NVIDIA’s earnings announcement scheduled for the 21st in Korean time has seen increased focus on issues with the overheating of their next-generation AI semiconductor, Blackwell. This comes amidst ongoing rumors about production disruptions, previously highlighted due to inferior power component performance.
As NVIDIA prepares to release its fiscal 2025 third-quarter (August to October) earnings after the close of the New York Stock Exchange on the 20th (local time), speculation about potential production issues that could impact results has resurfaced. The US IT media outlet The Information reported on the 17th (local time) that NVIDIA’s Blackwell AI chip has experienced overheating issues when connected to custom-designed server racks, prompting NVIDIA to request several server rack design changes from suppliers to address the issue.
Additionally, some customers are reportedly concerned that applying Blackwell may not allow enough time to operate new data centers, according to an anonymous source. These reports emerged after Japan’s SoftBank Group announced on the 13th that it is developing a supercomputer equipped with Blackwell chips.
Prior to this, companies like Microsoft and Alphabet, which pre-ordered Blackwell, were reportedly considering increasing orders for existing AI products from the Hopper line, such as the H100 and H200, instead of Blackwell. NVIDIA explained that their engineering team is cooperating with cloud service providers to resolve the issue.
However, Edgewater Research previously highlighted concerns about performance issues with power management components supplied by Monolithic Power Systems for Blackwell, suggesting that these components might cause a delay in Blackwell’s production. Consequently, the stock prices of US semiconductor-related companies Monolithic and Micron Technology, seen as beneficiaries of Blackwell, dropped about 25% and 14% respectively from the 11th to the 15th of the month, alongside a roughly 4% decline in NVIDIA’s stock price.
Blackwell is NVIDIA’s next-generation AI semiconductor product, unveiled last March, known for responding up to 30 times faster than existing products in tasks like generative AI chatbots. Although NVIDIA initially announced a Q2 release this year, reports indicated that defects found during production led to a delay of at least three months for its official launch.
Despite the reinvigorated rumors of Blackwell production issues since November 11th, Wall Street remains predominately bullish on NVIDIA’s stock. Since that time, at least eight firms, including Citi, Wedbush, and Mizuho, have raised their target prices based on AI optimism and NVIDIA’s market dominance. The highest target price was set by the British investment bank HSBC, which raised NVIDIA’s target price from $145 to $200.