Broadcom Warns of AI Chip Shortage, Citing TSMC Capacity and Supply Chain Issues

TLDR

• Broadcom indicates TSMC is reaching production capacity limits because of rising AI chip demand
• The bottleneck will likely continue through 2026, with capacity expansions scheduled for 2027
• Supply shortages are affecting not just chips but also lasers and PCBs
• PCB delivery times have extended from 6 weeks to 6 months for optical transceivers
• Businesses are entering 3–5 year supply contracts to secure capacity

---

(SeaPRwire) –   Broadcom has highlighted that supply chain restrictions are intensifying throughout the technology industry, identifying its manufacturing partner TSMC as a primary bottleneck. The caution was issued on Tuesday by Natarajan Ramachandran, director of product marketing in Broadcom’s Physical Layer Products division.

Broadcom Inc., AVGO

Ramachandran informed journalists that TSMC is “hitting production capacity limits.” He noted that he would have characterized TSMC’s capacity as “unlimited” only a few years prior — that situation has changed.

TSMC stands as the world’s leading manufacturer of advanced AI chips. Its key clients comprise Broadcom, Nvidia, and Apple. The Taiwan-based firm confirmed capacity constraints in January and stated it was striving to narrow the supply-demand gap.

Per Ramachandran, TSMC intends to boost capacity through 2027. However, that expansion hasn’t come quickly enough. “That has turned into a bottleneck, or that has essentially constrained the supply chain in 2026,” he stated.

The shortfalls aren’t confined to semiconductors. Ramachandran indicated that supply limitations are expanding into other areas of the technology supply chain.

He highlighted a shortage in the laser sector, observing that even with several vendors in the marketplace, capacity stays constrained. Printed circuit boards have likewise become an unforeseen stress point.

PCB Lead Times Surge

PCBs utilized in optical transceivers are experiencing lead times extending from approximately six weeks up to six months. Both Taiwanese and Chinese PCB manufacturers are encountering capacity constraints, according to Ramachandran, though he declined to identify particular firms.

This comes after previous reports of limitations affecting chip producers Intel and AMD. Intel’s server product pricing increased by roughly 10%, while AMD faced extended delivery periods. AMD stated at that time it stayed confident in satisfying demand via its supplier contracts, including its partnership with TSMC.

Long-Term Supply Contracts on the Rise

Reacting to these pressures, numerous customers are currently signing long-term supply deals spanning three to four years. Samsung verified last week it too is moving toward lengthier agreements of three to five years with key clients.

These contracts signify a wider transformation. Clients desire supply assurance, while vendors seek protection against demand fluctuations.

Ramachandran expressed he isn’t particularly concerned about the long-term perspective. He anticipates that new market participants and capacity growth will alleviate restrictions eventually.

This announcement follows Broadcom’s declaration of a collaboration with OpenAI to develop 10 gigawatts of custom AI accelerators. OpenAI will handle chip and system design while Broadcom assists with development and implementation.

TSMC declined to reply to a request for comment by publication time.