Trade policy and tariffs have emerged as the semiconductor industry’s dominant risk, overtaking labor and talent constraints for the first time in more than two decades, according to KPMG’s latest Global Semiconductor Outlook. The shift reflects how geopolitical pressure is reshaping investment decisions, network design, and long-term capacity planning across an industry that remains deeply global even as governments push for localization.
Trade Policy Moves to the Center of Strategic Planning
In KPMG’s 21st annual survey of 151 semiconductor executives conducted in the fourth quarter of 2025, tariffs and trade policy ranked as the industry’s top concern, displacing talent challenges that have dominated recent outlooks. More than half of respondents said geographic diversification of supply chains is now a primary focus, while 45% identified building flexibility against geopolitical shocks as their top strategic priority.
The urgency reflects rising uncertainty around cross-border flows of components, equipment, and finished chips. Despite aggressive domestic investment plans, the U.S. semiconductor supply chain remains import-intensive. According to trade data, the U.S. imported $22.6 billion worth of semiconductors last year, with large volumes sourced from Vietnam, Thailand, and Malaysia. That exposure helps explain why executives view diversification less as a cost optimization exercise and more as a continuity safeguard as demand accelerates.
Notably, these concerns have not dampened growth expectations. Ninety-three percent of respondents still expect revenue to rise in 2026, buoyed by sustained demand from artificial intelligence workloads and data center expansion. In the survey, 73% of executives cited AI as the top revenue driver, followed by cloud and data centers, wireless communications, and automotive applications.
Incentives Spur Investment as Tariffs Raise New Tradeoffs
Government policy is reshaping the operating environment through a mix of incentives and penalties. On the incentive side, funding and tax credits under the CHIPS and Science Act continue to draw capital into U.S.-based manufacturing. Texas Instruments announced plans in June to invest more than $60 billion across seven fabrication facilities, supported in part by up to $1.6 billion in federal funding. Amkor Technology has also expanded its U.S. footprint, breaking ground on a packaging and testing campus in Arizona backed by $407 million in CHIPS Act funding.
At the same time, executives remain conflicted. Just over half of survey respondents said domestic advanced fabrication is necessary, yet an equal share warned that accepting government funding could limit market agility or constrain innovation. Those concerns sharpen as policy shifts toward enforcement. The Trump administration’s proposal of 100% tariffs on semiconductor imports, later delayed, signaled a willingness to use duties as leverage, a move publicly supported by the Semiconductor Industry Association.
Operational constraints add another layer of complexity. Thirty-four percent of executives flagged access to energy as a major concern over the next three years, underscoring the infrastructure demands of advanced fabs. Labor availability remains an issue as well, even if it no longer tops the risk list.
Still, demand momentum is difficult to ignore. Taiwan Semiconductor Manufacturing Co. reported a 40.8% year-over-year increase in third-quarter revenue, driven largely by AI-related orders. That performance helped lift KPMG’s Semiconductor Industry Confidence Index to 63, its third-highest level in two decades.
When Resilience Becomes a Cost Variable
The survey suggests the industry is entering a phase where resilience itself is being priced into strategy. Diversification, domestic capacity, and policy alignment are no longer abstract goals; they are becoming fixed considerations alongside yield, utilization, and capital efficiency. As trade policy hardens, the competitive gap may widen between companies that can absorb higher structural costs in exchange for supply assurance and those still optimized for a frictionless global model that no longer exists.
