Rosmorport to spend 99.2 million rubles on an All-NVMe storage system and servers for AI workloads

Rosmorport is ready to invest 99.2 million rubles in high-performance servers and an All-NVMe storage system. The procurement is necessary to update the enterprise's IT infrastructure and is directly designed for artificial intelligence and machine learning workloads.

What Rosmorport Is Purchasing

At the center of the tender is the supply of high-performance server equipment and an All-NVMe storage system. For a state company, this is not a cosmetic upgrade but a replacement of the fundamental layer of computational infrastructure—it determines how quickly data can be processed, new services deployed, and resource-intensive scenarios supported.

Notably, the procurement description explicitly mentions AI and machine learning workloads. This means it's not just about archival storage and standard corporate systems, but a more demanding digital environment.

The mention of All-NVMe is particularly telling. Such storage systems are built on solid-state drives and are typically chosen where data access speed, low latency, and stable performance under parallel loads are critical. For machine learning tasks, this is a sensitive parameter: slow storage quickly becomes a bottleneck even with powerful servers. Thus, the procurement architecture itself shows that the enterprise is focused not just on volume but on the performance of the entire data processing pipeline.

Why This Enterprise Needs It

Rosmorport operates in a complex infrastructure environment where IT systems must be stable and predictable. If the company is building AI workload capability into the upgrade, it may signal a shift from isolated pilots to more systematic use of analytics, automation, and machine learning models. In such projects, it's important not just to deploy a model but to rapidly read large data sets, store results, run multiple processes simultaneously, and maintain performance as loads grow.

In practice, such a procurement is rarely done for a single service. Rather, it creates a platform on which analytics tools, internal digital products, and new computational scenarios for different departments can be deployed in parallel.

In the case of Rosmorport, scale also matters: when modernization is valued at nearly 100 million rubles, it's no longer a local procurement for a single department but an infrastructure circuit with headroom for future expansion.

• Fast access to data sets for models
• Acceleration of corporate analytics and file processing
• More predictable performance under peak loads
• A unified foundation for new AI and ML services

Why This Signal Matters

The procurement volume itself—99.2 million rubles—makes the news noteworthy. On the Russian market, this is a good indicator that demand for AI infrastructure is coming not only from technology companies, cloud providers, or banks. Organizations for which IT has been primarily a support function are now beginning to make such procurement decisions.

Now infrastructure is becoming part of operational logic: without it, it's harder to increase data processing speed, automate processes, and build application models on top of internal data stores.

What's also telling is that the emphasis is on high-performance hardware and fast storage, not on abstract "digital transformation." Such procurement language typically indicates a more mature approach: first a computational foundation is built, then services, models, and application scenarios are deployed on top of it.

For the market, this is an important marker. AI in the corporate segment is increasingly seen not as an experiment for demonstration purposes, but as a workload that requires infrastructure designed in advance.

What This Means

The Rosmorport tender story illustrates a straightforward shift: large organizations are already investing not just in software and pilots, but in the foundation for AI projects. When servers and All-NVMe storage for machine learning are included in the procurement from the start, it marks a transition from talking about AI to building an environment where such workloads can actually run at scale in production without constant trade-offs in speed and availability.