Samsung's HBM Chips Failing Nvidia Tests: Sources Reveal Heat and Power Consumption Woes!

Samsung's HBM chips are reportedly failing Nvidia tests due to concerns related to heat and power consumption, according to sources.

Samsung Electronics' newest high bandwidth memory chips are facing difficulties in passing Nvidia's tests for integration into the U.S. firm's AI processors. This is reportedly due to issues related to heat and power consumption, as informed by three individuals briefed on the matter.

The issues are impacting Samsung's HBM3 chips, which represent the fourth-generation HBM standard widely used in graphics processing units for artificial intelligence. Additionally, the fifth-generation HBM3E chips, introduced to the market by Samsung and its competitors this year, are also affected, as indicated by the sources.

This marks the first report on the reasons behind Samsung's failure to pass Nvidia's tests.

In response to Reuters, Samsung stated that HBM is a customized memory product requiring optimization processes aligned with customers' requirements. It mentioned ongoing efforts to optimize its products through close collaboration with customers but refrained from commenting on specific clients.

In separate statements issued after Reuters initially published this report, Samsung denied the claims of failing due to heat and power consumption, asserting that testing was proceeding smoothly and according to plan. Nvidia chose not to provide a comment.

HBM, a type of dynamic random access memory (DRAM) standard introduced in 2013, involves vertically stacking chips to save space and reduce power consumption. It plays a crucial role in processing the vast amounts of data generated by complex AI applications. With the increasing demand for sophisticated GPUs, particularly amid the growth of generative AI, the demand for HBM has also surged.

Satisfying Nvidia, which holds approximately 80% of the global GPU market for AI applications, is considered essential for the future growth of HBM manufacturers, both in terms of reputation and profit momentum.

Samsung has been striving to pass Nvidia's tests for HBM3 and HBM3E since last year, according to three sources familiar with the matter. In April, the results of a recent failed test for Samsung's 8-layer and 12-layer HBM3E chips were reported, although it remains unclear if the issues can be easily resolved. These failures to meet Nvidia's requirements have raised concerns in the industry and among investors that Samsung may fall further behind rivals SK Hynix and Micron Technology in the HBM market.

The sources, two of whom were briefed on the matter by Samsung officials, spoke anonymously due to the confidential nature of the information. Samsung's shares fell 2% in early Friday trade, slightly underperforming the broader market.

In contrast, domestic rival SK Hynix serves as the primary supplier of HBM chips to Nvidia and has been supplying HBM3 since June 2022. It also began supplying HBM3E in late March to an undisclosed customer, rumored to be Nvidia. Micron, another major HBM manufacturer, has also committed to supplying Nvidia with HBM3E.

This week, Samsung replaced the head of its semiconductor unit, a move believed to reflect concerns about its lagging position in the HBM market. Market expectations were high for Samsung to swiftly pass Nvidia's tests, given its status as the world's largest memory chip maker. However, specialized products like HBM often require time to meet customers' performance evaluations, noted Jeff Kim, head of research at KB Securities.

Although Samsung has yet to secure a position as a supplier of HBM3 to Nvidia, it supplies customers such as Advanced Micro Devices and aims to commence mass production of HBM3E chips in the second quarter, as stated in its response to Reuters.

Analysts suggest that SK Hynix has invested significantly more time and resources in HBM research and development over the past decade, contributing to its technological advantage. Nonetheless, Samsung emphasizes its history of HBM innovation, highlighting its development of the first commercial HBM solution for high-performance computing in 2015.

GPU manufacturers like Nvidia and AMD are eager for Samsung to refine its HBM chips, providing more vendor options and potentially weakening SK Hynix's pricing power, according to the sources.

At an Nvidia AI conference in March, Nvidia CEO Jensen Huang expressed enthusiasm for Samsung's 12-layer HBM3E, underscoring the company's interest in Samsung supplying these chips. HBM3E chips are expected to become the mainstream HBM product in the market this year, with shipments concentrated in the second half of 2024, according to research firm Trendforce.

Separately, SK Hynix anticipates a significant increase in demand for HBM memory chips, estimating an annual growth rate of 82% through 2027.

Investors have taken note of Samsung's relatively weaker position in the HBM market, reflected in its share performance compared to SK Hynix and Micron. While Samsung's shares are down 2% year-to-date, SK Hynix's stock has risen by 42%, and Micron's shares have increased by 48%.

This article was generated from an automated news agency feed without modifications to text.