Semiconductor Memory
High Bandwidth Memory
High Bandwidth Memory is an advanced memory technology, first standardized by JEDEC in 2013, that provides the high-speed data access modern AI accelerators require. It's now the dominant memory technology for AI chips, with the market in an AI-driven supercycle BofA estimated at $54.6 billion in 2026, up 58% year-over-year.
Additional Intelligence Signals
Patent citation lineage, earnings-call mentions, and federal contract disclosures — automatically collected, not yet visible anywhere else on the site.
Earnings-call mentions
Mentioned by Aehr Test Systems
“We also continue to pursue opportunities in memory, including NAND flash and potential high bandwidth memory (HBM) applications with our WLBI solutions roadmap.”
SEC 8-K exhibit (0001654954-26-006655) · Jul 14, 2026
Mentioned by Onto Innovation
“Dragonfly G5 system qualified at a leading 2.5D logic customer and a high-bandwidth memory customer, establishing a new level of performance and flexibility to support current and future demand.”
SEC 8-K exhibit (0001193125-26-206365) · May 5, 2026
Sources
Every claim traced to a primary source — evidence, recent activity, and insider filing behavior, all in one place.
Evidence
2 sources · 78% avg. source confidence
- SK hynix investor materials / industry market research (BofA, Patsnap) — HBM market enters AI-driven supercycle, BofA estimates 2026 market at $54.6B (up 58% YoY)80% source confidence
- DatacenterDynamics — Samsung and SK Hynix scale up memory production capacity in 2026 to meet AI demand75% source confidence
Dependency Map
What High Bandwidth Memory depends on below, and who depends on High Bandwidth Memory above — click any node to make it the new center, 2 levels deep.
Connection type
Click any node to make it the new center. Scroll to zoom, drag to pan.
The Story So Far
Auto-generated from this entity's dated milestones, relationship updates, and sourced evidence — not AI-written, just sorted.
Emerging Signals
High-Bandwidth Memory Demand Growth
CriticalExpansion of artificial intelligence infrastructure is increasing demand for advanced memory technologies.
High Bandwidth Memory's Timeline
A sourced, dated history of High Bandwidth Memory's key moments — founding to present.
Jan 2013 · JEDEC standard introduced
High Bandwidth Memory was first standardized by JEDEC, later becoming the dominant memory technology for AI accelerators.
Jan 2026 · AI-driven supercycle
The HBM market entered an AI-driven supercycle, with BofA estimating the 2026 market at $54.6 billion, up 58% year-over-year.
How These Connections Evolved
How a relationship changed over time, not just its current state — sourced, dated, and traceable.
High Bandwidth Memory ↔ NVIDIA
Jan 2024
HBM3E became the standard memory for NVIDIA's Blackwell and Hopper AI platforms, with NVIDIA qualifying multiple suppliers (SK hynix, Micron, Samsung) to secure capacity amid tight supply.
Jan 2025
HBM4 qualification progressed across NVIDIA's supplier base ahead of next-generation platform launches, with SK hynix maintaining the lead in samples shipped.
Now
High-bandwidth memory is integrated into advanced NVIDIA AI computing platforms.
High Bandwidth Memory ↔ Micron Technology
Jan 2024
Micron ramped HBM3E production and shipments to NVIDIA, positioning itself as the third major HBM supplier behind SK hynix and Samsung.
Jan 2025
Micron advanced HBM4 qualification and expanded high-bandwidth memory capacity to compete for a larger share of AI accelerator demand.
Now
Micron is a major producer of high-bandwidth memory products.
High Bandwidth Memory ↔ SK hynix
Jan 2024
SK hynix held the leading share of the HBM market, supplying the majority of HBM3E used in NVIDIA's AI accelerators.
Jan 2025
SK hynix led early HBM4 sample shipments and qualification, maintaining its position as NVIDIA's primary HBM supplier.
Now
SK hynix is a major producer of high-bandwidth memory for artificial intelligence computing systems.
High Bandwidth Memory ↔ Samsung Electronics
Jan 2024
Samsung worked to pass NVIDIA's qualification for HBM3E after earlier delays, competing with SK hynix and Micron for AI-accelerator memory supply.
Jan 2025
Samsung advanced HBM4 development while working to regain share lost to SK hynix in the HBM3E generation.
Now
Samsung Electronics develops and manufactures high-bandwidth memory products.
High Bandwidth Memory ↔ Advanced Semiconductor Packaging
Jan 2024
Rising HBM stack heights (8-hi to 12-hi) and CoWoS integration requirements made advanced packaging capacity a binding constraint on AI accelerator supply alongside memory production itself.
Now
Integration of high-bandwidth memory with advanced processors depends on sophisticated semiconductor packaging technologies.
High Bandwidth Memory ↔ AI Accelerators
Jan 2024
HBM3E capacity and bandwidth became a key differentiator between competing AI accelerators (NVIDIA Blackwell/Hopper, AMD MI300X), with memory supply often gating overall accelerator output.
Now
High-bandwidth memory supports the performance requirements of advanced AI accelerators.
High Bandwidth Memory ↔ Blackwell
2024–2025
NVIDIA's Blackwell GB200/GB300 platforms shipped with HBM3E memory, with supply constraints and packaging bottlenecks contributing to the widely reported Blackwell production ramp delays.
Now
Blackwell AI systems depend on high-bandwidth memory for high-performance data processing.
High Bandwidth Memory ↔ Micron Technology
2024–2025
Micron invested heavily in HBM3E and HBM4 development and manufacturing capacity, growing its share of the AI memory market from a smaller base than SK hynix and Samsung.
Now
Micron develops and manufactures high-bandwidth memory products for advanced computing systems.
High Bandwidth Memory ↔ SK hynix
2024–2025
SK hynix led global HBM development, shipping the first HBM3E products to NVIDIA in 2024 and advancing HBM4 samples through 2025 to defend its market-leading position.
Now
SK hynix develops and manufactures high-bandwidth memory used in advanced artificial intelligence systems.
High Bandwidth Memory ↔ Advanced Micro Devices
2024–2025
AMD's MI300X and MI350X accelerators used HBM3e memory with a capacity advantage over competing NVIDIA chips, a differentiator AMD marketed directly to hyperscaler customers.
Now
Advanced AMD AI accelerators depend on high-bandwidth memory.
High Bandwidth Memory ↔ AI Accelerators
2024–2025
HBM supply and stack-height generation (HBM3E to HBM4) became a binding constraint on AI accelerator output across NVIDIA, AMD, and custom-ASIC programs.
Now
Advanced AI accelerators depend on high-bandwidth memory for high-performance data processing.
High Bandwidth Memory ↔ Lam Research
2025–2026
As HBM shifted to increasingly tall 3D-stacked structures, Lam Research's etch and deposition tools became a bottleneck-critical input for HBM3E/HBM4 production ramps.
Now
Lam Research dominates the etch and deposition equipment market that physically builds the vertical 3D NAND and DRAM structures high-bandwidth memory requires.