High Bandwidth Memory
PulseAugur coverage of High Bandwidth Memory — every cluster mentioning High Bandwidth Memory across labs, papers, and developer communities, ranked by signal.
- employed by Sanjay Mehrotra 95%
- partners with SK Group 90%
- located in Cheongju 90%
- developed zHBM 90%
- used by Nvidia Rubin Gpu 90%
- instance of 3D dram structure with high mobility channel 90%
- competes with High Bandwidth Flash 80%
- used by Hbm4 70%
- developed by Hbm4 70%
- affiliated with Hbm4 70%
- instance of Hbm4 70%
- instance of Hbm4e 70%
- 2026-10-21 product_launch Samsung has sold out its High Bandwidth Memory production capacity through 2027, leading to anticipated price increases. source
- 2026-06-02 research_milestone TrendForce projects HBM contract prices to potentially double by 2027 due to ongoing DRAM shortages and manufacturing complexities. source
18 day(s) with sentiment data
Samsung's HBM production delays due to bonus dispute to impact Q3 2026 AI accelerator availability
The ongoing bonus dispute at Samsung, leading to worker revolts and production halts, directly impacts the testing and packaging of AI memory chips. This disruption is likely to cause delays in Samsung's HBM production schedules, potentially leading to a shortage of critical components for AI accelerators in Q3 2026.
Micron's US DRAM expansion to partially offset HBM-driven shortages in automotive and defense sectors
Micron's investment in US-based DDR4 DRAM production, supported by the CHIPS Act, aims to alleviate shortages in critical sectors like automotive and defense. While HBM demand pulls capacity away from these sectors, Micron's expansion could provide a much-needed supply buffer for older memory standards.
HBM demand is significantly constraining supply for traditional memory markets
The increasing demand for HBM in AI data centers is consuming a larger portion of wafer capacity, leading to a shortage of traditional DDR and LPDDR memory. This shift is driving up prices for consumer electronics and impacting the availability of memory for non-AI applications.
HBM supply constraints to persist through 2027, impacting consumer electronics availability
Evidence suggests memory manufacturers are prioritizing high-margin HBM for AI, leading to shortages and price increases for traditional DDR and LPDDR memory used in consumer electronics. With new capacity not expected until late 2027, this constraint is likely to continue affecting the availability and affordability of devices like smartphones.
Samsung's HBM packaging advancements may target mobile integration
Samsung is developing advanced HBM packaging technology that could be integrated into mobile devices. This is notable given the current trend of HBM consuming significant wafer capacity for AI data centers, potentially indicating a future where high-performance memory becomes more prevalent in consumer mobile hardware.
-
Memory chip CEO warns of severe 2027 shortage, AI demand driving prices
The CEO of memory chip manufacturer Team Group has warned of a severe global shortage of DRAM and NAND flash chips extending through the first half of 2027. This shortage is driven by AI servers consuming a large portio…
-
Google's Elephant-Goldfish AI aids software design with memory features
Google has developed a new AI model called Elephant-Goldfish, designed to assist in software development by incorporating memory capabilities and enabling fresh review of code. This development comes as the broader tech…
-
Sam Altman meets Samsung, SK Hynix amid AI chip crunch
OpenAI CEO Sam Altman recently met with executives from Samsung and SK Hynix to discuss AI capacity and the increasing demand for High Bandwidth Memory (HBM). While no new memory deals were finalized, the meetings highl…
-
New TACT-KV method compresses 3D medical VLM caches, boosting batch size 8x
Researchers have developed TACT-KV, a novel method for compressing the key-value (KV) cache in vision-language models (VLMs) that process 3D medical imaging data. Unlike previous methods that only consider 1D or 2D tran…
-
China's AI sector faces GPU chip and HBM supply constraints
China's AI sector is facing significant challenges due to restrictions on US-supplied GPU chips and global High Bandwidth Memory (HBM) availability. This supply chain constraint is leading to the repurposing of some AI …
