Global Technology: Innovating the Next-Generation Memory

Research evidence excerpt

Global Technology: Innovating the Next-Generation Memory

Global Insight

Industry View Key TakeawaysMGreater China Technology Semiconductors | • AI compute power continues to improve rapidly, but memory capacity

Asia Pacific remains constrained, bandwidth is limited, and costs are elevated.

Attractive • This report addresses this simple but critical question: how can the industry

overcome the memory bottleneck so AI can continue to scale?

• The challenge extends well beyond HBM. Innovation will be required across

What’s Changed design, process, packaging, peripherals, integration and materials.

From To • Enablers of the entire memory ecosystem, across interface, connectivity,

Montage Technology Co Ltd (6809.HK) packaging, system design, could see re-rating opportunities — not just

Price Target HK$310.00 HK$432.00 traditional memory vendors.

Montage Technology Co Ltd (688008.SS)

Price Target Rmb274.00 Rmb377.00 AI is entering a new phase, with the focus shifting from compute scaling to data

efficiency: Generative AI has driven strong HBM demand, and agentic AI is set to

further increase memory requirements (Rise of the AI Agent – Global Implications).

If GPUs determine how fast AI can run, memory determines how far it can scale.

Memory is becoming a key monetization pillar, we think, with agentic AI potentially

driving 26–77% of incremental DRAM demand by 2030. Based on our current

forecasts, cloud memory spending could reach US$418bn by 2030, implying 8%

CAGR from 2026.

The memory wall is emerging as a key bottleneck, driven by constraints in

capacity, bandwidth and cost: While AI models continue to scale, memory supply

is struggling to keep pace, creating structural imbalances. We estimate memory