It’s wartime, not peacetime for software
of this year Morgan Stanley The lineup for the Technology, Media and Telecommunications conferences is stacked: Dario Amodei and Sam Altman (both at the center of the Pentagon drama), Jensen Huang, Satya Nadella and dozens of enterprise software CEOs are trying to convince investors that they will survive the AI computation.
I spent a day researching this topic and then sat down with David Chen, Morgan Stanley’s head of global technology investment banking, to explain what he heard.
Chen said the biggest change from last year was the nature of the questions. In 2025, companies were talking about using AI to cut expenses, such as saving a few percentage points on operating costs with co-pilots and automation tools. Chen said this is table stakes now.
“I don’t think investors really want to hear about how people are becoming more efficient with AI,” he said. “Are you really, truly the beneficiary, or is AI threatening your overall business?” They wanted to hear the question.
And to their credit, companies have been more direct about answering that question this year. Especially enterprise software companies that have seen trillions of dollars of market cap evaporate in just one week this year.
Chen drew a line between companies whose software does something deterministic, like calculating payroll, sending invoices, etc., where making a 2% error is a real problem, and companies that essentially organize public data and put it behind a pretty interface.
He argued that there was still a moat in the first one. The latter is in serious trouble.
“AI doesn’t kill software,” Chen said. “He’s mixing it up again.”
But he didn’t sugar coat it either. For companies on the wrong side of this line, he described this as “wartime, not peacetime.” And he made an interesting observation about leadership: In this environment, boards are starting to favor product-focused CEOs over sales and marketing types. If you need to reinvent your company’s backend to be AI-specific, you want someone who understands the architecture, not just the pipeline.
CNBC producer Jasmine Wu coined a phrase that captures the shift nicely: We’ve moved from SaaS (software as a service) to SaaaS (agent-oriented software as a service). An idea that emerged from my conversation with Box CEO Aaron Levie earlier in the week. He told me that the agencies now have a new client base and he can see the business becoming 10 times bigger than the current one. This means a lot: The software that survives won’t be the software people use. This will be the software that agencies use.
Regarding the development of artificial intelligence, Chen was asked whether infrastructure spending would be higher or lower in 2027. His answer was: “Probably a similar level.”
If that’s the case, at least the hyperscalers’ AI capex cycle may be nearing its peak.
Chen predicts a rebalancing between winners and losers in enterprise software for the coming year; cybersecurity stands out as a segment that has all the characteristics of a competitive moat and is a clear beneficiary of AI rather than being a victim.
He also pointed to a new wave of companies focused on solving bottlenecks in connectivity, computing and energy that constrain AI development in semiconductors and systems.
Summary of the whole week: AI has moved from the idea that this is going to be big to realizing that this is already big, so show me you embrace that.
