IT Trends 2023: key developments in software and hardware

Why summarize 2023 and how to read trends

Tech news often feels like a race: “everything’s obsolete,” “you must change now,” “otherwise you’ll fall behind.” Summarizing 2023 helps separate what actually affects people and organizations from the noise around loud announcements.

A simple filter: will an event affect your daily processes in the next 6–12 months? If it makes automating text or customer support easier — that’s practical. If a “new standard” isn’t yet supported by your systems and vendors, it’s often wiser to watch rather than rush to buy.

It’s useful to split trends by layers. Software updates fast (AI models, office tools, security), while infrastructure and server hardware live longer and require planning. A typical 2023 mistake was trying to "fix" a problem with a new application when the bottleneck was the network, storage, backups or server performance.

To read news without panic or marketing hype, keep a short checklist:

• what exactly improves: speed, cost, security, convenience;
• what’s needed to make it work for you: data, people, budget, support;
• what’s the risk if you do nothing until the next quarter.

Even without big budgets you can start with the basics: tidy up data, review backups, enable minimal cyber hygiene and honestly measure where infrastructure bottlenecks are. Then the IT trends of 2023 become a checklist of what to verify in your IT estate.

Main topics of 2023

IT trends in 2023 are best seen by the questions that became “default” for businesses and governments: where to get capacity, how to protect data, how to avoid supply interruptions, and how to calculate TCO for years ahead.

What was talked about most

Five themes repeatedly appeared in projects and procurements.

First — generative AI: moving from “try it” to pilots with clear benefit (knowledge-base search, document drafts, operator assistance). Second — infrastructure for AI and analytics: accelerators, fast networks, storage and data preparation. Third — cybersecurity as a minimum: MFA, backups, network segmentation, access control, updates.

Fourth — the hybrid model, where some systems run in the cloud and some on-premises due to requirements, cost and control. Fifth — import substitution and technological independence: more attention to local manufacturing, service and transparent supply chains.

What accelerated, what slowed and what became more expensive

Adoption of AI tools and basic security practices accelerated: after high-profile incidents many stopped postponing hygiene. Large-scale migrations done “for trend” slowed — teams preferred targeted improvements with quick impact.

Prices rose for scarce components and support: accelerators, fast SSDs and licenses, plus service and spare parts. Against this backdrop, infrastructure and security returned to the top of priorities: downtime, leaks or lost data usually cost more than planned modernization.

Experiments more often moved to production via pilots, limited scopes and clear metrics: request processing speed, reduced downtime, recovery time after failure.

Generative AI in 2023: benefits, limits, risks

In 2023 generative AI became notable not because it “learned to think” but because it learned to quickly produce drafts: text, answers, ideas, summaries. This is where IT trends of 2023 most quickly turned conversations into pilot projects.

The benefit is clearest for tasks where speed and volume matter more than perfect accuracy: first-line support answers based on templates and a knowledge base, draft emails and instructions, Q&A-style internal search, short summaries of calls and meetings.

But expectations often didn’t match reality. The problem is usually not “intelligence” but responsibility: a model can sound confident and still be wrong, confuse facts or invent details. A second pain point is data. If you feed the system unchecked documents, it will confidently repeat garbage. If those documents contain personal or confidential information, the risk of leakage increases.

So many companies adopted a simple rule: AI drafts, a human approves the result. In practice this led to a few clear measures:

• don’t input sensitive data without strict rules;
• label AI-generated materials and require verification;
• keep sources and versions so it’s clear who changed what;
• predefine “allowed tasks” where errors are not critical.

From a people perspective, roles became more important than model names: a process owner (what to automate), an editor/reviewer (quality and tone), a data specialist (which documents can be used), and a security owner (what must never be done).

Infrastructure for AI: why GPUs, data and network matter

In 2023 it became clear that AI project success often depends not on a “clever model” but on hardware and data. Training and active use of models require compute, memory, fast networks and storage. So infrastructure — not just software — firmly entered the IT trends of 2023.

GPUs and other accelerators are needed where heavy workloads occur: training, image and voice processing, large language models, large-scale stream analysis. But they aren’t needed for every scenario. For a knowledge-base chatbot or forecasts from small tables, a good CPU server and correct data handling can be sufficient, especially if you use a pretrained model.

Data matter more than model “magic”

In 2023 the approach to data became more down-to-earth. Questions that were previously postponed started being asked right away: where are the data, who has access, can they be used legally, and how good are they? If a dataset contains many duplicates, errors or murky usage rights, AI will produce confident but incorrect answers.

To prevent AI infrastructure from becoming unexpectedly costly, break down TCO into clear parts: electricity and power limits, cooling (especially in dense racks), network (often the bottleneck between storage and accelerators), storage (fast NVMe for active data and a separate zone for backups), maintenance and spare parts.

A good start for a university or clinic is a pilot: one server for experiments, separate storage for datasets and clear access rules. It’s easier to scale when you can see what actually delivers value.

Server hardware 2023: focus on efficiency and support

In 2023 server choice became more practical. The priority shifted from “most powerful” to “how much useful work we get per watt and per rack unit.” These IT trends favored energy efficiency, density and predictable scalability: add resources as workloads grow, rather than replacing the whole platform.

