TrendForce: DRAM Supply Shortage to Persist Through 2027… NVIDIA Considering Downgrading Rubin Ultra's HBM Configuration - According to TrendForce's latest memory industry research, the DRAM supply shortage is expected to continue through 2027, and with uncertainties remaining in the HBM4e qualification timeline, NVIDIA has been re-evaluating Rubin Ultra's HBM configuration since Q3 2026. - While the original plan was to adopt a 12-high (12hi) HBM4e configuration, NVIDIA is now evaluating multiple designs in parallel — including HBM4e 8hi, HBM4 12hi, and HBM4 8hi — and the final specification has not yet been confirmed. - Beyond NVIDIA, some CSPs are also reportedly considering reducing the HBM capacity on their next-generation in-house ASICs. - NVIDIA maintained 12hi HBM4e as the baseline design for Rubin Ultra from 2025 through the first half of 2026. However, starting in early Q3 2026, it began reviewing lower-spec alternatives. This shift stems from two major supply-side constraints: first, the overall DRAM shortage expected in 2027 limits the wafer capacity that memory makers can allocate to HBM production; and second, uncertainty remains around the qualification timeline for 12hi HBM4e and the pace of mass-production yield improvement. - TrendForce explains that NVIDIA's top priority for the Rubin Ultra generation is increasing I/O speed, with expanding GPU shipments as a secondary priority. If NVIDIA ultimately decides to lower the HBM specification, it is expected to do so by reducing the number of DRAM stack layers. - Ultimately, whether HBM4e can complete qualification and enter mass production on schedule will determine whether Rubin Ultra's I/O speed can be raised from the previous generation Rubin's 8–11.7 Gbps to 14–16 Gbps, or whether it will remain at the 11–12 Gbps level through optimized HBM4 designs. - Meanwhile, within the same product generation, the number of DRAM stack layers determines the trade-off between HBM capacity per GPU and the number of GPUs that can be shipped. - TrendForce also believes the final configuration will depend on memory makers' wafer allocation decisions. On the supply-demand front, HBM bit shipments in 2027 are projected to grow 50–60% year-over-year, but even this is expected to fall short of keeping pace with demand growth. - With these supply constraints persisting, HBM suppliers are expected to retain pricing power throughout 2027. The industry already widely anticipates a substantial increase in HBM prices. As a result, AI chip makers will face the dual burden of limited HBM supply and rising procurement costs — further strengthening the incentive to adopt configurations with lower HBM capacity. $DRAM $SKHY $MU

TrendForce: DRAM Supply Shortage to Persist Through 2027… NVIDIA Considering Downgrading Rubin Ultra's HBM Configuration - According to TrendForce's latest memory industry research, the DRAM supply shortage is expected to continue through 2027, and with uncertainties remaining in the HBM4e qualification timeline, NVIDIA has been re-evaluating Rubin Ultra's HBM configuration since Q3 2026. - While the original plan was to adopt a 12-high (12hi) HBM4e configuration, NVIDIA is now evaluating multiple designs in parallel — including HBM4e 8hi, HBM4 12hi, and HBM4 8hi — and the final specification has not yet been confirmed. - Beyond NVIDIA, some CSPs are also reportedly considering reducing the HBM capacity on their next-generation in-house ASICs. - NVIDIA maintained 12hi HBM4e as the baseline design for Rubin Ultra from 2025 through the first half of 2026. However, starting in early Q3 2026, it began reviewing lower-spec alternatives. This shift stems from two major supply-side constraints: first, the overall DRAM shortage expected in 2027 limits the wafer capacity that memory makers can allocate to HBM production; and second, uncertainty remains around the qualification timeline for 12hi HBM4e and the pace of mass-production yield improvement. - TrendForce explains that NVIDIA's top priority for the Rubin Ultra generation is increasing I/O speed, with expanding GPU shipments as a secondary priority. If NVIDIA ultimately decides to lower the HBM specification, it is expected to do so by reducing the number of DRAM stack layers. - Ultimately, whether HBM4e can complete qualification and enter mass production on schedule will determine whether Rubin Ultra's I/O speed can be raised from the previous generation Rubin's 8–11.7 Gbps to 14–16 Gbps, or whether it will remain at the 11–12 Gbps level through optimized HBM4 designs. - Meanwhile, within the same product generation, the number of DRAM stack layers determines the trade-off between HBM capacity per GPU and the number of GPUs that can be shipped. - TrendForce also believes the final configuration will depend on memory makers' wafer allocation decisions. On the supply-demand front, HBM bit shipments in 2027 are projected to grow 50–60% year-over-year, but even this is expected to fall short of keeping pace with demand growth. - With these supply constraints persisting, HBM suppliers are expected to retain pricing power throughout 2027. The industry already widely anticipates a substantial increase in HBM prices. As a result, AI chip makers will face the dual burden of limited HBM supply and rising procurement costs — further strengthening the incentive to adopt configurations with lower HBM capacity. $DRAM $SKHY $MU

Source:https://x.com/jukan05/status/2084536701532451009