Tiger 123
09-09

The biggest change overnight is a sharper split between AI fundamentals and the broader market. The S&P 500 fell 0.58%, Dow 1.18% and Nasdaq 0.32% as Brent approached US$100 and Treasury yields stayed near 4.8%. Yet semiconductor names tied to custom AI infrastructure held up better, helped by a major new Amazon–Qualcomm agreement.

Micron is already shipping HBM4 in volume for Nvidia Vera Rubin, delivering >2.8 TB/s bandwidth and >20% greater power efficiency than HBM3E. SK hynix meanwhile approved about ₩54T of new fab investment specifically to address growing AI-memory demand. HBM remains the strongest fundamental layer, but after recent sharp memory rallies coupled with imminent interest rate increases, better entry points key rather than Supercycle over. 

HBM, SSDs and Permits — Which Memory Signal Comes First?
Memory stocks rebounded Friday after Thursday’s AI-policy selloff: Micron gained 4.75%, SanDisk 3.57%, and SK Hynix 2.78%. Micron must show HBM contracts holding on price and volume; SanDisk faces cloud-inventory and SSD-pricing tests; Hynix needs U.S. permits and a clear expansion timeline. Bulls say supply remains tight; bears cite approval delays, cloud budgets, and cyclical pricing risks. Which offers the first hard proof: HBM contracts, SSD orders, or permits?
Disclaimer: Investing carries risk. This is not financial advice. The above content should not be regarded as an offer, recommendation, or solicitation on acquiring or disposing of any financial products, any associated discussions, comments, or posts by author or other users should not be considered as such either. It is solely for general information purpose only, which does not consider your own investment objectives, financial situations or needs. TTM assumes no responsibility or warranty for the accuracy and completeness of the information, investors should do their own research and may seek professional advice before investing.

Comments

  • IrmaBurke
    09-09
    IrmaBurke
    That HBM4 jump matters more for inference than people think — Rubin class systems get more headroom on memory-bound loads, but the real bottleneck may shift to packaging and qualified supply.
Leave a comment
1