with extremely low latency and stable throughput. When the data layer cannot meet these demands, even more advanced GPUs can only sit idle and wait. This is precisely why architectures such as NVIDIA MGX are attracting increasing attention. MGX uses a modular design that enables system builders to flexibly combine compute, networking, and storage resources according to specific AI workload requirements. Under this architecture, storage is no longer merely supporting infrastructure, but a key element that provides performance support for achieving scale deployment. Sandisk is helping to define this critical data layer.
The role of high-performance storage
Modern NVMe SSDs play a central role in this transformation, providing the low latency and high throughput needed to keep AI pipelines running continuously and efficiently. Sandisk's enterprise storage portfolio, represented by the SANDISK® SN861 NVMe SSD, is built precisely for such scenarios, designed to deliver the outstanding performance and consistency required for model training and inference pipelines.
As AI deployments continue to scale, storage must not only deliver high-speed performance but also integrate seamlessly into increasingly dense, modular, and thermally constrained systems. This is why Sandisk continues to drive innovation in evolving areas such as connectivity, thermals, and form factors.
Keeping data flowing with high-speed connectivity
In AI systems, data continuously moves between storage, GPU, and network layers. PCIe Gen 5 high-speed connectivity helps Sandisk SSDs meet data transmission demands, delivering data to accelerators fast enough to minimize bottlenecks. The standard NVMe interface enables smooth integration on the NVIDIA MGX platform.
The hardware integration density of AI servers is approaching its limit. GPUs, network equipment, and storage devices are tightly packed into limited space, further intensifying thermal pressure. Sandisk enterprise SSDs (eSSDs) are designed to maintain stable performance under continuous load, even in such constrained environments.
Form factors suited to modular architectures
The E1.S specification has become the mainstream option for high-density NVMe deployments, balancing efficient airflow management and convenient device servicing in large-scale deployments. Sandisk was also one of the earlier vendors to bring the E1.S specification to modular AI systems.
Looking ahead, the industry is evolving toward the E3.S specification to support higher power and better thermal management, including adaptation to liquid-cooling scenarios, while providing stronger capacity expansion capabilities to meet the needs of next-generation AI workloads.
Ecosystem-wide deployment at scale
Sandisk's enterprise SSD product matrix is compatible with modular architectures such as MGX and can be widely used across major original equipment manufacturer and cloud service provider ecosystems. This helps products such as the SANDISK® SN861 NVMe SSD scale flexibly across diverse AI system designs and multiple capacity points.
The bigger vision
AI factories represent the next stage in data center evolution, a stage that continuously generates intelligent outcomes at scale. From product platforms such as the SANDISK® SN861 NVMe SSD to form factors such as E3.S, Sandisk continues to iterate its enterprise SSD technology to help build a solid data storage foundation on which AI factories depend.