PC sales rebound by the end of 2025 amid a component crisis and a transition towards AI

Global PC shipments surged by 9.6% in the fourth quarter of 2025, reaching 76.4 million units, according to IDC. This growth is attributed less to structural demand and more to anticipated purchases spurred by the impending end of Windows 10, supply chain tensions, and the introduction of new architectures featuring AI-dedicated components.

After several hesitant quarters, the PC market experienced a significant rebound towards the end of the year. IDC reports that manufacturers and enterprises advanced some of their orders to secure configurations ahead of anticipated supply chain challenges in 2026. Jean-Philippe Bouchard, Vice President of Worldwide Mobile Device Trackers at IDC, describes the environment as “extremely volatile,” characterized more by inventory fluctuations than by lasting demand.

Throughout 2025, a total of 284.7 million PCs were shipped globally, marking an 8.1% growth. Lenovo maintained its lead with 70.8 million units shipped, followed by HP Inc. with 57.5 million and Dell Technologies with 41.1 million. This stable hierarchy reflects a consolidation in the professional market around a few key players capable of weathering supply chain disruptions.

Memory Shortages Create Imbalances

The primary vulnerability in the market now lies with DRAM and NAND memory, a growing portion of which is consumed by data centers and AI infrastructures. IDC warns of a potential sustained tightening, which could directly impact the configurations available to businesses.

Jean-Philippe Bouchard indicates that “PC memory specifications may be adjusted to preserve available inventories.” This could mean that machines are delivered with less memory than anticipated, particularly as component prices influence the pricing strategies of manufacturers and distributors.

Rise of AI-Enabled PC Architectures

The market rebound is also accompanied by a technological transformation. New professional platforms from Intel, AMD, and Qualcomm are now incorporating neural processing units alongside traditional processors and graphics chips to facilitate local AI inference tasks.

These architectures allow for local processing of voice, image, and text, as well as local data indexing, but they do not empower PCs to train or execute large language models. Limitations in memory, bandwidth, and thermal dissipation restrict their use to compact models and supported recovery scenarios applied to local data.

Persistent Gap Between Hardware and Use Cases

In business settings, these so-called AI PCs primarily function as contextual computing terminals. They store and index sensitive information locally and orchestrate requests to remote platforms for heavy computations. Consequently, workstations are becoming integral parts of hybrid architectures rather than standalone AI engines.

The market paradox stems from this discrepancy. Investments in specialized architectures are growing rapidly, yet the practical applications remain focused on document automation, semantic search, and content production assistance. In an environment marked by component volatility, the value of professional PCs will increasingly depend on the ability of software vendors and integrators to transform these new platforms into measurable productivity gains for enterprises.

Scroll to Top