In This Article:
China's DeepSeek recently published a research paper that shocked Wall Street. The startup company claims it spent less than $6 million to train an artificial intelligence (AI) model whose performance matches or nearly matches that of leading U.S. models. Comparatively, OpenAI spent more than $100 million on its GPT-4 model.
Nvidia (NASDAQ: NVDA) stock declined sharply on the news. Its market value fell nearly $600 billion in a single day, the largest daily loss by any listed company in history. The logic behind the crash is straightforward: If DeepSeek built a good AI model for less money, U.S. companies can use the same training methods to achieve similar efficiencies.
Consequently, investors are worried U.S. companies would spend less than anticipated on Nvidia GPUs, which are usually the most expensive line item in AI infrastructure budgets. But the market may have overreacted. Nvidia shareholders just got good news from Meta Platforms (NASDAQ: META) and Microsoft (NASDAQ: MSFT), two of its largest customers.
Read on to learn more.
Good news from Meta Platforms: CEO Mark Zuckerberg views heavy investments in AI as a strategic advantage
Meta Platforms CFO Susa Li told analysts on the fourth-quarter earnings call that capital expenditures would increase as much as 66% to $65 billion in 2025 to support its generative AI efforts and core business. That represents a material acceleration from the 39% increase in capital expenditures last year.
CEO Mark Zuckerberg also provided context, telling analysts Meta's ability to spend heavily on AI is going to be a "strategic advantage" over time. He also said more efficient training methods do not necessarily reduce the need for AI chips. Instead, he sees recent breakthroughs as an opportunity to apply more computing power to inference workloads to "generate a higher level of intelligence and a higher quality of service."
Good news from Microsoft: CEO Satya Nadella says demand will increase exponentially as AI becomes more affordable
Not to be outdone by DeepSeek, Microsoft CEO Satya Nadella said on the latest earnings call: "We ourselves have been seeing significant efficiency gains in both training and inference for years now. On inference, we have typically seen more than 2x price-performance gain for every hardware generation and more than 10x for every model generation due to software optimizations."
However, Nadella thinks the consequences will be favorable for Nvidia. "As AI becomes more efficient and accessible, we will see exponentially more demand," he told analysts on the call. Nadella also posted on X: "Jevons paradox strikes again! As AI gets more efficient and accessible, we will see its use skyrocket."