Nvidia's AI Chip Orders Remain Strong Despite Jensen Huang's Departure

Despite the departure of CEO Jensen Huang, Nvidia's demand for artificial intelligence (AI) semiconductors remains robust. Following better-than-expected earnings and revenue forecasts, the company has significantly expanded its long-term supply agreements focused on memory, signaling positive developments for Samsung Electronics and SK Hynix's high bandwidth memory (HBM) business. Domestic semiconductor firms are prioritizing the increase in memory demand driven by expanded AI infrastructure investments over uncertainties stemming from management changes.


On November 18, Nvidia reported that its revenue for the third quarter of fiscal year 2027 (August to October) reached $115.8 billion, a 103% increase compared to the same period last year. Data center revenue surged to $107.6 billion, up 110%. The revenue forecast for the next quarter is projected at $130 billion, exceeding market expectations of $125.3 billion.


A key focus for the domestic memory industry is the expansion of long-term supply agreements. Nvidia's agreements for component supply and production capacity have increased to $390 billion, up $111 billion from the previous quarter's $279 billion. Nvidia explained that the primary reason for this increase is the memory purchase contracts expected over the coming years. Notably, the allocated amount for fiscal year 2028 is set to reach $203 billion, indicating a focus on securing long-term supply chains.


The ramp-up in shipments of the next-generation AI semiconductor, 'Vera Rubin,' is also supporting HBM demand. Nvidia has begun mass production shipments of the Vera Rubin NVL72 this quarter, which features sixth-generation HBM, HBM4. The expansion of next-generation GPU production could lead to increased HBM4 supply from domestic memory companies.


Samsung Electronics is accelerating its HBM4 supply expansion. In the third quarter, HBM4 revenue more than tripled compared to the previous quarter. The company has signed long-term memory supply agreements with ten major clients and is negotiating with an additional ten.


In its third-quarter report released on November 16, Samsung stated, "Despite our efforts to increase production, demand growth continues to outpace our capabilities." Consequently, the company plans to expand its supply of high-value AI-related products, including HBM, server DRAM, and solid-state drives (SSDs).


SK Hynix has also secured supply volumes based on long-term contracts. In its earnings announcement last month, the company stated that all DRAM and HBM volumes planned for production over the next three years, until 2029, are already booked. The company is expanding production capacity at its Cheongju M15X facility to respond to next-generation HBM supply.


In its recent third-quarter report, SK Hynix assessed that "the server memory market continues to show structural growth as AI infrastructure investments transition from learning to inference and agent stages." With the expansion of AI service utilization, demand for high-performance memory is expected to continue rising.


Price increases are also a positive factor for domestic companies. Market research firm TrendForce projected at an event held in Seoul that HBM4 prices will rise from over $2.05 per gigabit this year to between $3 and $5.5 next year. The price of the next-generation product, HBM4E, is also reportedly being discussed at over $5.


However, Nvidia's strong performance does not immediately translate into new order confirmations for domestic companies. A significant volume has already been secured through long-term contracts, and future investments in AI data centers and changes in Nvidia's management strategy remain variables.


Han Ji-young, a researcher at Kiwoom Securities, commented on Nvidia's performance, stating, "The results have somewhat alleviated the risks associated with Nvidia following the recent CEO resignation." However, she added, "Nvidia's performance alone does not resolve all uncertainties regarding funding burdens for AI companies and data center regulations."





* This article has been translated by AI.