EPAM Systems has cut its full-year revenue-growth expectations despite delivering a stronger second quarter, adding to evidence that the enterprise AI investment boom is reaching software-engineering and consulting suppliers less evenly than the infrastructure companies selling chips, data-centre capacity, and electrical equipment.

Second-quarter revenue reached $1.415 billion, up 4.5% from a year earlier and 3.4% on an organic constant-currency basis. GAAP operating income increased 20.4% to $152.2 million, while the operating margin rose from 9.3% to 10.8%; on EPAM’s adjusted measure, the margin increased from 15% to 16.4%.

Earnings also improved, with GAAP diluted earnings per share increasing 26.3% to $1.97 and non-GAAP diluted earnings per share rising 22% to $3.38. EPAM nevertheless reduced its 2026 revenue-growth forecast to between 3.2% and 4.2%, compared with the 4.0–6.5% range it set after the first quarter in May.

Expected organic constant-currency growth has narrowed from 2.5–5.0% to 2.0–3.0%, while third-quarter revenue is forecast at between $1.410 billion and $1.425 billion. At the midpoint, that would represent about 1.7% year-on-year growth, leaving the company with improving profitability but a much more restrained top-line outlook.

AI spending is moving unevenly

Much of the current technology cycle is being measured through exceptional demand elsewhere in the supply chain. Semiconductor manufacturers, data-centre developers, electricity suppliers, and equipment makers are booking large orders as companies and cloud providers build the physical capacity required to run increasingly demanding AI models.

Professional-services companies sit further along the adoption curve. Their revenue depends on organisations moving from infrastructure spending and experimentation into projects requiring application redesign, data engineering, systems integration, software development, process change, and sustained implementation work, and those decisions can be delayed even while the underlying technology budget remains substantial.

EPAM has been positioning itself inside that transition through AI-native engineering and integrated consulting, supported by partnerships with major cloud, data, and AI technology providers. The aim is to help customers rebuild applications and operating processes rather than simply install another software product.

Its second-quarter result shows that model still producing growth, while stronger margins suggest EPAM is managing utilisation and delivery costs effectively despite the reduced revenue outlook. Total headcount stood at roughly 62,850 at the end of June, including about 56,650 delivery professionals, only slightly higher than three months earlier.

Stable workforce numbers are notable for a services company because labour capacity has historically placed a direct constraint on growth. Generative AI changes that relationship by allowing engineers to produce, test, document, and analyse more code with automated assistance, potentially letting suppliers handle greater workloads without increasing headcount at the same rate.

The commercial consequence is more complicated. If AI makes an engineering team materially more productive, customers may expect projects to require fewer billed hours, while suppliers will attempt to price around outcomes, proprietary methods, reusable tooling, or higher-value expertise instead of the volume of labour assigned to an account.

Those pressures can improve margins while putting conventional services revenue under pressure. EPAM’s figures contain some of that tension: non-GAAP operating income rose 14.7% against revenue growth of 4.5%, while the adjusted operating margin expanded by 1.4 percentage points.

Stronger profitability alongside slower revenue can reflect successful productivity management, but it does not yet show that AI transformation work has created a growth engine comparable with previous waves of cloud migration and digital product development. Procurement remains slower where projects depend on data quality, access permissions, model risk, security, regulation, and integration with existing applications.

Those dependencies create work for consultancies and engineering suppliers, but they can also make projects harder to move from experimental budgets into large programmes. At the same time, capital committed to accelerators, cloud capacity, data platforms, or model subscriptions is not simultaneously available for external consulting, allowing different parts of the same AI strategy to experience very different demand conditions.

EPAM’s distributed engineering model adds another dimension. The company built much of its organisation around international delivery and technology labour markets across Europe and other engineering centres, while AI-assisted development reduces the advantage attached purely to the number and location of engineers.

Architecture, specialist domain expertise, customer proximity, and the ability to implement systems inside complicated organisations consequently become more valuable. EPAM is responding with what it calls a forward-deployed engineering approach, placing technical expertise closer to customer problems while combining consulting with software-development capability.

The direction resembles changes across the wider services industry, where suppliers are attempting to move away from labour-intensive outsourcing towards smaller teams supported by automation and reusable AI tooling. Existing commercial models cannot be replaced immediately, however, because large organisations still operate legacy systems, long procurement cycles, regulated processes, and complicated data estates.

Customers also have access to increasingly capable AI coding tools themselves, raising the threshold for external work that can command premium rates. Much of the opportunity therefore lies in areas where the difficulty sits in organisational systems, architecture, governance, and integration rather than generating code in isolation.

EPAM still expects higher earnings than its weaker revenue trajectory might imply, with full-year non-GAAP operating margin forecast between 15.5% and 16% and adjusted diluted earnings per share expected at $13.08–$13.24. The company is therefore reporting slower growth rather than a collapse in profitability.

The combination provides a useful counterpoint to industrial companies booking record AI-related orders. Enterprise investment is clearly moving through data centres, chips, networks, and power systems, but services suppliers have to wait for that capacity to become deployed applications, redesigned processes, and funded implementation programmes. EPAM’s reduced forecast suggests that conversion remains uneven even while AI begins to change the economics of the engineering work itself.