The market turmoil came after DeepSeek unveiled an AI tool that appears capable of competing with those created by top U.S. companies, but at a fraction of the cost. The development calls into question the wisdom of the massive AI spending wave that has swept through Silicon Valley in recent years. Analysts at Bank of America, however, say that lower costs to run AI models will benefit the software sector, which could also accelerate investment throughout the broader tech ecosystem.

“Over the longer term, we expect advances at the model level to drive accelerated enterprise AI adoption and usage as chatbots, copilots, and agents become simultaneously smarter and cheaper,” BofA’s Brad Sills, Andrew Moss, and Alkesh Shah wrote in a note Tuesday, “a scenario known as Jevons paradox.”

Microsoft CEO Satya Nadella also cited that economic theory on Monday. When the company reports quarterly earnings after the bell on Wednesday, he is widely expected to be asked about whether DeepSeek’s success carries any implications for his company’s AI spending going forward.

Jevons paradox strikes again! As AI gets more efficient and accessible, we will see its use skyrocket, turning it into a commodity we just can't get enough of. https://t.co/omEcOPhdIz

— Satya Nadella (@satyanadella) January 27, 2025

Brian Mulberry, a director and client portfolio manager at Zacks Investment Management, said Big Tech CEOs must choose whether to legitimize DeepSeek as a threat.  

“What they say or, maybe even more importantly, what they don’t say about the future could be very important for the health of those companies going forward,” he noted.

Dave Wagner, head of equity and a portfolio manager at Aptus Capital Advisors, believes CEOs will be measured on their earnings calls. He also cited Jevons paradox as a reason why Big Tech won’t feel pressured to change their capital expenditure trajectory—or cut down on buying GPU chips.

“If I was a management team, why would I make comments in still a period where we need to be researching what’s going on,” he said. “[And] where we don’t have all the answers.”

Why Microsoft benefits from lower AI computing costs

There’s plenty of fog surrounding DeepSeek’s accomplishments, and in particular its claims about developing its new model for under $6 million and using solely reduced-capacity chips—claims that appear to be misleading or untrue. BofA, though, said that discussion misses the bigger picture, at least when it comes to software.

“We expect qualitative commentary on ramping enterprise AI adoption during 1H25 earnings to evolve into quantitative indications of incremental revenue in 2H25,” Sills, Moss, and Shah wrote, “before monetization becomes meaningful in 2026/2027.”

A separate note argued this is also a big win for Microsoft. The company’s Azure platform, which accounts for almost a quarter of the cloud computing market, is already seeing a third of its growth generated by running ChatGPT workloads via a partnership with OpenAI.

“If DeepSeek is capable of generating efficiencies at lower scale, this would bode well for Microsoft’s ability to continue lowering the cost of AI computing and drive scale on capital expenditures,” Sills and Carly Liu wrote Tuesday.

Microsoft Copilot, meanwhile, is an example of agentic AI, or models that run digital agents capable of enhanced reasoning and decision-making. BofA already believed the trend could send software stocks soaring in 2025. Like Salesforce, Microsoft plans on having customers pay as they go for each use case, Sills and Liu noted, and lower AI computing costs could push costs lower and margins higher.

Amid all the optimism, Sills and Liu did say DeepSeek’s advancements could represent a headwind for Oracle. Thanks to a partnership with Nvidia, Oracle rents access to advanced GPUs, the chips at the heart of the AI boom, through its cloud platform. Demand for that service could decline if companies find they can meet their AI needs with substantially less computing power.