-
AI efficiency gains unlikely to lower RAM prices amid surging demand
Despite advancements in AI model memory efficiency, the price of RAM is unlikely to decrease. While techniques like lower KV-cache usage can reduce memory needs per task, the overall demand for memory is increasing due …
-
Looped transformers' HBM traffic savings on Nvidia B200 analyzed
A recent analysis explores the efficiency of looped transformer architectures, specifically the Huginn model, in managing High Bandwidth Memory (HBM) traffic on Nvidia's B200 hardware. The study found that while sharing…
-
AI chip smugglers exploit loophole to bring banned HBM into China
Smugglers are exploiting a loophole in trade controls to bring banned High Bandwidth Memory (HBM) chips from South Korea into China. These expensive HBM chips, crucial for AI GPUs, are paired with cheaper, less regulate…
-
Sam Altman explores AI data centers and HBM supply in South Korea
Sam Altman, CEO of OpenAI, is in Seoul to strengthen ties with South Korean tech giants, specifically SK Hynix and Samsung. The discussions focus on securing High Bandwidth Memory (HBM) for OpenAI's Stargate project and…
-
South Korea invests $16B in AI, driving record chip exports
South Korea is making a significant investment in AI and related technologies, allocating approximately $16 billion for the upcoming year to boost AI, chips, and data centers. This strategic focus is already showing eco…
-
OpenAI CEO Sam Altman seeks more AI chips from Samsung amid shortages
OpenAI CEO Sam Altman met with Samsung leadership in South Korea to discuss securing additional High Bandwidth Memory (HBM) for AI data centers. The meeting, which included executives from Samsung's electronics, biotech…
-
LLM power capping ineffective during decode, new research finds
A new research paper challenges the effectiveness of power capping on GPUs for Large Language Model (LLM) serving, particularly during the autoregressive decode phase. The study found that power caps have minimal impact…
-
Intel stock plummets amid PC market woes; Samsung Electronics holds strong in HBM
Intel experienced a significant stock decline of 28.6%, contributing to a 7.2% drop in the SOX index, as the KOSPI closed at 5,730.74 on October 30th. The primary driver for Intel's fall was its exposure to the PC marke…
-
Samsung's HBM Supply Sold Out Through 2027 Amidst High Demand
Samsung has sold out its High Bandwidth Memory (HBM) supply through 2027, driven by strong demand from hardware buyers. The chipmaker's success is attributed to long-term contracts and a significant increase in operatin…
-
China's CXMT prioritizes AI memory production despite losses · 1 source tracked
China's Changxin Memory Technologies (CXMT) is prioritizing the production of high-bandwidth memory for AI accelerators, even at a financial loss. This strategic move, supported by significant state subsidies, aims to f…
-
AI memory chip shortage to last until 2030, warns SK Hynix
SK Hynix, a key supplier of High Bandwidth Memory (HBM) for Nvidia, has warned that the global shortage of memory chips is expected to persist until 2030. The company has already committed over 50% of its production cap…
-
SK Hynix: Memory chip shortage to last until 2030, AI demand surges
SK Hynix, a leading memory chip manufacturer, has issued a stark warning that the global shortage of essential semiconductors could extend until 2030. The company reported record quarterly revenue and operating profit, …
-
New paper reviews cloud LLM architecture, orchestration, and inference
A new paper published on arXiv details a comprehensive review of Large Language Model (LLM) deployment in cloud data centers. It addresses challenges such as High Bandwidth Memory limitations, KV cache expansion, and se…
-
Hopper GPU value skyrockets amid HBM scarcity
Nvidia's Hopper GPUs, previously thought to be in low demand, have seen their value significantly increase. This surge is attributed to a scarcity of High Bandwidth Memory (HBM), a critical component for AI accelerators…
-
ASI may be an economic loop, not a robot uprising, author suggests
Artificial Superintelligence (ASI) may manifest not as a robot uprising, but as an economic loop where money is converted into tokens to justify hardware purchases, which in turn generate more tokens. This cycle, remini…