TL;DR
High Bandwidth Memory has become the part driving the 2026 memory squeeze, according to Thorsten Meyer AI’s late-June series on the market. Suppliers are shifting fab capacity toward HBM for AI accelerators, tightening supply for DDR5 and GDDR7 while HBM demand remains sold out through 2026.
High Bandwidth Memory has become the component driving the 2026 memory crunch, according to a late-June report from Thorsten Meyer AI, as suppliers divert wafer capacity from standard DRAM toward the stacked memory used in AI accelerators.
The report says HBM, once a niche part, is now shaping the price and availability of much of the memory market. The part is used beside high-end AI GPUs, including Nvidia H100, H200 and B200 systems, and is expected to be central to Nvidia’s coming Rubin platform.
HBM stacks eight to 16 DRAM dies vertically and links them through thousands of through-silicon vias. That design gives AI chips far more bandwidth than conventional graphics memory, but the report says each HBM bit can consume roughly three to four times the wafer area of a DDR5 bit.
Thorsten Meyer AI, citing market sources including TrendForce, DigiTimes and Reuters, said HBM demand is sold out through 2026. It also reported that HBM3E pricing rose about 20% for 2026 and that HBM revenue could grow from roughly $35 billion to about $100 billion by 2028.
HBM ate the fab
The thing the factories make instead of your RAM is a tower of stacked memory bolted to every AI chip. In three years it went from niche part to the component that sets the price of nearly all the world’s memory — and now a chunk of its GPUs.
A tower, not a sheet
HBM stacks DRAM dies vertically, links them with thousands of through-silicon vias, and sits beside the GPU to deliver 5–10× the bandwidth of normal graphics memory. AI is bandwidth-bound — without it, the world’s most expensive silicon sits starved for data. But stacking is inefficient: one HBM bit eats 3–4× the wafer area of DDR5, and one defect can ruin a whole tower.
≈ 8 HBM stacks wrap every AI GPUThis isn’t artificial scarcity — AI really is bandwidth-bound, HBM really is the fix, and it really does eat 3–4× its weight in fab capacity. The discomfort is structural: one component, coupled to one customer’s demand, now sets the price of nearly all memory and a slice of GPUs. The market is now $35B → ~$100B by 2028, ~41% of all DRAM revenue (was 8% in 2023), and sold out through 2026. The one hope: with all three suppliers finally racing on HBM4, competition can add supply. The matching risk: if AI demand corrects, HBM is where it breaks first. Next: DDR5 now, DDR6 soon.
AI Memory Demand Hits Consumers
The shift matters because memory fabs are finite. When manufacturers allocate more wafers to HBM stacks, less capacity remains for ordinary DDR5 RAM and graphics-card memory. The report frames the shortage as a structural supply problem rather than a simple pricing cycle.
The pressure is also reaching consumer GPUs. Thorsten Meyer AI said suppliers’ focus on HBM has contributed to a shortage of GDDR7, the memory used in new graphics cards, and cited reports that Nvidia reduced RTX 50-series production by a third or more in the first half of 2026. Nvidia has not confirmed that figure in the provided source material.
High Bandwidth Memory HBM modules
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
How HBM Took Fab Priority
HBM is valuable because AI workloads are often limited by memory bandwidth rather than raw compute. A fast accelerator can sit underused if memory cannot feed it data quickly enough. HBM addresses that bottleneck by placing stacked memory close to the GPU and delivering about five to 10 times the bandwidth of ordinary graphics memory, according to the report.
The market is led by SK Hynix, with the report estimating its share at 50% to 62% and saying about 90% of its HBM goes to Nvidia. Samsung is described as holding about 28% to 40%, while Micron is estimated at 5% to 10% and sold out for 2026.
“The thing the factories make instead of your RAM is a tower of stacked memory bolted to every AI chip.”
— Thorsten Meyer AI
suevery Prebuilt Gaming PC White
High-Performance Processor: Experience the unparalleled power of the 13th I9 13900HX(Beats I7 14700F) 24Cores 32Threads Up to 5.4G…
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Supply Relief Remains Unproven
Several details remain unconfirmed or fast-moving. Per-stack HBM prices in the report are described as estimates, including about $200 for HBM3, $300 for HBM3E and an estimated $500 for HBM4.
It is also unclear how quickly Samsung and Micron can increase HBM4 output, how much added supply will reach non-Nvidia customers, and whether AI demand will keep expanding at the pace assumed by current capacity plans.
NVIDIA RTX PRO 4000 Blackwell Graphics Card – 24GB GDDR7 ECC Memory, PCIe 5.0 x16, 4X DisplayPort 2.1b, Single Slot Full Height AI Workstation GPU, Retail Packaging
Professional GPU with Blackwell Architecture
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
HBM4 Output Becomes Test
The next marker is HBM4 production. Thorsten Meyer AI said all three major suppliers had qualified for HBM4 by June 2026, shifting attention to yield, volume and customer allocation.
Readers should watch for HBM4 shipment volumes, DDR5 price changes, GDDR7 availability and any update from Nvidia, SK Hynix, Samsung or Micron on capacity plans. A slowdown in AI accelerator orders could ease pressure, but it could also expose the HBM market first.
ASUS ROG Strix GeForce RTX 4090 OC Edition Gaming Graphics Card (PCIe 4.0, 24GB GDDR6X, HDMI 2.1a, DisplayPort 1.4a), 3 Year Warranty
NVIDIA Ada Lovelace Streaming Multiprocessors: Up to 2x performance and power efficiency
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is HBM?
High Bandwidth Memory is stacked DRAM placed close to an AI processor or GPU. It provides much higher memory bandwidth than standard DDR5 or GDDR memory.
Why does HBM affect ordinary RAM prices?
HBM uses more wafer area per bit than DDR5 and is more profitable. When fabs prioritize AI memory, less capacity is available for consumer and server RAM.
Is the shortage confirmed to last through 2026?
The report says HBM is sold out through 2026, citing market sources. The duration of pressure on DDR5 and GDDR7 is still developing.
Which companies dominate HBM supply?
The report identifies SK Hynix as the leader, followed by Samsung and Micron. Their HBM4 ramp will shape supply through late 2026 and 2027.
Source: Thorsten Meyer AI
