AI Trade Loses Steam as Infrastructure Boom Faces Reality Check

The AI trade spans semiconductors, memory chips and the data-center supply chain and has been fueled by rising capital spending among cloud providers. Investors are now assessing whether spending on GPUs, high-bandwidth memory and data centers can generate sufficient returns. If hyperscalers such as Microsoft and Amazon slow construction, chip demand, valuations and capital flows could all come under pressure.

Samsung Electronics reported second-quarter 2026 revenue of $113 billion, below the market estimate of $113.9 billion. Operating profit reached $59.1 billion, up 19-fold year on year, but its shares still fell nearly 7% on July 7. Micron and SanDisk each dropped about 6% in premarket trading the same day, while SK Hynix retreated 25% from its record high. China’s Zhipu AI is also considering developing chips in-house to support growing demand for its open-source GLM models.