Loading IndicatorLoading Indicator

Nvidia’s $115.8 Billion Quarter and Vera Rubin Rollout Boost Memory Outlook for Samsung, SK Hynix

Source
Korea Economic Daily

Summary

  • Nvidia said it posted record results with $115.8 billion in revenue and $107.6 billion in data center revenue for the third quarter of fiscal 2027.
  • The company said shipments of its next-generation AI platform 'Vera Rubin NVL72' and its 20.7TB of HBM4 would expand demand for high-performance memory.
  • Counterpoint Research said demand for high-performance memory including HBM4, server DRAM and LPDDR5X would rise, creating more growth opportunities for South Korean memory companies.

Forecast Trend Report by Period

Loading IndicatorLoading Indicator

Nvidia sets another record as memory gains for Samsung and SK Hynix grow

Nvidia posts $115.8 billion in fiscal third-quarter revenue

Data center remains the main driver, up 110% from a year earlier

Vera Rubin shipments begin, raising expectations for broader HBM4 demand

Photo: Shutterstock
Photo: Shutterstock

Nvidia more than doubled revenue in its fiscal third quarter from a year earlier, delivering another record performance. As shipments begin for its next-generation AI platform, Vera Rubin, the results are opening new growth opportunities for Samsung Electronics and SK Hynix. Demand is poised to rise in particular for HBM4, the latest high-bandwidth memory product, and server DRAM.

Nvidia revenue hits $115.8 billion, with data center making up 93%

According to industry data on Nov. 18, Nvidia reported in earnings released on Nov. 17 that revenue for the third quarter of fiscal 2027, covering August through October, reached $115.844 billion. That was up 103% from a year earlier and 20% from the previous quarter. It also topped the market forecast of about $109.1 billion. Operating profit rose 115% from the same period a year earlier to $77.5 billion.

The data center business drove the increase. Third-quarter data center revenue surged 110% from a year earlier to $107.6 billion, accounting for about 93% of total revenue. Continued AI infrastructure spending by global big tech companies appears to have boosted sales of related chips and systems.

Nvidia projected fourth-quarter revenue of $130 billion for fiscal 2027, with a margin of error of plus or minus 3%. That was above the market estimate of about $125 billion. The forecast excludes revenue from its computing business for data centers in China. The fourth quarter also includes one more week than a typical fiscal quarter.

The start of shipments for the next-generation AI platform Vera Rubin NVL72 also stood out. In its earnings release, Nvidia said it is ramping up Vera Rubin production and has begun supplying the product to customers. Major cloud providers including Google Cloud, Microsoft Azure, Oracle Cloud and CoreWeave are adopting the system.

HBM4 reaches 20.7TB across 72 GPUs, creating a bigger opening for Samsung and SK Hynix

Vera Rubin is Nvidia's next-generation AI platform following Blackwell. Its core system, NVL72, integrates 72 Rubin graphics processing units and 36 Vera central processing units into a single rack. Each GPU carries 288 gigabytes of HBM4, bringing total HBM4 capacity to about 20.7 terabytes.

Counterpoint Research said Nvidia's transition to the next-generation platform would be positive for South Korean memory chipmakers. As AI models grow larger and computing workloads increase, high-performance memory that can process data quickly is becoming more important. The HBM4 used in Vera Rubin offers higher bandwidth and greater capacity than the previous generation.

Samsung Electronics and SK Hynix are also accelerating efforts to expand HBM4 supply. SK Hynix said in its third-quarter earnings release last month that fourth-quarter HBM4 shipments and revenue would both surpass HBM3E this year. The company plans to raise output at its Cheongju M15X plant to the maximum level by year-end to expand HBM4 production.

Samsung Electronics also said it sharply increased HBM4 sales in the third quarter. In the fourth quarter, it plans to expand sales of AI-related products including HBM, server DRAM and solid-state drives. The company sees demand rising mainly for memory products tailored to next-generation CPU and GPU platforms.

Rising investment in AI servers is affecting not only HBM but also the broader DRAM market. As GPU computing power increases, the memory capacity required to support CPUs is also rising. Counterpoint Research said memory capacity per CPU socket in Vera Rubin systems reaches as much as 1.5 terabytes, about three times the level of the previous generation. The system uses LPDDR5X, a low-power DRAM.

Recent DRAM market data show revenue growth in general-purpose products outpacing HBM. According to figures Counterpoint Research released earlier this month, third-quarter revenue for DDR and LPDDR rose about 500% and 400%, respectively, from a year earlier. The gains were driven by both higher memory content in AI servers and rising prices.

As demand for high-performance memory grows, securing production capacity is becoming more important. Counterpoint Research said supply constraints may persist as AI systems become more complex, particularly in memory bandwidth, advanced packaging, networking and power infrastructure. That helps explain why major memory makers are expanding capacity while increasing long-term supply contracts.

Counterpoint Research said the shift to the next-generation AI platform Vera Rubin is set to accelerate demand for high-performance memory, including HBM4. Expanding AI server investment could support not only HBM but also demand for server DRAM, creating additional growth opportunities for South Korean memory companies.

Hong Min-seong, Hankyung.com reporter mshong@hankyung.com

#Semiconductor
Korea Economic Daily

Korea Economic Daily

hankyung@bloomingbit.ioThe Korea Economic Daily Global is a digital media where latest news on Korean companies, industries, and financial markets.

What do you think about this news?

‌
‌
‌
‌
‌
‌
‌

PiCK News

‌
‌
‌
‌
‌

Hashtag News

‌
‌
‌
‌