VAST Data and the Evolving Neoclouds

VAST Data has secured its position as the dominant storage supplier for neoclouds following a 600 PB capacity win at Australia’s Sharon AI, according to reports from Blocks and Files. This win reinforces a shift in the cloud market where neoclouds are establishing sovereign datacenters and moving into the managed service provider (MSP) market to compete with established hyperscalers.

The structure of the cloud market has fundamentally changed since 2018. Following a crypto market crash, miners transitioned from cryptocurrency mining to using GPUs for 3D rendering, scientific simulations, and machine learning services. The launch of ChatGPT in late 2022 created a sustained frenzy for AI foundation model training. Because neoclouds held the necessary GPUs and infrastructure while hyperscalers did not prioritize GPU-based compute, the neoclouds captured the initial boom.

The financial scale of this shift is evident in recent revenue trends. Neocloud suppliers saw revenues of $23 to $25 billion in 2025, more than triple their 2024 revenues. Market leader CoreWeave reported $5.13 billion revenues in 2025. This collective neocloud revenue was significantly more than the $15 billion-plus AI revenue run rate AWS held earlier this year. While AWS overall revenues were far greater at $128.7 billion in 2025, the hyperscalers are now in a catch up position, attempting to use their own silicon alongside Nvidia’s hardware.

The market is currently moving away from a reliance on a few large AI model training customers toward a broader base as AI inferencing expands across the enterprise. This transition requires highly efficient storage architectures. VAST Data has emerged as the primary supplier for these neoclouds by focusing on all-flash storage efficiency.

As AI workloads scale, the industry must move away from older methodologies like SSD-HDD tiering that risk leaving GPUs idle. The winners in this era will be those who can maintain high data reduction and flash efficiency to keep compute resources active.

How will hyperscalers bridge the infrastructure gap to reclaim their position against specialized neoclouds?