nvidia's huang dismisses deepseek threat, predicts surge in chip demand
Shenzhen, Thursday, 20 March 2025.
nvidia ceo jensen huang addressed concerns about deepseek’s r1 model impacting chip demand. he stated that the model would increase the demand for computing infrastructure. huang believes worries about decreasing chip demand are unfounded. deepseek’s r1 model, developed in two months for under $6 million, rivals openai’s o1. this model initially caused nvidia’s stock to drop nearly 17%. huang’s statement reinforces nvidia’s growth prospects amidst increasing ai development.
gtc conference insights
huang addressed analysts and investors at nvidia’s gtc conference on wednesday, march 19, emphasizing that the market’s understanding of deepseek’s r1 model was “completely wrong” [1]. he argued that these models would increase computing demands significantly [1][5]. nvidia is trying to convince industries to invest in ai equipment, promising economic benefits [2]. the company faces scrutiny in 2025 regarding sustained ai infrastructure spending from its clients [2]. at the gtc conference in san jose, approximately 25,000 attendees were present, including representatives from microsoft, google, waymo, and ford [4].
market reaction and nvidia’s response
the release of deepseek’s r1 model in january caused nvidia’s stock to plummet nearly 17% [1]. it raised questions about the extensive investments in ai models and data centers by major tech companies [1]. however, nvidia’s major clients reaffirmed their ai spending plans in recent earnings reports [1]. some analysts noted that data center operators’ spending is growing faster than expected [1]. nvidia’s stock experienced a 2.4% increase on wednesday, though it is still down 14% year-to-date [2].
competitive landscape and nvidia’s edge
huang addressed concerns about customers developing their own chips to replace nvidia’s ai accelerators [1][2]. he specifically mentioned google’s collaboration with broadcom on asic chips [1]. huang countered that many asics are not suitable for data centers [1][2]. he stated that clients need better chips to generate more revenue from their infrastructure, not cheaper alternatives [1][2]. huang asserted that competitors’ chips do not compare to nvidia’s hopper design, and the blackwell platform is 40 times more powerful [1].
future strategies and economic outlook
huang highlighted that nvidia is receiving significantly more orders for blackwell than hopper [1]. these orders are primarily from cloud providers, with further spending expected from other companies in ai data centers [1]. he suggested that in an economic recession, companies might increase investments in ai equipment as a growth area [1][2]. huang stated that proposed tariffs would have minimal short-term impact, as nvidia is shifting manufacturing of key product lines to the us [1][3]. nvidia is utilizing tsmc’s arizona plant and plans to increase usage as tsmc expands its capacity [1][2][3].