Visibility that drives elasticity: Introducing Capacity Blocks and Capacity Dashboard
Visibility that drives elasticity: Introducing Capacity Blocks and Capacity Dashboard
Continuous scarcity of AI compute has become a major decision-making factor for ML teams developing and serving GenAI models. AI leaders are often forced to adjust their business goals, development plans and even headcount expansion, based on the availability of GPU resources. In this environment, any opportunity to secure compute becomes a short-term competitive advantage, giving teams the ability to accelerate experiments, shorten iteration cycles and move product roadmaps forward faster.
As a purpose-built AI cloud provider, Nebius has always treated elasticity as a core value of our platform. But true elasticity requires more than simply having GPUs available — it also depends on the visibility, transparency and predictability that help teams plan and use their compute efficiently. From day one, we’ve offered on-demand access, long-term reservations and, more recently, introduced preemptible VMs to make every unit of capacity accessible.
Today, we’re expanding that elasticity even further with the introduction of Capacity Blocks and Capacity Dashboard — two new capabilities that bring additional clarity to GPU capacity management and give users full confidence in the compute they rely on.
Build confidently with Capacity Blocks
Capacity Blocks introduce a clear and intuitive way to understand the GPU resources available to your organization at any moment. They show exactly how much reserved capacity you have in Nebius and help you align that capacity with the needs of your workloads and development plans.
With Capacity Blocks, teams can easily track how their GPU resources are distributed over time: when each portion of capacity is booked, when additional expansion is planned, and how future availability aligns with upcoming experiments or production milestones (Figure 1). This visibility allows organizations to plan confidently, ensuring their pipelines run smoothly without unexpected pauses or unused windows of paid capacity.
Figure 1. Capacity Blocks shows different categories of booked capacity allocated for the selected month
In addition, Capacity Blocks streamline the process of creating instances. Users can associate each VM with a specific capacity interval, keeping workloads organized and simplifying capacity management across teams and projects.
Figure 2. Project-based capacity allocation within each Capacity Block
Within each Capacity Block, you can also see how GPUs are allocated between projects (Figure 2). This brings an additional layer of transparency for ML and research teams, enabling more efficient planning, collaboration and workload distribution.
Operate predictably with Capacity Dashboard and API
We’re also introducing a public Capacity API that brings real-time transparency to on-demand and preemptible GPU availability across Nebius AI Cloud. This API gives users direct access to accurate capacity data, enabling them to automate decisions, integrate availability checks into their workflows and build more predictable compute operations.
Figure 3. Example of an API response showing available GPUs
This visibility improves the experience of launching services and virtual instances across the platform. Whether users are creating clusters for any of our managed services or apps, or spinning up individual VMs, they can rely on up-to-date information to choose the right region, fabric and GPU type with confidence.
Our partners who resell or integrate Nebius compute also benefit from this accuracy — the API provides consistent, reliable capacity data that can be surfaced directly in a partner’s own tools and provisioning flows.
For those working in the Nebius web console, the same information is available through the new Capacity Dashboard, a visual widget that shows real-time GPU availability in each region. Instead of checking availability during VM creation, users can explore capacity visually and select resources more easily. The Capacity Dashboard and API enhance transparency across the platform and help teams make faster, better-informed decisions on consuming infrastructure.
Building a more transparent platform
Availability of AI compute remains a critical factor for ML teams and AI-driven businesses, shaping both day-to-day productivity and long-term strategic decisions. Even small gaps in available infrastructure can slow down experimentation, disrupt research timelines or leave valuable capacity underutilized.
At Nebius, we’re committed to giving our users not only powerful GPU infrastructure but also the clarity and control needed to use that infrastructure to its fullest. With Capacity Blocks and Capacity Dashboard, we’re bringing greater transparency and predictability to capacity management, making elasticity more accessible and more actionable for teams of all sizes.
We will continue to evolve the Nebius platform with features that enhance visibility, improve user experience and ensure that AI practitioners always have the compute they need, exactly when they need it.
