Chinese AI startup DeepSeek burst on the scene this week with its latest AI model, which the startup claims performs as well as leading AI from OpenAI, Meta, and Anthropic — but at a far lower cost to develop. DeepSeek said their total training costs amounted to just $5.576 million — much cheaper than the $100 million to $1 billion Anthropic CEO Dario Amodei says it costs U.S. startups to train AI today.
However, in earnings calls on Wednesday, leaders at Microsoft and Meta separately affirmed their plans to continue to spend heavily on AI, even though DeepSeek showed that it was possible to spend less and still develop a competitive AI model. Executives said spending on AI has already led to business gains and spending more on AI is necessary to stay competitive in the long term.
Related: OpenAI Says AI Industry Disruptor DeepSeek May Have Copied Its Work as Rivals Race to Catch Up
Microsoft says AI revenue is up
Microsoft CEO Satya Nadella said on the company’s earnings call that its AI business was up 175% in revenue year-over-year, for an annual revenue run rate of $13 billion. Overall revenue at the company was up 12% from the previous year, reaching $69.6 billion.
“As AI becomes more efficient and accessible, we will see exponentially more demand,” Nadella predicted.
He highlighted that Barclays, the University of Miami, and Pearson used Microsoft’s AI for 10,000 or more users for the quarter ending December 31.
Satya Nadella. Photographer: Dimas Ardian/Bloomberg via Getty Images
Microsoft stated earlier this month that it would spend $80 billion on AI data centers in the fiscal year ending in June. Microsoft’s chief financial officer Amy Hood said on the call that more than half of the company’s AI-related spending was “on long-lived assets that will support monetization over the next 15 years and beyond.”
Meta says AI investments give it an advantage
On Meta’s earnings call on Wednesday, CEO Mark Zuckerberg defended plans to spend up to $65 billion on AI this year, stating that “investing very heavily” in AI infrastructure “is going to be a strategic advantage over time.”
“At this point, I would bet that the ability to build out that kind of infrastructure is going to be a major advantage for both the quality of the service and being able to serve the scale that we want to,” he said.
Meta’s chief AI scientist Yann LeCun stated in a Threads post on Saturday that DeepSeek was able to succeed because of open-source AI models, which allow developers to build on each other’s work. Meta famously made its AI open source in 2023 after spending millions developing it, and DeepSeek used parts of it to create its latest model.
Related: Meta Is Building AI That Can Write Code Like a Mid-Level Engineer, According to Mark Zuckerberg
Meta made $164.5 billion in the twelve months ending on December 31, a 22% increase from 2023.
Is the DeepSeek model the future of AI spending?
With DeepSeek topping U.S. app stores, even ahead of ChatGPT, the competition will be watching to see if massive spending is worth it.
DeepSeek’s popularity has already rattled tech stocks, causing AI chipmaker Nvidia to lose $590 billion in market value in one day on Monday.
“DeepSeek has shown that innovation doesn’t need a trillion-dollar price tag,” Lukman Otunuga, senior market analyst at global broker FXTM told Entrepreneur in an emailed statement. “If US tech leaders fail to convince investors of their edge, AI stocks could face further pressure this week.”
