Nebius Group (NASDAQ:NBIS) detailed a series of new cloud platform features during a product release event, highlighting additions across AI assistance, orchestration, cost controls, security, storage and developer programs.
The company’s presentation centered on Nebius Echo, a built-in AI agent designed to help users interact with Nebius Cloud through the console or command-line interface. The operator demonstrated Echo by asking it to create a virtual machine on a single H100 GPU. Echo returned the proposed configuration for approval before creating the VM, which the operator later confirmed appeared in the console.
“Our vision under Echo is, in general, that this is an AI that basically works as a very good trained customer architect or customer support that knows everything about you, like your tenant,” Tatevosyan said.
Nebius Adds Managed SkyPilot Support
The company also introduced a managed SkyPilot instance within the AI orchestration section of the console. SkyPilot is an open-source project for running AI workloads across cloud infrastructure. The operator said that, before this release, Nebius users had to deploy and manage their own SkyPilot API server. With the new managed option, Nebius hosts the control plane, allowing users to deploy an instance by selecting a name, platform and CPU preset.
Tatevosyan said Nebius is aiming to support multiple ways customers run AI workloads, including SkyPilot, Slurm, Anyscale and serverless options. He said Nebius does not charge for the SkyPilot control plane and charges only for compute used on its platform.
“We want customers to build AI the way they want to build,” Tatevosyan said. He added that Nebius sees SkyPilot gaining traction because it can help aggregate scarce infrastructure across clouds and data centers while offering primitives that are more accessible to AI researchers than some DevOps-focused tools.
New Budgeting and Key Management Tools Target Enterprise Readiness
Nebius also rolled out a Budget feature for spending alerts. Users can set a spending limit for an account, team, project, product or specific conditions over a defined period, such as monthly usage. If spending exceeds a selected threshold, Nebius sends an email notification. The operator noted that the console will not automatically stop resources when a budget threshold is exceeded.
Tatevosyan said the feature had been requested by customers and is part of a broader push toward enterprise readiness. He said Nebius is building features for large organizations with multiple teams operating within secured and governed environments, and noted that billing data can also be exported into external FinOps systems.
The company also introduced Key Management Service, allowing users to create customer-managed encryption keys to protect workloads. The operator said users can create symmetric keys for encrypting and decrypting data or asymmetric key pairs for use cases such as authentication, signing and secure communication. Users can also define key rotation periods.
Tatevosyan said requests for KMS came mainly from enterprise customers, though Nebius also received positive feedback from AI-native users. He said the service currently uses Nebius-managed hardware security modules and that the company plans to enable customers to bring their own keys, potentially by the end of the year or later.
Intelligent Object Storage Introduced for Tiering
Nebius announced intelligent object storage, a tiered storage option intended to reduce costs for data that is not accessed frequently. The operator said data initially sits in a warm tier and, if unused for 30 days, moves to a cold tier where users pay less until the data is accessed again.
Tatevosyan said the primary value is cost efficiency without requiring customers to configure tiering themselves. He said the feature is particularly relevant for customers working at storage scales of tens of petabytes, including those training models, running simulations or working with synthetic data in areas such as physical AI and multimodal models.
“If you use storage in serious scale, when you see that your storage cost is hitting, you need to think about tiering,” Tatevosyan said.
Builder Program, Certifications and Roadmap Updates
The event also highlighted the Nebius Builder Program, a free registration program that includes credits, planned office hours with Nebius engineers, community access and free access to Nebius certifications. The operator said two certifications were already live.
Tatevosyan said the company is working to support builders with credits and compute, while also managing abuse risks. He said credit availability and free-tier access may change over time as Nebius refines the program.
Looking ahead, Tatevosyan said Nebius plans to offer GB300 access on its platform and described that as part of the company’s effort to make advanced infrastructure available at smaller scales than are typically accessible to only the largest customers. He also said Nebius plans to continue investing in pay-as-you-go models, including preemptible machines with an auction model, as well as multi-data-center capabilities.
Other roadmap items discussed included cloud interconnect for connecting customer data centers to Nebius, production-ready serverless endpoints with autoscaling and multi-data-center capabilities, and broader model hosting support. Tatevosyan said Token Factory can already support bring-your-own weights for LLM models using vLLM or SGLang through support, while broader serverless support for other model types is planned for later.
During a security-focused Q&A, Tatevosyan said Nebius does not yet enforce post-quantum encryption everywhere but is preparing to move toward post-quantum encryption during this year or next year. He said the company views quantum-related risk as real but still low for the broader industry today.
About Nebius Group (NASDAQ:NBIS)
Nebius Group N.V., a technology company, builds intelligent products and services powered by machine learning and other technologies to help consumers and businesses navigate the online and offline world. The company’s services include Nebius AI, an AI-centric cloud platform that offers infrastructure and computing capability for AI deployment and machine-learning oriented solutions; and Toloka AI that offers generative AI (GenAI) solutions at every stage of the GenAI lifecycle, such as data annotation and generation, model training and fine-tuning, and quality assessment of large language model for accuracy and reliability.
