Nasdaq Rises Over 1% to Recover Previous Two Days of Losses. AI Hardware Stocks Resume Dominance in Market Rally as JPMorgan Points to Three Tailwinds Supporting AI Infrastructure Demand

TradingKey - On July 6, Eastern Time, the three major U.S. stock indexes diverged as AI hardware stocks rebounded. JPMorgan Chase stated that triple tailwinds are supporting demand for AI infrastructure. The Nasdaq Composite recovered its losses from the previous two days, the Philadelphia Semiconductor Index rose over 3%, and the Dow turned down after hitting a record high.

As of press time, the Dow Jones Industrial Average fell 0.05% to 52,874.57 points; the Nasdaq Composite Index rose 1.35% to 26,182.40 points; and the S&P 500 Index rose 0.71% to 7,536.62 points.

[Source: FutuBull]

At the sector level, chip stocks and memory stocks led the gains; AMD ( AMD) rose 8.62%, breaking through the $560 mark; Qualcomm ( QCOM) rose 6.65%; TSMC ( TSM) rose 5.38%; Western Digital ( WDC) rose 7.3%; Seagate ( STX) rose 4.76%; Micron ( MU) and SanDisk ( SNDK) both rose over 3%.

In its latest research report, JPMorgan Chase stated that LLM call volumes are still expanding rapidly, token spending is re-accelerating, GPU rental prices in the non-hyperscale cloud provider market continue to rise, and DRAM spot prices remain strong. The institution believes that the decline in model prices has not dampened AI demand; instead, it is driving further expansion of inference demand through lower barriers to entry and broader use cases.

According to OpenRouter data tracked by the bank, the total volume of LLM tokens in June grew by 70% month-on-month, significantly higher than the 33% in May and 5% in April; the year-on-year growth rate reached 20 times, higher than the 12 times in May and 15 times in April. Token spending also recovered in tandem, growing 70% month-on-month and 16 times year-on-year in June, ending the growth slowdown of the previous two months.

Based on this, JPMorgan Chase judges that under the premise of samples and simplified assumptions, even with relatively conservative token pricing, the unit economics of model providers still show an improving trend. This does not mean that all model companies have achieved stable profitability, but at least it indicates that the market demand side has not yet shown signs of "price wars crushing business models."

In addition, GPU rental prices in the non-hyperscale cloud provider market continue to rise. Bloomberg data tracked by JPMorgan Chase indicates that GPU rental prices in the non-hyperscale cloud provider market continued to rise in June. The average rental price for the A100 was $1.63 per GPU hour, up 6.3% month-on-month, marking the fifth consecutive month of increases.

Absolute market demand for the latest generation of GPUs remains robust. However, with supply gradually increasing, alternative products proliferating, and customers choosing equipment more based on unit computing costs, the scarcity premium of the B200 is becoming more rational. For Nvidia, this does not necessarily mean weaker demand; rather, it reflects that the AI computing market is gradually moving from "demanding only the latest product" to more refined performance and cost optimization.

Lastly, DRAM prices are still surging, with the spot price of DDR5 16Gb rising to $43.14 in June, up 10% month-on-month, up 740% from $5.13 in the same period last year, and marking the third consecutive month of increases.

For DRAM and HBM suppliers such as Micron, SK Hynix, and Samsung, the strength in DRAM prices is beneficial for improving profitability and product mix; for storage-related companies such as Western Digital, SanDisk, and Seagate, NAND prices remain high, but future profit elasticity may be weaker than that of DRAM.