Tuesday saw tech stocks rebound according to Reuters. Chip designer NVIDIA, whose GPU chips are used to train AI models and unlock AI capabilities, and are also used to run AI applications, had plunged 17% on Monday losing $593 billion in valuation due to the low cost of the DeepSeek model. NVIDIA rebounded 8.9% on Tuesday and remains the AI stock that all professionals want in their portfolios. Over the last year, NVIDIA remains up over 106%.
“Yesterday was an initial reaction. Today investors are asking if anybody did any sort of homework and made sure DeepSeek is exactly what they say it is. Can we have more proof that they really built it for so much less?”-JJ Kinahan, president of Chicago’s tastytrade brokerage in Chicago
DeepSeek’s placement at the top of the App Store’s Free Apps list set off a huge tech sell off on Monday. | Image credit-Apple
The word is that unlike NVIDIA, Apple benefits from DeepSeek’s emergence. That’s because the excitement and popularity of DeepSeek’s App Store app surely generated a large number of new customers for Apple’s app storefront. Furthermore, DeepSeek could be so efficient that it might be able to run on the iPhone without raising costs although this is still something that is not expected to take place right now. Analysts at Jefferies said, “DeepSeek’s success offers some hope but there is no impact on AI smartphone’s near-term outlook.”
Like other AI chatbots developed in China, DeepSeek gets a little touchy when Chinese leader Xi Jinping’s policies are brought up and tries to move away from this topic. DeepSeek’s R1 reasoning model was released earlier this month and there are varying answers about how many GPUs were used to develop DeepSeek R1. Jefferies estimates a “training cost of only US$5.6m (assuming US$2/H800 hour rental cost). That is less than 10% of the cost of Meta’s Llama.”
