HBM Powers Growth for Micron and Samsung
Micron Technology has announced better-than-expected quarterly earnings and expressed optimism for the current quarter, despite Morgan Stanley's pessimistic report predicting an oversupply of semiconductor chips and a decline in prices.
The U.S. chipmaker saw its revenue for the fourth fiscal quarter, spanning June to August, grow by 93% year-over-year. This growth was driven by a surge in demand for data center DRAM and high-bandwidth memory (HBM), which are critical for artificial intelligence applications. Both memory and NAND chip revenue hit all-time highs during this period.
"We are entering fiscal 2025 with the strongest competitive position in Micron's history. We expect record revenue in the first quarter, along with substantial growth in profitability throughout fiscal 2025," said Sanjay Mehrotra, President and CEO of Micron Technology. He added that the company has already sold all HBM products planned for production in 2024 and 2025.
Micron’s Computing and Networking business unit, responsible for AI-related chips, saw its revenue soar by 152% year-over-year and 17% quarter-over-quarter in the fourth fiscal quarter. Micron is one of the top three global suppliers of HBM chips, alongside SK Hynix and Samsung Electronics. HBM is a high-performance memory designed to handle massive data processing at exceptional speeds.
Looking ahead, Micron anticipates first-quarter revenue to hit a record $8.7 billion, with a margin of $200 million, and forecasts gross margins to jump to approximately 39.5%. The company's earnings projections slightly surpass analysts' expectations, standing in stark contrast to Morgan Stanley's outlook for the semiconductor sector.
On September 13, Morgan Stanley suggested that the DRAM and HBM markets may enter a downturn next year due to weak demand and oversupply. The firm also forecasted that HBM production could more than double by 2025.
However, Micron is not the only company predicting growth in the HBM sector. Samsung Electronics, the world's largest memory chipmaker, also painted a promising picture for the HBM market, expecting demand and supply to reach equilibrium at 25.0 GB by 2025. Samsung projects that wafer input for HBM will rise from 16% in 2024 to 28% in 2025 and that the custom AI memory market will expand from 12.9 GB in 2023 to 229.4 GB by 2028.
Analysts at Meritz Securities noted that the market is shifting toward custom memory products, with server DRAM and HBM now making up half of DRAM demand, giving manufacturers more leverage in pricing negotiations. Despite concerns about significant facility investments, industry experts believe these expenditures are primarily aimed at increasing the production of high-value chips like HBM and modernizing older production lines. As a result, they argue, these investments are unlikely to lead to an oversupply issue.
Tech giants such as Google, Microsoft, and others are making substantial investments in AI servers. Nvidia is set to begin mass production of its next-generation AI chips by the end of this year, with expectations to generate $10 billion in revenue in Q4. These upcoming AI chips will be equipped with eight HBM chips, and the AI servers will include both AI processors and solid-state drives.
Global consulting firm Bain & Company echoed this positive outlook, warning of a potential AI chip shortage. Their report predicts that by 2026, demand for HBM and NAND will grow by 60%-65% and 30%-35%, respectively, compared to 2023.
