The sharp increase in global semiconductor revenues by 21 per cent to $793 billion in 2025 is a signal to policymakers far more important than a simple cyclical recovery. The data confirms a fundamental reallocation of capital in the IT sector. Infrastructure budgets are no longer the domain of general computing, shifting aggressively towards specialised AI acceleration. For CFOs and CIOs, this means the end of waiting for component prices to fall – high infrastructure costs are the new structural norm, not a temporary anomaly.
A key finding from the analysis of the revenue structure is the growing role of critical components. AI-related semiconductors already account for almost a third of total sales, and the bottleneck has shifted from processors to memory. High-bandwidth memory (HBM) has generated more than $30 billion in revenue, taking 23 per cent of the DRAM market. For supply chain strategists, this is a clear wake-up call: the availability of AI servers in the coming quarters will be dictated not so much by computing power as by the availability of HBM memory dice, of which SK Hynix and Samsung are the main beneficiaries.
Equally important for long-term planning is the unprecedented consolidation of market power in the hands of a single supplier. Nvidia, with revenues of $125.7 billion, has become the de facto monopoly of the computing standard, distancing Samsung by $53 billion. This raises a significant risk of vendor lock-in for companies building their own data centres. On the other hand, Intel’s dramatic drop in market share to just 6 per cent is forcing organisations to re-evaluate their technology partnerships and retention strategies for legacy systems based on x86 architecture, which is losing its market leadership status.
Gartner’s forecasts, suggesting an increase in AI infrastructure investment to $1,300 billion by 2026 and the dominance of AI chips in sales by 2029, point in the right direction. Organisations that fail to factor the growing cost share of specialised silicon into their CAPEX models risk underestimating digital transformation budgets on an unprecedented scale.
