Claude Cowork the newest fuel for an AI-driven de-rating of software stocks
Software stocks are in the wilderness: fears of disintermediation by AI mean it’s difficult to think of them as growth stocks going forward, but they’re not necessarily cheap enough to be considered value stocks, either.
Software companies started off 2026 with a record underperformance of chip stocks.
Things haven’t gotten any better since.
The iShares Expanded Tech Software ETF is off more than 4% year to date, with most of the stocks in the fund showing a discouraging trend: just 31% are trading above their 200-day moving average.
The launch of Claude Cowork by Anthropic, which was mostly built using its Claude Code tool, has reignited traders’ desire to get out of software stocks for fear that they’ll be disintermediated by AI tools and agents.
A smattering of formerly high-flying, highly valued software companies have suffered significant valuation compression to converge around enterprise value-to-estimated sales ratios of less than 5.
(Many thanks to modestproposal1, a member of my finance twitter Mount Rushmore, for bringing this to our attention.)
To modify “Anna Karenina,” each member of this software family is unhappy in a similar way, despite being very different when it comes to top-line growth, margins, market caps, or the customer needs their businesses address. Nevertheless, they’ve all arrived at essentially the same valuation destination by way of a unifying cause.
The business prospects of established software firms have taken a hit because of the ease with which AI agents are able to develop software and handle the tasks and processes that served as the core value proposition provided by these companies. Or more simply: if the marginal corporate dollar goes directly to AI, rather than software or labor, it makes sense that investment dollars would follow, too.
“First, the arrival of truly capable AI agents is no longer a 2027 or 2028 story, it’s happening now. The timeline has collapsed. Second, the classic ‘build versus buy’ calculation that has governed enterprise software decisions for decades has been fundamentally altered,” Jordi Visser of 22V Research wrote in a note from January 7. “When a domain expert can build sophisticated technical systems in hours rather than months, the economics of custom development versus off-the-shelf solutions shift dramatically.”
For investors who primarily hold broad market ETFs, this state of affairs is more than a bit annoying: many of the seemingly disrupted are multibillion-dollar market caps in popular benchmark indexes, while the disruptor, in the case of Anthropic, isn’t publicly traded.
Typically, software firms trade at higher valuations than chip companies because they’re asset-light, historically higher-margin, and tend to generate high amounts of reliably recurring revenues, whereas semiconductor companies are subject to the whims of volatile manufacturing cycles. But the sell-off and concurrent de-rating of software stocks leaves the parent index for the iShares Expanded Tech Software ETF near a similar valuation as the Philadelphia Semiconductor Index. That’s a signal about the perceived strength of this trend: investors are increasingly willing to pay up for the products that lay the foundation for the potential disintermediation of software companies, rather than these sticky revenue generators.
(Note: This software fund also counts both AI software beneficiaries like Palantir, D-Wave Quantum, and some crypto treasury companies as some of its most richly valued members, most of which I would struggle to call software stocks.)
While valuations for many software companies are cheap relative to their history, they still trade at a significant premium to the S&P 500 on most valuation metrics (including EV to sales). Therein lies the rub: a growth stock that’s sold off is not a value stock. It’s a stock left to wander the wilderness.
“For those trying to buy software because they are cheap and fade semiconductors, I think people are missing what these engineers have said this year,” Visser added. “The competitive moats around enterprise software businesses begin to look dangerously shallow in this world.”
The bull case for software? Well, for starters: this bear case, and the fact that everyone’s seemingly betting on these negative trends to persist. Positioning data from Morgan Stanley suggests that institutions hate software stocks at the moment.
Even if AI eats software, it’ll have a lot of chewing and digesting to do along the way, since software’s already eaten the world first.
