5 Key IT Trends of 2026: From AI Agents and Zero Trust to Sovereign Clouds

To understand how the IT sector will change in the coming years, let’s look at five key technological directions that will determine the development of the industry. We will discuss what to expect in 2026 and explain what this means for business.

Artificial Intelligence and Machine Learning: From Image Generation to AI Agents

A few years ago, artificial intelligence was perceived by many as a tool for generating images or writing texts. Today, it is becoming the foundation of business processes and is transitioning from auxiliary functions to independently performing complex tasks.

The scale of AI usage is impressive: according to the McKinsey Technology Trends Outlook 2025 report, the share of companies applying AI in their work has increased to 78%. In Russia, the volume of the artificial intelligence market is also growing rapidly and may reach 500 billion ₽ in 2026.

Every year we observe an increase in demand for high-performance computing power for AI tasks. In response, Selectel aims to offer businesses not only IaaS solutions but also ready-made pre-configured platforms for training and inferring models. This approach will allow us to comprehensively meet business needs, helping companies build and scale their own AI solutions and significantly reduce project launch times.

Alexander Tugov

Director of AI Vertical at Selectel

One of the directions for the development of AI in 2026 will be the implementation of AI agents. Such systems analyze data, make decisions, and perform complex tasks without constant supervision. Research from HSE shows that 59% of Russian companies are considering the possibilities of implementing AI agents.

Unlike conventional chatbots or generative models, AI agents operate on the principle of "receive a task - plan a solution - execute." Here are specific examples of AI agents that businesses can implement:

• Analytical agents. Automatically analyze sales over the month, identify anomalies, and provide recommendations for management.

• Customer service. Functions as a team of specialized agents, where one consults clients, another processes sales, and a third updates product listings.

• Demand forecasting agents. Process large data sets and help businesses make informed decisions regarding purchases and assortment.

Despite the active development of technologies, the financial effectiveness of AI raises questions. According to NANDA, only 5% of companies have fully recouped their investments in AI. This is due to the fact that the level of autonomy of AI agents is still limited: systems may make errors in real-world scenarios.

According to Gartner's forecasts, by 2027, AI agents will automate up to half of all business decisions. The company highlights several key areas, including retail, logistics, marketing, and customer support.

Cybersecurity: Proactive Protection and Zero Trust Architecture

According to Selectel, in just the first half of 2025, 61,212 DDoS attacks were recorded and mitigated, totaling more than 9,652 hours. This is almost double compared to the same period in 2024, when 31,436 attacks were recorded.

At the same time, attackers have changed their strategy: they are shifting the focus from extremely powerful attacks to more frequent, prolonged, and distributed scenarios. This approach is associated with an increase in the size of botnets to nearly 5 million devices by 2025.

Today, cybersecurity issues are about resilience and often the survival of a business. Imagine: you have just deployed your infrastructure and launched a digital service, and within minutes, you are being attacked, ports are being scanned, or passwords are being guessed. Responding to incidents after the fact is no longer sufficient.

Dmitry Polyansky

Director of Client Security at Selectel

Companies can no longer rely on the old protection model — building a strong perimeter and hoping that everything inside is safe. The perimeter has become more vulnerable as employees have transitioned en masse to remote work, and corporate data and workflows have spread across different platforms and services.

The main risk factor has become single corporate sign-on — SSO and OAuth. This technology allows employees to access multiple systems with a single set of credentials. If an attacker gains access to a private network through a vulnerability, they can move freely between information systems and steal confidential data.

Zero Trust is becoming increasingly relevant for businesses of any size. In Russia, for example, the demand for Zero Trust solutions grew by 22% in the first quarter of 2025 alone. At the same time, the global market for such solutions is estimated at $38.37 billion and is expected to triple by 2030.

In addition, companies are transitioning from reactive protection (responding after an attack) to proactive protection. This includes not only security systems that learn to recognize anomalies in behavior but also choosing a reliable IT infrastructure provider that will ensure a secure foundation for business services. Artificial intelligence analyzes how employees work and instantly notices strange actions: accessing files during non-working hours, attempting to copy thousands of documents at once, connecting from a new device in another country.

Against this background, Selectel not only provides a secure infrastructure but also actively counters threats: in just the first half of 2025, the company recorded and blocked more than 61,000 DDoS attacks. Self-learning systems adapt to new threats faster than antivirus updates can be released.

The shift to Zero Trust is driven by a whole range of new threats, including:

• Deepfake attacks. Malicious actors create fake video calls impersonating executives and urgently demand money transfers or access credentials. Voices and faces look so realistic that even experienced employees fall for them.

