-+ 0.00%
-+ 0.00%
-+ 0.00%

Demand far exceeds supply! Bank France and Pakistan: The semiconductor industry is collectively bullish on 2026, and electricity and ASIC risks are overestimated

智通財經·12/15/2025 01:17:01
語音播報

The Zhitong Finance App learned that the BNP Paribas Research Department pointed out that many semiconductor companies are still positive about the supply and demand situation moving towards 2026.

The global financial firm hosted a Silicon Valley bus tour last week with 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 (STX.US), Maywell (MRVL.US), and Western Digital (WDC.US) Guan had a meeting.

“In the field of artificial intelligence computing, all companies we have come into contact with are very optimistic about the market performance in 2026 and the subsequent rise in production capacity. Demand far exceeds supply,” BNP Paribas analysts David O'Connor and Carl Ackman said in an investor report. Investors' main concerns are power supply, ASIC competition, and the ability of suppliers to finance.”

Both AMD and Nvidia said electricity supplies across the US are becoming tight. However, both companies believe that the US government is taking steps to ease electricity restrictions; this is more of a short-term issue.

“All parties agree that electricity is the main bottleneck in the artificial intelligence arms race,” O'Connor said. Although Nvidia admits that electricity consumption is tight, they don't think there are energy barriers, and construction is expected to accelerate. Given that the US is most limited in terms of power capacity, foreign investment is likely to increase. “The visibility of the data center ecosystem has been extended to multiple quarters to accommodate hyperscale enterprises' multi-year roadmap and Nvidia's 9 to 12 month delivery cycle, which has improved supply chain efficiency and price dynamics.”

Another potential concern discussed is the introduction of custom chips, such as tensor processing units (TPUs) built by Google (GOOGL.US) using application-specific integrated circuit (ASIC) technology.

“Following the recent TPU announcement, ASIC competition became the focus of attention,” O'Connor pointed out. Companies involved in the computing field emphasized that TPU is optimized for specific cloud service providers/workloads (such as Anthropic, GCP) and is not for all cloud services (unlike GPUs), so their market share growth should not be extrapolated to vendors other than existing TPU users.”