Big Tech’s earnings roll call kicks off later this afternoon, with Microsoft (MSFT) and Meta (META) set to announce their results after the bell. If you’d asked anyone last week about the biggest thought on Wall Street’s mind heading into earnings, they likely would have given you a number of responses ranging from the impacts of the new Trump administration to capital expenditures.
That all changed Monday when the market went into full meltdown over DeepSeek AI. The China-based company sent Wall Street into panic mode over claims that its latest AI model, DeepSeek-R1, matches the performance of models produced by Silicon Valley’s biggest tech names, including OpenAI and Meta, for just a fraction of the cost.
Now the big question on investors’ minds is how Big Tech will respond to what could mark a titanic shift in the ever-changing AI space. It’s not only a question of how companies like Microsoft, Meta, Google (GOOG, GOOGL), and Amazon (AMZN) are spending the billions they’ve poured into developing AI models, but whether Nvidia (NVDA), the biggest winner of the AI explosion, is wildly overvalued.
Wall Street already hammered Nvidia Monday, sending shares plummeting and wiping out nearly $600 billion from its market cap. But it’s not all doom and gloom for the AI trade. Some analysts see DeepSeek as a net positive for the AI industry.
In a recent investor note, Bernstein analyst Stacy Rasgon said that if DeepSeek’s claims are correct and it’s managed to improve overall AI training efficiency, tech companies will likely see an increase in demand for AI technologies and that any computing power saved by DeepSeek’s approach will end up being used up to meet that new demand.
Rasgon also points out that it’s highly unlikely that competing AI firms were unaware of the techniques DeepSeek used to develop its latest model. That’s not a stretch either. DeepSeek announced its DeepSeek-V3 model in December and R1 on Jan. 20.
DeepSeek-V3 is the model the company said it spent just $5 million to train, compared to the estimated $100 million other companies have spent training their models. Despite knowing about both V3 and R1, Meta went ahead and announced its plans to spend upwards of $65 billion in 2025 on AI.
Microsoft, for its part, said that it will spend $80 billion on its own AI buildout in the company’s fiscal 2025, though it made the announcement in early January.
Other analysts still have questions about DeepSeek’s claims, with Truist Securities analyst William Stein saying it’s difficult to confirm the company’s statements as to the number and type of chips it used to train its models and how long training took.
“We highlight DeepSeek’s indication that its cost claims do not include ‘the costs associated with prior research and ablation experiments on architectures, algorithms, or data,” which may make DeepSeek’s training expense claims not properly comparable to those from ChatGPT and others,” Stein wrote.
Lost in all of this is the fact that DeepSeek, like Meta, is offering its models as open-source software, which means companies around the world can access, use, and adjust it to their liking.
That’s a massive benefit to the broader AI industry, as it helps smaller companies build out their own AI offerings over time, increasing competition and forcing ever more advances in the space.
Now it’s up to Silicon Valley’s tech giants to explain just that.
Email Daniel Howley at dhowley@yahoofinance.com. Follow him on Twitter at @DanielHowley.
Read the latest financial and business news from Yahoo Finance