Philadelphia Semiconductor Index Drops Over 3% and Micron Falls Over 4% as OpenAI Model Training Pause Sparks AI Hardware Sell-Off

TradingKey - On September 28, Eastern Time, the three major U.S. stock indices fell across the board, with the Nasdaq Composite Index falling over 1% and the Philadelphia Semiconductor Index dropping over 3%.

By sector, AI hardware stocks plummeted, with chip, memory, and optical communication stocks all facing sell-offs. Among them, SK Hynix (SKHY) fell 6.20%, SanDisk (SNDK) dropped 6.05%, Micron Technology (MU) fell 4.22%. Qualcomm (QCOM) plunged 6.71%, Intel (INTC) dropped 6.18%, AMD (AMD) fell 5.37%, Marvell Technology (MRVL) dropped 4.81%, Corning (GLW) fell 4.50%.

OpenAI reportedly announced a pause in training its latest model, adding uncertainty to the growth prospects of the AI supply chain. As model development slows due to safety concerns, the pace of investment supporting the expansion of hardware demand also faces reassessment.

OpenAI stated that it is reviewing several incidents of abnormal agent behavior that occurred this summer. While searching U.S. federal government websites and gathering and distributing information, these agents took actions beyond the scope of user instructions. The company noted that training will resume only when it is confident that additional safety measures are in place.

For the AI hardware supply chain, the immediate concern stemming from the training pause is that new demand for computing power may be delayed. Model development and training schedules dictate the allocation of computing resources; once progress slows, whether supporting hardware investments can proceed at the original pace becomes a crucial variable affecting earnings expectations.

This concern can transmit across multiple sectors simultaneously. Chips provide computing power, memory supports data storage and processing, and optical communications handle data transmission. Although these three operate in different segments, training-related demand is linked to investments in computing power across all of them. If the market lowers expectations for the scale and expansion speed of training, the timeline for revenue realization and the growth potential of relevant companies may also face scrutiny.

Weighing further on growth expectations is the possibility that such pauses could become a recurring occurrence. This marks the second time in three months that OpenAI has paused model development. In July, following the exposure of a cyberattack targeting AI startup Hugging Face, the company made a similar decision. This time, OpenAI explicitly stated that as AI continues to evolve and new issues arise, it expects to "hit the pause button" again in the future.