Micron's new memory boasts 30% density boost; DeepSeek AI model faces chip shortage
Boise, Friday, 27 June 2025.
Micron has begun sampling its LPDDR5X memory chips, utilizing EUV lithography. The new process increases density by 30% and improves performance, while cutting power use. DeepSeek’s development of its R2 AI model is facing delays. The delays are due to a shortage of Nvidia’s H20 chips in China. This highlights the critical role of chip availability in AI advancements and Nvidia’s strong market influence.
Micron’s memory innovation
Micron’s announcement of sampling LPDDR5X memory chips, created with its 1γ process using EUV lithography, signals a leap in memory technology [1]. The 1γ process, Micron’s 6th generation 10nm-class production node, promises a 30% increase in bit density [1]. This new technology also slashes power consumption by 20% and boosts performance by 15% compared to the prior 1β technology [1]. Micron CEO Sanjay Mehrotra stated the company is making excellent progress on the 1γ DRAM technology, with yield ramping ahead of the pace of the 1β node [1].
Strategic implications for Micron
Micron’s strategic move to leverage 1γ across its DRAM product portfolio demonstrates its commitment to advanced memory solutions [1]. Currently, Micron is producing 1γ DRAM in Japan, where its first EUV tool was installed in 2024 [1]. Plans are in place to expand EUV capacity in both Japan and Taiwan [1]. This positions Micron to potentially capture a larger market share in the high-performance memory sector. The 1γ process combines EUV with DUV multi-patterning and introduces an upgraded high-k metal gate and redesigned BEOL circuitry [1].
Nvidia’s H20 chip shortage impacts DeepSeek
DeepSeek’s R2 AI model development faces delays due to a shortage of Nvidia’s H20 processors in China [2]. The U.S. government restricted sales of Nvidia’s H20 processors in mid-April [2]. DeepSeek previously utilized a cluster of 50,000 Hopper GPUs, including 30,000 H20s, to train its R1 model [2]. The CEO of DeepSeek, Liang Wenfeng, indicated dissatisfaction with the R2 model’s current capabilities [2]. This bottleneck underscores the reliance of Chinese AI companies on American hardware, especially Nvidia [2].
Nvidia’s strategic partnerships and market position
Nvidia is actively expanding its reach through strategic partnerships [3]. Nvidia and Deutsche Telekom recently announced a collaboration to build Germany’s first industrial AI cloud [3]. Trend Micro announced the implementation of the NVIDIA Agentic AI Safety blueprint [3]. Nvidia’s CEO, Jensen Huang, noted Europe is in a golden age of AI development but lacks infrastructure [3]. These moves, combined with the high demand for Nvidia’s chips, reinforce Nvidia’s dominant position in the AI infrastructure market [2][3].
Market reaction and financial outlook
Nvidia’s stock currently trades at 155.020, up 0.46% with a total market capitalization of 3.78 trillion [6]. Its price-to-earnings ratio is 50.01 [6]. Recent news highlights Nvidia’s continued importance in AI and related fields [3]. OpenAI is developing an AI-integrated office software to compete with Google and Microsoft [3]. Nvidia is reportedly designing a B30 AI GPU for China as a workaround to export restrictions [3]. These developments suggest sustained investor interest in Nvidia [6].