In March, Nebius signed a staggering $27 billion deal with Meta for five years of AI cloud capacity, signaling an unprecedented bet on specialized AI infrastructure. This landmark agreement followed the release of Aether 3.6, Nebius’s latest quarterly update to its AI Cloud, which became available in March, offering new tools for developers.
Nebius is making multi-billion dollar investments in specialized AI cloud infrastructure, but the broader AI cloud market is still evolving rapidly with many competing solutions. This creates a tension between hyper-specialization and the need for broader market appeal.
Nebius appears poised to become a formidable, specialized player in the AI cloud space, potentially reshaping the competitive landscape for large-scale AI deployments and setting a new bar for dedicated AI infrastructure.
New Features for Efficiency and Productivity
Aether 3.6 introduces Nebius Echo, an AI agent embedded in the web console that answers questions about infrastructure and executes commands, according to Techzine Global. This feature fundamentally changes how users interact with and manage their cloud resources, moving beyond traditional console interactions.
The update also includes a new Intelligent object storage class that automatically shifts data to a lower-cost tier after 30 days of inactivity, as reported by Techzine Global. Nebius focuses on both developer productivity through AI assistance and cost optimization for data storage, balancing high-performance needs with operational efficiency.
Strategic Investments and High-Performance Infrastructure
Nebius acquired inference optimization specialist Eigen AI for $643 million in May, according to Techzine Global. The acquisition confirms Nebius’s strategy: not merely scaling compute, but deeply optimizing for high-demand AI inference. Raw compute power alone is insufficient for sophisticated AI workloads.
Standalone application on NVIDIA H200 NVLink with Intel Sapphire Rapids costs $4.50 per GPU hour from June 1, 2026, according to Docs Nebius. The $4.50 per GPU hour high-tier pricing confirms Nebius's commitment to providing cutting-edge hardware for the most demanding AI tasks. For example, the NVIDIA RTX PRO 6000 Blackwell Server Edition GPU, with 96 GB of GDDR7 memory, exemplifies the specialized infrastructure Nebius offers. The focus on premium, high-performance hardware positions Nebius as a critical provider for enterprises tackling the most intensive AI computations.
Current Offerings and Pricing
Standalone application on NVIDIA L40S PCIe with Intel Ice Lake costs $1.35 per GPU hour, as detailed by Docs Nebius. This pricing offers a more accessible option compared to the premium H200 instances. Nebius aims to capture a wide spectrum of AI workloads, from bleeding-edge research to broader developer adoption, by segmenting its market with varied performance tiers.
Upcoming Serverless Capabilities
The DevPods serverless feature is in private preview and expected to be generally available in the next three months, according to Nebius. The expansion into serverless AI offerings aims to simplify deployment and scaling for developers. Nebius is building an AI-driven operational ecosystem, integrating AI agents and serverless features to create intelligent, managed AI development environments.
Frequently Asked Questions
Is Nebius AI Cloud Aether 3 available for serverless workloads?
Yes, Nebius supports serverless AI workloads. Jobs and Endpoints serverless features are currently in public preview, offering flexible options for developers.
Given its aggressive investment in specialized hardware and AI-driven operational tools, Nebius appears likely to solidify its position as a critical infrastructure provider for large-scale, high-performance AI deployments, particularly if the demand for dedicated AI cloud capacity continues its rapid ascent.
