Major technology stocks plunged in Monday’s pre-market trading after Chinese artificial intelligence startup DeepSeek unveiled its AI model, which reportedly matches OpenAI‘s ChatGPT capabilities at a fraction of the cost. These developments are stirring significant concern over U.S. technological leadership.
NVIDIA Corp (NVDA) bore the brunt of the sell-off, plummeting 8.15% to $131.00. Other chip manufacturers followed suit, with Broadcom Inc. (AVGO) down 8.66%, Super Micro Computer Inc. (SMCI) dropping 8.03%, and Taiwan Semiconductor Manufacturing Company (TSM) declining by 7.66%. Advanced Micro Devices Inc. (AMD) lost 4.89%, Micron Technology Inc. (MU) saw a 6.00% plunge, and even Intel Corp (INTC) and Qualcomm Inc. (QCOM) faced smaller declines of 2.45% and 2.28%, respectively. The sell-off wasn’t limited to chip peers, as major tech entities like Microsoft Corp (MSFT) fell 3.30%, Meta Platforms Inc. (META) dropped 3.22%, and Alphabet Inc. (GOOGL) also saw declines exceeding 3%.
DeepSeek’s newly released R1 language model was open-sourced just last week, boasting performance comparable to OpenAI’s models. Significantly, it reportedly runs on approximately 50,000 Nvidia H100 GPUs, raising questions about the necessity of heavy investments in GPUs—an element fueling the market valuations of chip stocks.
Scale AI CEO Alexandr Wang characterized DeepSeek’s emergence as “a wake-up call for America” on X, calling for rapid innovation and tougher chip export controls. The impact of the announcement rippled through financial markets, with investors bracing for significant losses. The Kobeissi Letter reported expectations of over $1 trillion wiped off the U.S. market capitalization during Monday’s trading session, with Nasdaq 100 futures plunging 662.50 points or 3.2%, and S&P 500 futures down by 1.9%.
This situation is particularly precarious for Nvidia, whose chips have been considered instrumental for cutting-edge AI applications. Adding more weight to market sentiment, billionaire investor Chamath Palihapitiya pointed out on social media how DeepSeek has managed to achieve sophisticated reasoning without relying on massive supervised datasets. This, he argued, could signal transformative progress within AI technology.
On another front, AMD isn’t sitting idly by. The company has announced the integration of the DeepSeek-V3 model within its Instinct MI300X GPU. This initiative reflects AMD's commitment to advancing artificial intelligence technologies. The DeepSeek-V3 model, widely recognized as the most powerful open-source large language model, is specially optimized for enhancing AI inference, which could bolster AMD’s performance through its GPUs.
The new model features advanced architecture known as the mixture of experts (MoE) with 671 billion parameters—a deep learning innovation activating 37 billion parameters per token. The architecture is complemented by multi-head latent attention (MLA) and unique load balancing strategies aimed at enhancing training cost-effectiveness and inference efficiency. Collaborations with the DeepSeek and SGLang teams have been noted as pivotal, allowing AMD to attain peak performance on its hardware.
AMD's ROCm platform supports FP8 (8-bit floating-point), which optimizes data transfer and minimizes latency issues. The Instinct MI300X GPU stands as a cornerstone for accelerating AI processes, paving the way for developers focusing on projects involving large language models, image recognition, and natural language processing.
Despite AMD’s advancements, market reactions remain mixed. While some financial analysts advocate for investing in competitors like NVIDIA and TSMC, others, such as Barclays, have recommended acquiring AMD shares, signifying diverse perspectives on the company’s AI competitiveness moving forward.
These interconnected developments highlight the dynamic and often volatile nature of the tech stock market amid the rapidly changing AI sector. Investors are left to navigate shifting landscapes as they anticipate upcoming reports and meetings, including the Federal Reserve's discussion slated for this week.