Omdia forecasts AI PCs will account for 60% of shipments in 2027, with 69% of workers willing to pay more for an AI PC.

IT Home, 10 30: Enterprise AI spending continues to surge. As leading AI vendors shift to token-based pricing, managing these costs has become increasingly unpredictable. For PC OEMs, ODMs, chipmakers, channel partners, and managed service providers, the core challenge is clear: where should AI workloads run, and what role will PCs play? Forecasts indicate that AI PC shipments will account for 60% of the total by 2027, with the installed base of local AI growing steadily through device refresh cycles. Enterprise AI spending is still climbing, and token-based pricing makes cost control difficult. Most enterprises plan to increase their AI investments next year. According to Omdia's "2026 AI Market Maturity Survey," 86% of organizations with an AI budget expect spending to rise further in 2027. Leading AI vendors are moving from per-seat or per-request models to token-based billing. Since the cost of a task remains unknown until execution completes, expenditure management becomes significantly harder. Agentic AI exacerbates this issue; a single task can trigger a long chain of pay-per-call model invocations. Despite these risks, enterprise AI adoption and spending continue to accelerate. Historically, most enterprise AI ran in the cloud, driving up costs as companies began reevaluating which workloads belong there. A hybrid infrastructure spanning public clouds, on-premises servers, and edge AI is emerging as the solution. However, large-scale deployment of local AI requires significant technical capabilities that many companies lack. This presents a key opportunity for the entire PC ecosystem: helping enterprises clarify how and where to run AI on devices. Pure cloud AI drives up costs and compliance risks, while edge AI has yet to fully convince enterprise buyers. Beyond cost concerns, as more company data leaves devices, pure cloud AI introduces worries around compliance, security, and data sovereignty. As agentic solutions move into production, these risks may intensify, often accompanied by untrackable management overhead. Blind spots also exist on the user side. The Omdia "PC Horizons Consumer Survey" reveals that 78% of respondents who use PCs for work utilize AI at least several times a week, yet only 12% of this group rely on AI provided by their workplace or school.