• Use of AI agents by malicious actors. Professional groups use AI agents based on large language models that independently explore infrastructure, search for vulnerabilities, and plan multi-stage attacks. One person with an AI agent can achieve results equivalent to an entire team of hackers.

• Polymorphic malware. Malicious software constantly changes its code, bypassing traditional antivirus programs. Each copy of the virus looks different, and classic detection methods do not work.

• Attacks on AI models. Malicious actors exploit vulnerabilities in training data, modify system prompts, or take advantage of weaknesses in AI frameworks.

The construction and development of a cybersecurity system should start with a comprehensive audit and risk assessment. Organizations need to clearly identify which assets are critically important, where sensitive data is stored, who has access to it, and, importantly, what threats and risks are relevant specifically to critical assets. Without understanding the current state of protection, it is simply impossible to build an effective strategy.

Denis Polyansky

Director of Client Security at Selectel

To determine the current state of protection, experts suggest starting from the principle of Assume Compromise — building defenses as if an attacker has already penetrated the network, and focusing on rapid detection, isolation, and elimination of threats. In the era of deepfake, AI agents, and polymorphic malware, a Zero Trust strategy ceases to be an option and becomes a necessity for companies.

Low-code/No-code platforms: when businesses create software without programmers

In 2025, the traditional approach to development faced a shortage of highly qualified IT specialists. Organizations were forced to wait while the IT department, for example, improved the CRM or automated business processes. To cope with the growing volume of tasks, companies began to delegate some of them to artificial intelligence:

• No-code is a builder where applications are assembled from ready-made blocks by dragging elements with a mouse. Not a single line of code needs to be written. A manager without technical education can create a request form, an approval process, or a simple portal.

• Low-code works differently: the platform provides ready-made visual blocks for 80% of tasks but allows adding custom code (JavaScript, Python) when complex logic is needed. It’s a hybrid that accelerates development but does not limit the capabilities of professionals.

Applications can be created not only by programmers but also by those who previously only wrote technical specifications. Gartner forecasts that by 2026, 80% of users of Low-code tools will be "citizen developers": business analysts, product managers, and department heads without formal IT training. At the same time, the popularity of Low-code/No-code (LCNC) will not diminish the need for highly skilled developers.

The LCNC market is growing at an explosive rate. By 2025, it is expected to reach $37.39 billion, and by 2032, it could grow to $264.4 billion. A similar situation exists in Russia. About 70% of companies have already implemented Low-code solutions. Almost half of the companies (48%) are creating prototypes in a few weeks instead of the usual 6–12 months for traditional development, and 57% of organizations have reduced their dependence on external contractors.

At Selectel, we strive to optimize development and use ready-made libraries or components whenever possible. Additionally, we actively implement No-code solutions specifically in internal services. A vivid example is project management systems, where business processes can be configured without a single line of code. The built-in automator provides complete freedom of action, so there is no point in spending time developing plugins.

Alexander Tugov

Director of the AI Vertical at Selectel

In large companies, LCNC can solve quite serious tasks. RPA scenarios automate routine tasks: the system independently checks the details in contracts, routes documents for approval, and processes vacation requests. The economic benefits from implementation are expressed in concrete figures: companies that use the LNCN approach reduce overall IT costs by 30–45%.

With the development of AI, LCNC solutions have reached a new level. If previously platforms were limited to ready-made templates and visual blocks, now users can create unique applications that address specific business tasks with the help of AI.

Alexander Tugov

Director of the AI Vertical at Selectel

LCNC has several limitations. Firstly, as businesses grow, it is difficult to scale such systems. Secondly, without an IT department, there may be issues with security and system compatibility during development. Thirdly, under extreme loads, LCNC applications may fall short compared to optimized code, so a completely no-code solution is not optimal for high-load systems.

LCNC is best suited for quick internal services, testing business hypotheses, and integrating small processes at minimal costs. For complex corporate platforms requiring deep customization and high loads, traditional development remains irreplaceable.

Cloud Technologies and XaaS Models: Hybridization, Localization, and Digital Sovereignty

The Russian cloud market is developing rapidly. In 2025, it grew by 30–35% and reached about 400 billion ₽. The growth outpaces global averages. Drivers include equipment shortages, demand for high-performance computing for AI, stricter data protection requirements, and a lack of highly qualified IT personnel.

Today, Russian companies increasingly trust cloud technologies. This trend is confirmed by the steady growth of the cloud services market: in 2025, its volume exceeded 400 billion ₽, and by 2030, it is expected to reach around 1 trillion ₽.

At the same time, Selectel notes a consistent trend towards a multi-cloud strategy. This allows businesses to flexibly distribute loads among different providers, take advantage of each, and minimize dependence on a single one.

