NEW YORK, February 6, 2026, 16:33 (EST)
- Micron stock slid in premarket trading after Semianalysis cut its forecast for Micron’s Nvidia HBM4 market share to zero.
- According to research firm estimates, SK Hynix is expected to account for about 70% of Nvidia’s HBM4 supply, with Samsung picking up the remaining 30%.
- Nvidia is pressuring Samsung for early HBM4 deliveries, DigiTimes reported, as memory shortages loom.
Shares of Micron Technology fell 0.5% ahead of Friday’s open after Semianalysis downgraded its forecast, now expecting Micron to miss out entirely on supplying HBM4 memory for Nvidia’s next products.
This decision matters. High-bandwidth memory, or HBM, is the stacked DRAM sitting alongside AI chips, feeding them data at breakneck speed. Getting selected by Nvidia can secure a supplier years of robust orders and pricing leverage. Miss that cut, and suppliers are left hustling for the next opportunity.
Investors are zeroing in on the surge in AI infrastructure outlays, as memory shortages rear up again as a serious bottleneck. Big U.S. tech names look set to shell out more than $630 billion on datacenters and AI chips this year. “The magnitude of the spend is materially greater than consensus expected,” MoffettNathanson analysts wrote. https://www.reuters.com/business/retail-co…
Micron popped almost 4% before the bell after Amazon set out a $200 billion capital spending plan for 2026, signaling demand for cloud infrastructure is holding up. That early gain evaporated, though, as the Semianalysis note circulated and shares gave up ground.
According to Semianalysis, there’s no sign Nvidia is placing HBM4 orders with Micron. Instead, the firm expects Nvidia to rely on SK Hynix and Samsung for its HBM4 supply. SK Hynix is projected to grab roughly 70% market share, while Samsung should account for 30%. Semianalysis attributed an earlier downgrade to “poor speed performance from their use of an internal base die.”
Nvidia has reached out to Samsung Electronics ahead of schedule for sixth-generation high-bandwidth memory—HBM4—according to DigiTimes, as the industry’s pivot toward HBM ramps up and supply constraints begin to take hold.
Nvidia is set to rely on SK Hynix and Samsung for HBM4 in its “Vera Rubin” series, passing over Micron altogether, TechPowerUp reports. https://www.techpowerup.com/346044/nvidia-…
If the projections hold, the three leading HBM suppliers could find themselves locked even closer together. SK Hynix keeps its lead. Samsung edges forward. Micron, on the other hand, is left chasing new HBM contracts or banking on another shot with Nvidia down the line.
HBM isn’t conventional commodity memory. It’s layered in stacks, with each stack tied to a logic “base die” that manages the stack and hooks it up to the processor package. Performance limits keep climbing, as AI chips demand ever more bandwidth.
Supplier allocations shift constantly—qualification cycles, yield surprises, timing slips. Nvidia won’t comment on parts sourcing for unreleased products. “The market just dislikes the substantial amount of money that keeps getting put into capex for these growth rates,” said Dave Wagner, portfolio manager at Aptus Capital Advisors. Mood around AI capex, and its suppliers, can sour fast. https://www.reuters.com/business/retail-co…
Micron holders now face questions about whether the company can carve out a substantial role in HBM4, with Nvidia prepping for the next AI hardware cycle. One supplier comment was all it took on Friday to overshadow what had looked like a robust demand backdrop.
