Nvidia's Challenges with TSMC: A Look at Production and Demand
Written on
Chapter 1: Nvidia's Order Reduction with TSMC
Nvidia is currently looking to decrease its orders with TSMC for the manufacturing of its GeForce RTX 4000 series. Since 2021, the company has allocated about $9 billion to secure production capacity on 5nm lines. However, Nvidia has realized that it will not fully utilize this booked capacity, leading to efforts to scale back these commitments—a process that proves to be quite complex.
Reports indicate that Nvidia's spending on TSMC's production capacity surged significantly, reaching $6.9 billion by the third quarter of 2021 and climbing to approximately $9 billion by February of this year (refer to sources like HardwareTimes for details).
Section 1.1: Intel's Ongoing Server Challenges
Amidst these developments, Intel is grappling with its own issues. The anticipated 7nm Xeon “Sapphire Rapids” processors may experience delays of up to 18 months, as the company acknowledges the challenges ahead.
Similar to Nvidia, Qualcomm also invested substantial sums into TSMC, having previously utilized Samsung for its last generation of products.
Chapter 2: The Current Hardware Market Landscape
Nvidia is facing a downturn in hardware demand, with significant quantities of current products circulating in distribution channels. The market is flooded with surplus cards, particularly from miners.
This situation echoes events from late 2013 and early 2014, when AMD's then-CEO Rory Read misjudged demand, leading to overstock and a decline in market value. The oversupply forced AMD to delay new product launches and compensate customers for losses. Ultimately, this mismanagement led to Read's exit and Lisa Su's ascendance, which turned the company's fortunes around.
Despite the similarities, Nvidia's circumstances differ in several key aspects. Unlike AMD in 2014, Nvidia possesses a new generation of products—Lovelace—which are substantially more powerful than the existing inventory. This enables Nvidia to transition the high-end mining cards to mainstream status while introducing new, higher-performing models.
Section 2.1: Upcoming Technological Advances
Looking ahead, PCI Express 7.0 is in development and is expected to deliver link speeds that are 16 times faster than the current standard.
However, the challenge of excess hardware in storage remains a concern. Nvidia is wary of devaluing its existing inventory, prompting them to consider delaying the release of the next generation of products and reducing their orders for 5nm capacity from TSMC.
Section 2.2: TSMC's Reluctance to Adjust Orders
TSMC, recognizing that customers may overbook and then scale back, is hesitant to alter contracted deliveries. Currently, they are only willing to postpone the initial batch of chips until the fourth quarter of this year, potentially extending into the first quarter of next year.
The ongoing chip shortage is projected to last several more years, impacting various industries, including automotive. Despite these challenges, Nvidia must find solutions to its unused wafer commitments with TSMC.
Even in a scenario where TSMC agrees to adjustments, Nvidia would still need to secure alternative clients for the surplus wafer capacity. Other customers, like Apple, have begun scaling back their orders, cutting the initial production of iPhone 14 components by 10%. AMD has also adjusted its orders, reducing 7nm generation wafer requests by approximately 20,000 units for the upcoming quarters.
I hope you found this analysis insightful. If you’re interested in more content like this, please explore my other writings. I enjoy sharing my thoughts and would love to hear your opinions in the comments—what are your top insights? Your feedback is always appreciated!