ACTUALITÉS TECH – After DeepSeek’s appearance, it was Nvidia that lost an unprecedented amount of market value!

Nvidia lost nearly $600 billion of its market value in one day on the stock market (a record by the way… its stock value dropped 17%) as investors speculated on the Chinese AI startup DeepSeek and its aftermath. This company has been able to keep up with U.S. companies in hardening models while working on a much smaller budget. Today’s stock market sell-off (which began shortly after the opening bell in Japan) hit the Asian country’s data center stocks, chip equipment maker ASML and Nvidia.

This was reflected in NASDAQ futures and the VanEck Semiconductor ETF. Shares of ASML (which is heavily exposed to the global semiconductor industry as the only supplier of advanced EUV chipmaking equipment) fell 11% in Europe ahead of its crucial earnings report later this week, which will set the pace for 2025 MI expectations. Shares were already down 22% in October when the company forecast weak demand for equipment in 2025, prompting investors to unload some of their demand expectations for MI. ASML ADRs traded on the NASDAQ exchange fell 11% in premarket trading, wiping about $30 billion off the company’s value. Over the past six months, ADRs are down 16% excluding today’s premarket losses. Japanese investment conglomerate Softbank, which earlier this month tried to make headlines with President Trump’s $500 billion AI initiative, fell 8.32% in Japanese markets. Chp testing equipment maker Advantest was also down 8.6%.

One of the hardest hit chip stocks after Nvidia is semiconductor maker Broadcom. Its shares jumped a staggering 38% in December when management announced that the company expects to generate tens of billions of dollars in revenue from sales of its AI chips by 2027. However, as investors question whether demand for AI chips will actually materialize given the current prevailing beliefs about DeepSeek’s training resources, they are fleeing Broadcom stock in droves. Its shares were down 10.5% in pre-market trading, wiping $100 billion off the company’s value from Friday. AMD was also down 5.3% in pre-market trading, and British chip designer Arm Ltd, whose CEO Rene Haas repeatedly insisted earlier this month that his company’s CPUs are on par with any MI GPU, fell 9.43%. Microsoft was down 6% and Meta also saw a 4.6% drop as investors ponder the wisdom behind their multi-billion dollar investments. Apple was also down (0.9%).

Over 1/3rd of Nvidia sales go to China, probably ~$40B last year.

The Singapore backdoor is real: Nvidia even says “shipments to Singapore are insignificant” while 22% of billings last qtr were to Singapore. pic.twitter.com/AGeArCgTI1

— Sheel Mohnot (@pitdesi) January 25, 2025

As a result, DeepSeek shocked the global technology industry with its groundbreaking R1 model, which is more than competitive with OpenAI’s o1 model, but at fifty times the typical cost of training a complex LLM. More recently, Scale AI founder and CEO Alexandr Wang stated in an interview that DeepSeek has about 50,000 H100s. Nvidia’s H100 GPUs cannot be sold to China under U.S. law, and Wang’s statement provoked mixed reactions, with some dismissing his claims outright. Then Elon Musk got into the act and backed up Wang’s statement about DeepSeek’s alleged H100 cache. This quickly turned into a speculative frenzy, with netizens pointing to Singapore as a possible vector for DeepSeek to acquire Nvidia’s H100 chips. Some have pointed out that Singapore accounted for 22% of Nvidia’s billings last quarter, but actual shipments were insignificant according to Nvidia’s filings with the SEC.

WCCFTech reached out to Nvidia, who responded with this statement: “DeepSeek is an excellent AI advancement and a perfect example of test-time scaling. DeepSeek’s work illustrates how new models can be created using this technique, leveraging widely available models and computation that is fully export control compliant. Inference requires a significant number of NVIDIA GPUs and high-performance networking. We now have three scaling laws: pre-training and post-training, which continue, and new test-time scaling.” The company believes that DeepSeek was able to achieve the breakthrough by fully complying with U.S. export controls, seemingly dispelling rampant speculation about the startup’s significant GPU inventory.

Is the AI balloon about to burst? Can we finally get to the point where Nvidia might release its graphics cards at a COMPETITIVE price? There hasn’t been much of an example since the GTX 900 and 1000 series…

Source: WCCFTech,

Obviously

— Elon Musk (@elonmusk) January 27, 2025

v