KB Securities: Memory
"As of August, only 60% of big techs’ memory and AI substrate needs are being met. Big techs seem to understand that overinvestment in AI carries less risk than underinvestment. Unmet demand in 2026 should spill over into 2027 and unfulfilled demand in 2027 should carry into 2028, resulting in a multi-year domino effect that is likely to last for the next three years. Since a memory fab requires over three years to be built and become fully operational, we expect the shortage to begin easing after 2029."
$MU $DRAM $EWY
SK HYNIX AND SANDISK HAVE UNVEILED THE INDUSTRY'S FIRST STANDARD SPECIFICATIONS FOR HIGH-BANDWIDTH FLASH (HBF), A NEXT-GENERATION STORAGE TECHNOLOGY, ABOUT SIX MONTHS AFTER THE TWO COMPANIES LAUNCHED A CONSORTIUM THAT ALSO INCLUDES GOOGLE AND TENSTORRENT.
2027 full-year DRAM (including HBM) capacity from the three major makers has already been fully pre-allocated/sold out. NAND Flash capacity is expected to be largely booked by the end of August 2026.
Samsung, SK Hynix $SKHY and Micron $MU
2027 memory capacity is already spoken for.
Industry sources say the three major DRAM makers have finished allocating their full-year 2027 output. Both standard DRAM and HBM are sold out.
Buyers are staying quiet on purpose. They do not want more players rushing in and cutting into their own share.
This is the real shift. AI demand has flipped the market into a lasting seller’s position. Long-term agreements of three to five years are locking capacity far ahead of normal cycles.
Hyperscalers and big brands fought hard for 2026 supply, sometimes almost begging for volume.
To avoid repeating that scramble, they moved early on 2027. The result is that most of next year’s output is already reserved.
Allocation is not limited to the biggest LTA customers. Mid-sized buyers who received 2026 volume are also getting notified of their 2027 quotas.
Deposits are required up front. Even so, many companies are getting less than they asked for or nothing at all.
Typical allocations run only 60-70% of the original request. Priority goes to cloud and AI customers.
That leaves traditional phone and PC makers with clearly smaller DRAM shares than in 2026.
Adata chairman Chen Li-bai has confirmed the picture.
The three big producers’ 2027 capacity is gone.
HBM and AI-server related demand will take nearly 70% of total DRAM output.
PC and smartphone supply will be tight.
Strong enterprise storage demand is also tightening NAND and hard-drive availability.
$MU $SNDK $SKHY
Samsung, Micron, and SanDisk have already pre-sold their full-year capacity.
Kioxia and SK Hynix are expected to finish allocation by the end of August. Buyers still have some room to negotiate on NAND, but the window is closing fast.
SK Group chairman Chey Tae-won has said AI semiconductor demand in 2027 could jump 60-100% from 2026 levels.
Overall storage demand may rise 50-60%. The supply gap looks set to widen.
Many industry voices now call 2027 the most severe shortage year on record, with the shortfall from 2026 simply carrying forward and growing larger.
Prices are expected to stay high, though the wild multi-fold spikes seen in 2026 may moderate. Final pricing will be settled closer to actual shipment dates.
Makers still control the allocation lever and continue to push higher quotes. High prices are becoming the new normal rather than a temporary spike.
Companies that have not locked volume yet face a harder reality.
2027 could leave them with little or no supply.
One practical step for smaller or mid-sized buyers is to keep pressing every available channel now, including secondary suppliers and any remaining flexible capacity, before the final cutoffs close.
#semiconductors #memory #DRAM #NAND
Looks great for $MU
Citing sources , all memory he entire production capacity for 2027 has been allocated.
FOR DRAM makers - confirmed that the three major manufacturers' 2027 production capacity has already been sold out.
tomorrows Digitimes, August 4th.
And its It has been revealed that Samsung Electronics, Micron, and SanDisk have already sold out their entire year's NAND production capacity
https://t.co/pp46rKEBCH
Saw some memory FUD resurface on Nvidia $NVDA reducing memory on its shipped Rubin Ultra in 2027 than originally planned.
One thing that needs to be reiterated is that $NVDA is not reducing HBM. They are reducing SOCAMM capacity for its Rubin Ultra in 2027, mainly because there simply isn't enough. Second reason is to lower costs per rack.
Is this bearish? No, it mostly underlines the unprecedented shortage in DRAM, which won't get better in 2027. NVIDIA wants to guarantee that the racks can be shipped, and it does so by reducing SOCAMM capacity.
SK hynix $SKHY, Micron $MU and Samsung $005930.KS are at the forefront to profit from the DRAM shortage, which will only become worse in 2027.
I ran the numbers on the memory oligopoly.
Samsung $SSNLF + SK Hynix $SKHY + Micron $MU control 90% of DRAM.
Add Kioxia + Sandisk $SNDK and you basically own the high end too.
Combined market cap $3.3T.
Next year profits ~$550-600B.
18% earnings yield. S&P is under 4%.
Shortage runs into 2028.
Still feels mispriced.