BNP Paribas Research highlights that most semiconductor companies remain optimistic about supply-demand dynamics heading into 2026. The global financial firm recently conducted a Silicon Valley bus tour, meeting with executives from AMD (AMD.US), NVIDIA (NVDA.US), Intel (INTC.US), Applied Materials (AMAT.US), Astera Labs (ALAB.US), Credo Technology (CRDO.US), Lumentum Holdings (LITE.US), Seagate Technology (STX.US), Marvell Technology (MRVL.US), and Western Digital (WDC.US).

"In AI computing, all companies we engaged with expressed strong confidence in 2026 market performance and subsequent capacity ramp-up, with demand far outstripping supply," said BNP Paribas analysts David O'Connor and Karl Ackerman in an investor note. "Key investor concerns revolve around power supply, ASIC competition, and supplier financing capabilities."

Both AMD and NVIDIA noted tightening power supply across the U.S. However, they believe the U.S. government is taking steps to alleviate power constraints, viewing this as a short-term issue. "There’s consensus that power is the primary bottleneck in the AI arms race," O'Connor stated. "While NVIDIA acknowledges power challenges, they don’t see an insurmountable energy barrier and expect construction to accelerate. Given U.S. power capacity constraints, foreign investment may rise."

The data center ecosystem’s visibility has extended to multiple quarters, driven by hyperscalers’ multi-year roadmaps and NVIDIA’s 9-12 month lead times, improving supply chain efficiency and pricing dynamics.

Another discussed concern is the rise of custom chips, such as Google’s (GOOGL.US) Tensor Processing Units (TPUs) using ASIC technology. "Following recent TPU announcements, ASIC competition has come into focus," O'Connor noted. "Industry players emphasize that TPUs are optimized for specific cloud providers/workloads (e.g., Anthropic, GCP) rather than general cloud services (unlike GPUs), so their market share growth shouldn’t be extrapolated beyond existing TPU adopters."