VAST Data is the leading storage vendor for neoclouds, strengthened by a recent 600 PB deal with Australia’s Sharon AI. Rapidly evolving neoclouds build sovereign data centers and expand into the MSP sector to compete with mainstream hyperscalers.
Six to seven years ago, the cloud market consisted of x86/Arm-based hyperscalers, low-cost niche MSPs and GPU-powered crypto mining facilities. The 2018 crypto crash drove miners to repurpose GPUs for 3D rendering, scientific simulation and ML cloud services. These workloads gradually evolved to support LLMs, giving rise to neoclouds that challenged hyperscalers, which overlooked GPU-centric cloud services at the time.
Late 2022’s ChatGPT launch sparked explosive, sustained demand for foundation model training. Neoclouds held abundant GPU infrastructure while hyperscalers lacked dedicated GPU resources, fueling their rapid revenue surge.Neoclouds generated $23–25 billion in collective 2025 revenue, over triple their 2024 earnings. Leader CoreWeave alone hit $5.13 billion, surpassing AWS’ $15B+ 2025 AI run rate, despite AWS’ $128.7B total annual revenue. Hyperscalers missed the GPU-as-a-service AI boom and are now catching up with custom and Nvidia GPUs.
The neocloud market is shifting from heavy reliance on large LLM training clients to wider enterprise adoption driven by mainstream AI inference workloads.As a pure all-flash storage provider, VAST Data dominates neocloud storage deployments. Co-founder and CTO Alon Horev cited 2:1 data reduction efficiency on neocloud flash workloads. He stressed traditional SSD-HDD tiering causes latency and GPU idling risks. Neoclouds now fully adopt all-flash architectures to maximize GPU utilization, pushing VAST to refine flash resource efficiency.
Cost concerns over long-term flash data retention are negligible for neoclouds. Their top priorities are operational simplicity and consistent GPU uptime, core competitive advantages. High GPU and infrastructure operational costs far outweigh flash storage expenses, making performance and reliability paramount.
Neocloud storage requires extreme reliability, as infrastructure outages paralyze entire GPU compute clusters. Horev noted large-scale video model training may consume up to 500 PB of VAST storage across thousands of GPUs, expanding VAST’s critical role in massive AI workloads.
Customers evaluate VAST against alternatives based on overall hardware and software costs, uptime stability, and the simplicity of its unified large-scale file and object storage.Neocloud operating models are maturing. Early custom, single-tenant dedicated deployments delivered high margins, while modern providers are shifting to hyperscaler-style operations, decoupling infrastructure from end users to enable flexible, on-demand resource consumption.
Enterprise clients prefer per-GB pay-as-you-go pricing over upfront large cluster investments, driving higher margins across GPU, CPU, storage and database services. Hyperscalers’ success hinges on automated, logically partitioned multi-tenant infrastructure, flexible short-term subscriptions and minimal operational friction — a model neoclouds are adopting.
Neoclouds are building large-scale multi-GPU data centers with secure, isolated multi-tenant partitioning for compute and flash storage. This flexible consumption model replaces rigid long-term contracts, unlocking greater profit potential.
Traditional telecom and MSPs are upgrading into neoclouds, evaluating GPU service margins and market differentiation amid strong AI compute demand. Yet AI inference services are far more complex than basic VM hosting, requiring up-to-date expertise in evolving open-source models, customized model deployment and elastic scaling.
Neocloud-grade AI infrastructure demands specialized skills including InfiniBand networking, liquid cooling and Kubernetes, replacing legacy virtualization expertise. Though challenging, this upgrade is achievable for existing infrastructure teams.VAST empowers neoclouds’ enterprise expansion with mature enterprise-grade capabilities. Unlike AI training workloads that only need fast storage, enterprise inference requires FIPS compliance, auditing, robust security, multi-protocol support and user data protection.
VAST’s fully validated multi-tenant architecture delivers strict data isolation, zero neighbor interference, end-to-end encryption and external key management. These enterprise security features help neoclouds meet compliance standards and accelerate enterprise customer onboarding.Neoclouds will not replace hyperscalers but will expand into MSP-style enterprise AI services. Meanwhile, traditional MSPs are upgrading to offer AI capabilities, bringing dual growth opportunities for VAST’s neocloud storage business.
Beijing Qianxing Jietong Technology Co., Ltd.
Sandy Yang/Global Strategy Director
WhatsApp / WeChat: +86 13426366826
Email: yangyd@qianxingdata.com
Website: www.qianxingdata.com/www.storagesserver.com
Business Focus:
ICT Product Distribution/System Integration & Services/Infrastructure Solutions
With 20+ years of IT distribution experience, we partner with leading global brands to deliver reliable products and professional services.
“Using Technology to Build an Intelligent World”Your Trusted ICT Product Service Provider!
Vietnamese
