J.P. Morgan is effectively rejecting the memory peak thesis by forecasting DRAM + NAND revenue surging to ~$1.8T in 2028. The note suggest that the market still grows another 27% in 2028 as HBM-driven mix, constrained supply and higher pricing reshape memory margins. The market is still pricing $MU, $SKHY and $SNDK for a cycle peak that J.P. Morgan now sees extending well into 2028.

J.P. Morgan says the market is ignoring 12 years of $AVGO execution at $GOOGL and would be “aggressive buyers at current levels.” The 5-year Google deal still locks Broadcom into TPU v8-v11 with rising revenue commitments while $META, $AAPL, OpenAI, Anthropic, ByteDance and SoftBank programs remain on track. J.P. Morgan sees over $56B of AI revenue in FY26 and another 2x to 2.5x growth in FY27 with Tomahawk 6 already sold out through next year and Samsung supply commitments implying $1T of cumulative AI revenue over 5 years.

J.P. Morgan expects a continued expansion of the global memory market in 2027. Driven by growth in both DRAM and NAND markets, with higher supply and ASPs forecasted across the board. Note - I would ignore 2028 numbers. It's too far away to model out demand, supply, and resultant ASPs imo. $DRAM