Teams looked more at steady performance under real loads: virtualization, databases, VDI, backups, and sometimes dedicated nodes for analytics. Equally important was how a server behaves under load: noise, heat dissipation, power requirements, rack compatibility and network integration.

Interest also grew in reliability and reparability. The reason is simple: downtime costs more than peak spec numbers. So teams checked spare parts availability, real recovery times and the ability to service equipment without waiting for rare components.

Before buying, teams typically verified:

• power consumption and cooling for typical loads, not peak;
• the ability to expand memory, disks and network without changing platform;
• a 3–5 year support and upgrade plan;
• parts availability and delivery speed;
• service conditions (diagnostics, on-site engineers, service hours).

Storage: NVMe, backups and recovery without surprises

In 2023 disk conversations increasingly started not with capacity but with how quickly a system can be restored after failure. These trends tie to data growth, remote work and the fact that downtime costs more than the drives themselves.

NVMe does speed things up, but not uniformly. The biggest differences appear where many small operations and parallel requests occur. For example, an accounting database or virtualizing multiple servers on one host often hits disk latency.

NVMe usually shows real benefits in databases and transactional systems, VDI and virtual machines, analytics with frequent reads, app caches, and build/test environments with many small files.

If your bottleneck is network, CPU or a slow application, switching to NVMe might be barely noticeable. A typical case is a file server where users copy documents over Wi‑Fi.

Why backups and recovery became more discussed than raw capacity? Because two numbers matter: how much data you could lose (recovery point) and how many hours to restore service (recovery time). Many organizations have backups but never tested recovery until the first incident.

A practical storage-tier rule: “hot” data are needed daily and must be fast; “warm” data are used occasionally and can be cheaper; “cold” is an archive kept long-term and must be reliable even if accessed rarely. For archives, object storage is often chosen: durability and cost matter more than access speed.

Before buying or starting a project, ask minimum questions about fault tolerance:

• what happens if one or two drives fail, or a controller fails;
• how backups are done and where they are stored (separate site or at least separate contour);
• how recovery tests are performed and who is responsible;
• realistic component replacement and support timelines;
• expected data growth in 12–24 months.

If you purchase servers and storage in Kazakhstan, specifically clarify who provides 24/7 support and where the service network is located. That can matter more than a 10% speed advantage on tests.

Clouds and hybrid: practical solutions over extremes

In 2023 many stopped arguing “cloud or on-prem” and asked a more useful question: where should each specific workload live? Hybrid architectures became a notable part of how IT trends were interpreted: businesses need fast provisioning, predictable costs and control.

Full migration to public cloud was often hindered not by “fear” but by data residency requirements, network latency, vendor dependence and complex approvals. The typical approach became: keep critical systems and data in your own environment, and put test environments, web services and temporary workloads where they’re easiest to spin up and shut down.

“Sovereign infrastructure” is not a slogan but a set of clear answers: where data are physically stored, who has hardware and update access, how you receive support, and what happens if supplies or a provider become unavailable.

Procurement rules and equipment origin strongly influenced choices. Where local content and supply-chain transparency mattered, teams preferred predictable on-prem infrastructure with clear support.

Hybrid works when responsibility boundaries, redundancy and recovery scenarios are defined in advance, not just when a “right” vendor is chosen.

Cybersecurity in 2023: the mandatory minimum

In 2023 it became clear: attackers don’t target only the big players. Common issues were phishing (email and messengers), ransomware, leaks from weak or reused passwords, and misconfigurations in cloud and servers. Many incidents began with one compromised account and ended with a department or entire organization stopping work.

Basic hygiene that reduces risk too much to ignore

In short, basic hygiene looks like this:

• MFA for email, VPN, admin panels and all critical systems;
• regular OS, server and application updates with a clear maintenance window;
• backups following the 3-2-1 rule and recovery tests, not just a checkbox;
• network segmentation so an infected workstation doesn’t lead to servers and backups;
• least-privilege principle and separate admin accounts for administration.

Employee training proved as important as new products. Explaining real examples — “an invoice from accounting” or “urgent delivery” — reduces clicks on malicious links faster than another blanket ban.

To avoid cybersecurity becoming report-driven, measure with simple metrics: share of accounts with MFA enabled, percent of critical updates applied on time, incident response time, success rate of recovery tests, number of devices with outdated software. Such figures help turn IT trends of 2023 into concrete tasks and budgets.

PCs and workstations: what changed in fleet refresh strategy

In 2023 many stopped chasing top clock speeds and maximum memory. Predictability, ease of support and a consistent user experience became priorities. These trends are especially visible in organizations with large fleets.

Demand grew in three areas: all-in-ones for front offices and classrooms (fewer cables, simpler workspaces), workstations for engineering and graphics tasks, and planned refreshes of standard PCs when old models start costing employees time due to slowness and compatibility issues.

The major shift was standardization. When a fleet includes 5–7 different configurations, spare parts inventory, diagnostic time and downtime risk grow. So organizations more often chose 1–2 standard models, unified OS images, common security policies and identical accessories.

