Smaller software stocks have been on a tear this year, and the sector broadly stands to benefit from lower costs for AI

Software stocks have beaten chip stocks this year — and more dramatically so over a six-month span.

Move over, semiconductor stocks. After a hot run for big software stocks, now smaller ones are getting in on the party — and there could be more room for gains across the whole sector.

While chip stocks have been on a zig-zag pattern so far this year — in the face of post-DeepSeek fears of an artificial-intelligence infrastructure overbuild, as well as worries about tariffs — software stocks are looking like a more stable bet to many. These companies and their customers stand to benefit from the DeekSeek developments, especially if computing costs decline.

Software stocks are performing slightly better versus semiconductors so far in 2025 — but over a six-month period, the shift is far more dramatic. The iShares Expanded Tech-Software Sector ETF (IGV), which tracks the performance of the S&P North American Expanded Technology Software Index, is up about 27% in that span. Meanwhile, the iShares Semiconductor ETF (SOXX), which tracks the ICE Semiconductor Index, shows how chips are lagging behind, with growth of less than 1% over the last six months.

Including more software stocks, or even swapping chips out for software in portfolios, is clearly happening. “It is a conversation being had at the institutional level,” said Brendan Connaughton, founder and managing partner of Catalyst Private Wealth, a financial advisory firm. “Software has a level of consistency” thanks to its “more reliable, consistent revenue growth,” he noted. “With chips, DeepSeek threw a curve ball into it.”

That consistency of licensing revenue also means possibly slower, but not as bumpy, growth. Plus, software companies are testing out new revenue streams that would allow them to charge for each time customers actually use their new AI features.

And since software is delivered via the cloud and resides inside the cloud, corporate data centers or on local systems, it is not a physical product that needs to be manufactured and shipped from countries outside the U.S. — and thus not generally subject to tariffs.

In the fourth-quarter corporate earnings season so far, software revenue growth is 12.15%, according to FactSet’s data on S&P 500 companies. On the hardware side, semiconductor and equipment revenue is up 11.34% in the fourth quarter, based on companies that have reported to date.

Nvidia Corp., which reports this month, will surely distort that average, with revenue for its fiscal first quarter forecast to grow 72%. When taking into account a combination of reported results and estimates for companies that haven’t yet posted numbers, semiconductor components of the S&P 500 are expected to see a 24.6% blended growth rate, versus 11.7% for software components.

That said, many software companies that have been particularly hot lately are smaller, faster growing and not part of the S&P 500. For example, project-management software company Monday.com Ltd. saw 32% revenue growth in the latest quarter, ahead of expectations, and its stock rose 27% in the session following its earnings. Data-streaming platform Confluent Inc. also beat estimateswith its 23% revenue growth, and its stock rose 25% in the first session following results.

“Software remains the most favored sector on the buy side in terms of looking for new long ideas that look and feel derisked,” said Jordan Klein, a managing director and TMT-sector specialist at Mizuho Securities, in his sales newsletter to clients.

He noted that while larger software stocks aren’t quite “working” for investors right now, investors have been hungry for small-cap and mid-cap software plays. “If you do not own enough of these names, you are probably not going to be outperforming broader tech and/or your key benchmark,” Klein added.

Whether or not software will keep its vaunted status right now is a big question. With most investors still focused on AI and its impacts on corporate spending, as well as its promise for corporate efficiency, software has not quite had the same kind of AI moment as physical hardware yet. And the massive capital-spending plans announced by big cloud and hyperscaler companies so far could deliver further revenue boosts to chip players.

“I don’t want to throw the chips in the trash can just because they had a little pothole,” Connaughton noted. “I think there is a marrying of both of those. Ultimately it comes down to names and companies that can execute in areas of high growth. Corporate software’s addressable market is huge.”

But if DeepSeek’s lower-cost revelation tells investors anything, software-application developers could come to benefit, said Patrick Walravens, head of technology equity research at Citizens JMP.

“The apps providers are doing better than the infrastructure providers,” Walravens, who covers 37 software companies, told MarketWatch. The bigger implication of the DeepSeek news is that the lower cost of compute will be beneficial. “If you are doing inference and using these [large-language models] to come up with answers and drive your products, your prices are going to go down,” he said, referring to the prices companies pay to use software that taps into these models.

Inference refers to the actual action of trained AI models — either a prediction, a statement or its own conclusions. Citing DeepSeek, Palantir Technologies Inc. Chief Technology Officer Shyam Sankar recently observed that “the price of inference is dropping like a rock.”

The next AI wave could involve more software purchasing — including software for data centers, both corporate and cloud, and AI-specific data centers that are being built now, especially the application layers.

“Our checks indicate improving demand-pattern inputs across software, including data/analytics and SaaS, and with cybersecurity remaining strong,” Gregg Moskovitz, an analyst at Mizuho Securities, wrote in a note to clients. He added that AI uptake, including the use of agents, is picking up. “We heard several reports of even more activity around customers readying their data estates for the AI wave,” Moskovitz said.

In addition, some software companies are trying to get leaner, a feature always endorsed by Wall Street. For example, just in the last two weeks, Okta Inc., Workday Inc. and Salesforce Inc. all saw some stock momentum after announcing rounds of layoffs — although Salesforce has since lost much of that boost, after the news last week that Chief Operating Officer Brian Millham is retiring in May. He will be replaced by Salesforce board member Robin Washington, in a newly created role as president and chief operating and financial officer.

Some midtier software companies like Snowflake Inc. and Confluent have already seen big surges in their share prices this year, with Snowflake up 21% and Confluent up 25%. But some larger-cap stocks could see renewed momentum after falling earlier this year, including ServiceNow Inc. and Oracle Corp.

Further upside could also come as more proof emerges that AI can actually deliver cost savings or other financial benefits. Some investors may be concerned that many software companies price by seat licenses — so if customers need fewer workers because AI tools become so helpful, they may need fewer software licenses.

“On one hand, you might worry the software companies can only charge half because their customers are using half the seats,” Walravens said. But companies are now adding consumption pricing to their contracts, he noted. “Every time you deflect a [call-center] conversation using AI, we will charge you a $1,” he said, as an example.

That could become a lucrative model as AI adoption becomes more widespread.