SK Hynix may cut Nvidia HBM4 shipments 20%-30% as Rubin ramps slowly

aibusinesstech 1 posts · 1 accounts

A detailed Korea-sourced report said SK Hynix may cut planned HBM4 shipments to Nvidia by 20%-30% from its original plan as Rubin ramps more slowly, redirecting volume to HBM3E and server DRAM.

From the sources (1 posts)

@jukan05

SK Hynix to Cut HBM4 Volumes This Year... Expanding HBM3E and Other Products Instead SK Hynix is reportedly pursuing a plan to reduce its 6th-generation High Bandwidth Memory (HBM) shipments to NVIDIA by approximately 20–30% compared to it

Market impact

55 medium

NVDA SOXX MU SMH

Why it matters: A forced HBM4 order cut of 20-30% signals slower-than-planned Nvidia Rubin ramp, a real demand read-through into the AI-accelerator/HBM supply chain that pressures NVDA and semis, though volume redirects to HBM3E/server DRAM soften the memory-maker hit.

At publish time: Headline states a specific supplier order cut of 20-30% tied to a slow Rubin ramp — a forced/negative supply-chain signal that is hard to fake and reads directly on Nvidia's near-term product cadence.

model confidence: medium

What happened next

1–2 weeks later

fizzled moved: flat

Follow-ups show Rubin ramping, not slowing: SOCAMM2 mass production, Nanya added, copper-foil demand up, Groq deal. HBM4-cut/slow-ramp bear thesis for NVDA didn't materialize.

based on follow-up coverage

Follow-up coverage

No follow-up coverage matched this story.

TSMC sees 2026 sales up 32% and capex at high end on AI demand IATA warns Europe could face late-May flight cancellations as jet fuel shortages deepen