Shares of Nvidia’s (NASDAQ: NVDA) major Asian suppliers tumbled on Thursday, following a sharp overnight decline in the U.S. tech sector. Growing concerns over a potential data center oversupply and shifting AI infrastructure demand sent shockwaves through the supply chain, weighing on semiconductor and chip-related stocks.
Leading the losses was Taiwan Semiconductor Manufacturing Company (TSMC) (TW:2330) (NYSE:TSM), Nvidia’s top supplier, which dropped about 2% in Taipei trading.
Foxconn (TW:2317) (SS:601138) saw a steeper decline of over 3%, while South Korea’s SK Hynix (KS:000660)—a key supplier of AI memory chips—fell by 3%. Japan’s Advantest Corp. (TYO:6857), which provides chip testing equipment, posted the sharpest drop, tumbling more than 7%.
The selloff followed a 6% overnight plunge in Nvidia’s stock, triggered by a note from investment firm TD Cowen, which reported that Microsoft (NASDAQ: MSFT) had canceled over 2 gigawatts (GW) of data center leases in the U.S. and Europe over the past six months.
Microsoft’s AI Strategy Shift Raises Red Flags
TD Cowen’s research highlighted Microsoft’s decision to scale back AI training workloads for OpenAI, leading to canceled data center leases. While Microsoft remains a dominant force in AI infrastructure, this move raised speculation about a potential cooling in hyperscale data center demand, particularly for AI model training.
Key takeaways from the report include:
- Over 2GW of canceled data center leases in the past six months, including terminations of existing contracts.
- Microsoft scaling back AI training expansion for OpenAI, shifting its focus to inference workloads.
- Concerns over a possible oversupply in data center infrastructure, a sector that has seen rapid growth.
Despite Microsoft’s pullback, TD Cowen noted that Google (NASDAQ: GOOGL) and Meta (NASDAQ: META) were stepping in to backfill some of the capacity, indicating that AI infrastructure demand remains dynamic.
Data Center Growth Faces AI Demand Uncertainty
The TD Cowen report comes amid broader warnings about a data center bubble, with Alibaba (NYSE: BABA) Chair Joe Tsai recently cautioning that many projects were being launched without clear demand drivers.
A key industry shift is also underway. The AI sector is moving from training large-scale models (a computationally intense process) to AI inference, where models generate and process data with lower computing requirements. This transition could reduce demand for high-performance GPUs, a cornerstone of Nvidia’s dominance.
However, analysts argue that strong consumer adoption of AI-powered applications could sustain high data center demand, though the pace of expansion remains uncertain.
As the AI industry recalibrates, investors are closely watching how hyperscalers allocate their infrastructure spending, with Nvidia’s supplier network particularly exposed to any shifts in strategy.