AI Regulation Could Boost Nokia, STMicroelectronics, BofA Says

Bank of America analysts are challenging the view that tighter artificial intelligence regulation will simply slow the sector, arguing it could instead redirect spending toward a broader array of hardware for AI safety and monitoring. This potential shift could benefit European technology companies like Nokia and STMicroelectronics, according to a recent BofA research note.

A Shift from Training to 'Policing'

The market's assumption that regulation will primarily weaken semiconductor demand by slowing the development of large frontier models is "too narrow," the bank's strategists said. They argue that production-ready AI agents will require extensive infrastructure for monitoring, verification, containment, and auditability.

This creates what BofA calls "policing workloads" that could scale with the complexity of AI agents, not just user query volumes. The bank cited Anthropic research indicating that using one AI model as a safety filter for another increased inference costs by approximately 25%. While BofA noted this figure is "architecture-specific" and not an industry average, it illustrates the potential for significant computational overhead from safety measures.

This trend suggests a potential spending pivot from massive, concentrated training runs toward more distributed infrastructure for inference, control, and connectivity.

European Beneficiaries and Hardware Impact

According to Bank of America, this spending shift would favor specific hardware categories, including custom ASICs, CPUs, networking switches, and optical interconnects. The bank identified two European companies as potential beneficiaries:

Ad

• Nokia: Viewed as a winner through its "datacentre switching and optical transport" offerings. BofA highlighted the company's portfolio of high-speed switches designed for AI training and inference workloads.
• STMicroelectronics: Cited for its exposure to the optical components market.

The bank also noted that CPUs should see increased demand to handle agentic tasks like scheduling and safety checks, while networking intensity will likely rise as AI agents communicate with multiple models, data stores, and security services.

Potential Risks to the Outlook

BofA's strategists also outlined two key risks to their thesis. First, AI developers could embed guardrails so efficiently that the demand for incremental computing power is minimized. Second, regulation could slow the overall deployment of AI more than the expansion of policing infrastructure is able to offset it.

Regarding semiconductor capital equipment makers, the bank said it views the group as "largely agnostic" to this potential spending shift. However, it sees a possible long-term benefit if custom ASICs and connectivity silicon increasingly adopt leading-edge manufacturing nodes.