July 2026 confirmed that companies no longer have any difficulty accessing technology. The challenge has become transforming it into measurable business results without losing control over costs, security and data.
AI: the infrastructure is ready, but business isn’t always
As many as 90 per cent of organisations state that they have an architecture enabling large-scale data-driven decision-making. At the same time, only 23 per cent have begun to implement agent-based AI more widely. The proportion of companies where operational efficiency exceeds expectations has fallen over the past year from 80 per cent to 65 per cent. The survey covered 1,000 IT and business managers from the US, Europe and the Asia-Pacific region.
This is a significant difference. Simply having models, cloud infrastructure and data layers does not yet equate to productivity. The phase of counting the number of pilot projects is coming to an end. The focus is now shifting to measuring the cost of individual operations, the impact on revenue, customer service times and the quality of decisions.
AI is becoming a balance sheet item
The scale of expenditure is illustrated by the results of the largest technology companies. Meta allocated $31.08 billion to capital expenditure in the second quarter. The company’s costs rose by 55 per cent, its operating margin fell from 43 per cent to 31 per cent, and free cash flow amounted to just $784 million. At the same time, revenue rose by 28 per cent.
Microsoft spent $31.9 billion on investments in the most recent reported quarter, with around two-thirds of this going towards processors and graphics chips. Azure’s revenue rose by 40 per cent, and demand continued to outstrip available computing power.
The cost of AI therefore encompasses computing, data storage, integrations, observability and access security. Without FinOps and clear usage limits, even a successful project can undermine a product’s economics.
Cybersecurity is accelerating alongside AI
According to Verizon, 31 per cent of breaches now begin with the exploitation of software vulnerabilities. Ransomware is involved in 48 per cent of breaches, and the proportion of employees using unauthorised AI tools has tripled to 45 per cent.
Google Cloud points out that the time between a vulnerability being disclosed and its active exploitation has shrunk from weeks to days. In the case described, the attackers exploited the trust between the CI/CD system and the cloud platform in less than 72 hours.
An annual security audit is too slow to keep pace with threats that evolve within hours. This highlights the need to automate updates, verify machine identities and monitor the AI tools used by staff.
The cloud is growing, but Europe remains uneven
53 per cent of businesses in the European Union use paid cloud services, whilst 20 per cent of firms use AI. In Poland, the proportion of businesses using AI stands at 8.4 per cent. At the same time, 32 per cent of Polish firms reported an ICT security incident, which was one of the highest figures in the EU.
Demand is shifting ever more clearly towards data and platforms. At IBM, revenue from data solutions rose by 19 per cent, and at Red Hat by 11 per cent. Revenue from consultancy, however, remained unchanged. The performance of a single company does not reflect the entire market, but it clearly illustrates the shift in customer priorities: fewer general transformation projects, and more investment in specific data environments, automation and the hybrid cloud.
Regulation is becoming part of product architecture
From 2 August, EU transparency requirements concerning, amongst other things, chatbots, deepfakes and AI-generated content will come into force. The maximum fine may amount to €15 million or 3 per cent of a company’s global annual turnover.
July showed that the market is entering a phase of maturity. The cloud is now widely available, the use of AI is growing, and regulations are beginning to influence how services are designed. This means less room for actions taken under pressure from trends, and more scope for calmly selecting processes where technology genuinely helps. The best decisions do not necessarily have to be the most spectacular. They should be useful, sustainable and understandable to the people who will be using these solutions.