Konstantin Ansimov

Product Director at Selectel

According to a Gartner study, data geopatriation is becoming one of the six key trends impacting infrastructure and operations in 2026. This involves the transfer of information to cloud services and data centers under Russian jurisdiction. Businesses want to control where their data is physically stored, who has access to it, and under what laws the provider operates. Only with the emergence of legislative requirements and industry standards have companies had the opportunity to move part of their services to the cloud.

Alongside the desire for control and isolation, a contrasting, more flexible approach is evolving — the XaaS (Everything-as-a-Service) model. Within this framework, a company does not purchase hardware and software but rather rents everything: infrastructure, development platforms, and ready-made applications. According to the auditing company "TeDo," the XaaS market in Russia is projected to reach up to 300 billion rubles by the end of 2024, and may reach 480-500 billion rubles in 2026.

XaaS forms the basis for the rapid implementation of AI, analytics, and automation. This is especially relevant for small and medium-sized businesses: not all can afford expensive servers for training neural networks, so they opt for cloud rental with GPUs. The demand for GPU infrastructure from companies has doubled, particularly in retail, finance, and industry.

Alongside the XaaS model, a complementary approach is developing — edge computing. Unlike the centralized "everything as a service" model, it involves processing data "at the edge of the network," closer to its source and user. Both approaches can effectively coexist within a single IT strategy of a company: flexible subscription to cloud services for core business processes and edge computing where response time is measured in milliseconds.

Data integration and processing: from disparate sources to a unified analytics framework

Previously, companies launched Big Data as an experiment. Now it is a matter of survival. When data comes from CRM, cloud systems, IoT sensors, social networks, and partner services, the ability to consolidate them and bring them to a unified format becomes critically important. According to HSE, 61% of Russian companies have already implemented data integration solutions, and another 26% plan to use them. This is no longer a trend — it is a basic requirement for growth.

Data integration is when information from dozens of different places is collected into one system. For example, customer purchase data from CRM is combined with their support interaction history, geolocation from their smartphone, and activity on social media. Data enrichment is the next step. This is when contextual information is added to the basic details. Information about what products the customer viewed beforehand, what the weather was like in their city, and what promotions competitors had is linked to the customer's purchase.

In retail, Big Data analytics increases the average check by 5–15% through precise personalization of offers. Artificial intelligence models for demand forecasting work only on complete and enriched data; otherwise, they provide erroneous forecasts. And a quick response to market changes is possible only when you have a unified analytics framework, rather than 15 disparate reports from different systems.

In the data era, the main competitive advantage is the ability to quickly integrate heterogeneous information and extract valuable insights from it. We see how companies that have created a unified data management system make more accurate decisions and respond faster to market changes.

Alexander Grishin

Head of Data Storage Product Development at Selectel

The Big Data market in Russia is growing rapidly. By 2025, its volume will exceed 430 billion ₽, with an annual growth of 25–35%. The Big Data Association forecasts that by 2030, the market could reach 1.2 trillion ₽. Cloud technologies are becoming the core of this growth: by 2026–2027, 75% of organizations will use them as a basic platform for working with data.

At the same time, demand for PaaS is also increasing. In 2025, platform services became an essential part of companies' technological strategies. However, businesses are facing a shortage of specialized personnel and expertise in managing complex platforms: Kubernetes, systems for data storage and processing, as well as solutions for working with AI/ML, including platforms for training and inference of models.

Ready-made PaaS solutions are becoming the optimal choice — they are not just infrastructure under SLA, but fully managed services that help automate processes, accelerate time-to-market for solutions, and use infrastructure resources more efficiently.

PaaS solutions are becoming critically important for working with big data and AI/ML. They allow companies to focus on developing analytical models and extracting value from data, rather than spending resources on configuring and maintaining complex infrastructure.

Alexander Grishin

Head of Data Storage Product Development at Selectel

Summary: from experiments to measurable effectiveness

The Russian IT market in 2026 is approaching the mark of 4.5 trillion ₽, but the nature of investments is changing. Companies are increasingly evaluating technologies not by novelty, but by return on investment and impact on business metrics. The mentioned trends do not operate in isolation, but form comprehensive requirements for a company's IT landscape. AI-based automation requires cloud infrastructure, data needs protection, and the low entry barrier provided by LCNC accelerates the implementation of solutions.

In 2026, the focus shifts to the practical application of technologies for specific business tasks. Now companies are automating processes that directly affect costs, implementing protection under the Zero Trust model, using cloud infrastructure with consideration for data localization requirements, and integrating disparate systems into a single framework for analytics. Technologies are becoming not a competitive advantage, but a necessary condition for maintaining market positions.