Selection criteria for schools, hospitals, government and offices usually boiled down to reparability, stable supply of identical configurations, quiet operation and ergonomics, compatibility with business applications and peripherals (printers, scanners, tokens), local servicing and predictable response times.

A realistic scenario: IT modernization in 2023 (example)

Imagine a regional clinic in Kazakhstan updating its server room and part of the workstation fleet. The budget is approved mid-year and the rollout must finish by quarter-end. There are also local content procurement requirements and clear expectations for support because downtime in a registry or lab quickly causes queues and complaints.

The team first divided everything into three buckets: keep, replace urgently, and move to virtual infrastructure. Old power-hungry, failure-prone servers were replaced, and some secondary services (test stands, internal portals) were virtualized to reduce physical machines and simplify backups.

Equipment selection took time not for dry specs but for practical questions: delivery times and spare parts availability in the country, engineers’ working hours, node replacement speed (disks, PSUs), regional service coverage and a clear escalation process.

The project closed on time: key servers and part of the PC fleet were updated, and support and repair procedures were agreed in advance. Two notable IT trends show up here: a preference for reliability and predictable support over peak specification numbers.

They postponed lower-impact items to the next year: storage expansion and a separate AI analytics pilot. That helped avoid overloading the team and stay within budget without risking stability.

Step-by-step: turn 2023 trends into a plan for 2024

Trends are useful only when turned into tasks with dates and owners. A 2024 plan starts not with buying hardware but with clarity: what must run without interruption and what happens if it stops.

5 steps that deliver practical results

1. Map services: email, 1C/ERP, file shares, video surveillance, VDI, backups, customer-facing services. Mark critical items (how many hours of downtime are acceptable) and single points of failure.

2. Prioritize into three groups: security (MFA, updates, segmentation), resilience (redundancy, backups, DR), performance (latency, response times, maintenance windows).

3. Check bottlenecks: CPU load, insufficient RAM, IOPS and disk latency, network throughput, overheating and power. Often the issue is not the whole server but storage or network.

4. Prepare 2–3 architecture options and compare by TCO: purchase, licenses, electricity, rack space, support, delivery times, downtime risks. The “cheaper now” option is often costlier in a year.

5. Plan rollout and support: pilot, migration, training, runbooks, monitoring, spare parts. For example, a hospital can update servers and backups in Q1 and user workstations and peripherals in Q2 to avoid interrupting care.

Common mistakes when choosing IT and hardware after 2023

After 2023 many decisions were emotional: “we must have what leaders have” or “let’s wait.” Budgets were spent poorly and issues surfaced in operation when changes became expensive.

What usually goes wrong

The most typical mistake is buying equipment for an unlikely peak load. Daily workloads sit idle while money is spent. It’s far better to model the realistic load profile and leave a reasonable margin.

Another frequent error is underestimating support. It’s not just warranty years but repair times, spare parts stock and regional service. For organizations with branches this is critical: “waiting for a part from another city” quickly offsets any procurement savings.

People also forget the “surroundings” of the server: power, UPS, racks, cooling and cabling. New server hardware can expose overheating, tripped breakers and messy racks.

A separate problem is a heterogeneous fleet of generations and incompatible platforms. Mixed approaches complicate support and troubleshooting becomes a lottery.

How to reduce risk

A simple rule set helps:

• size for typical load and growth over 12–18 months, not a one-time maximum;
• test support in practice: response times, parts stock, regional service;
• audit power and cooling before procurement so new equipment runs in normal conditions;
• reduce platform diversity so updates and monitoring are unified;
• bake security in from the start: segmentation, updates, backups and access control.

Quick checklist before procurement and next steps

When reading IT trends of 2023 it’s easy to want the newest thing. More useful is to check fundamentals that determine long-term stability and support costs.

A short checklist before finalizing specs and budget:

• load requirements: which apps, how many users, peak patterns, growth margin for 12–18 months;
• reliability: acceptable maintenance window, need for failover cluster, target RPO/RTO;
• timing and logistics: launch date, delivery risks, compatibility with existing systems;
• regulation and procurement: local content (if applicable), certifications, data storage rules;
• support: 24/7 coverage, response times, spare parts, who will physically come onsite and where.

A short example: if you refresh workstations while introducing a new internal server, decide in advance who tests typical scenarios (printing, access to 1C/ERP, video calls) and who signs off acceptance. Without this, procurement “checks the boxes” but launch drags on for weeks.

Next steps

Hold a 60-minute session with IT, security and procurement and agree the target architecture: what stays on-prem, what goes to cloud, where you need headroom for GPU/network, and where it’s not justified.

Then compare 2–3 options by TCO and risk: “minimal changes,” “normalization” (unified models and support), and a locally produced hardware option if supply timelines and transparency matter.

If you need practical guidance for infrastructure and support in Kazakhstan, rely on local manufacturers and integrators’ experience. For example, GSE.kz manufactures computers and servers in Kazakhstan and provides 24/7 technical support through a nationwide service network, which helps reduce the risk of prolonged downtime.

Assign an owner for ongoing support in advance: who is responsible for updates, monitoring, backups and incidents. That’s often more important than another 10–15% of theoretical performance on paper.