How has software sentiment evolved this year?
Matthew Hedberg: The software space has been one of the most interesting sentiment roller coaster rides that I've seen in my career. Back on April 10, the IGV software index was down about 27%, which was really the height of the so-called SaaSpocalypse, where investors were asking: if AI can write code, build apps, and automate workflow, what exactly is the value of traditional software?
Since then, the recovery has been pretty remarkable. The IGV software index has rallied back to essentially flat to the year. The conversation has shifted from AI is going to kill software to AI is fundamentally going to change what it means to be a software vendor.
"The conversation has shifted from ‘AI is going to kill software’ to ‘AI is fundamentally going to change what it means to be a software vendor.’"
Matthew Hedberg, Head of Global TIMT Research, RBC Capital Markets
I characterize sentiment today as cautiously optimistic in subsegments of software such as cyber, infrastructure, data, and AI, rather than universally bullish on the entire space. Data is becoming more valuable because AI is only as useful as the data it can access, and investor enthusiasm reflects that trend.
We've seen AI-native vendors raise substantial funding rounds at high valuations, while others have been acquired by strategic buyers. In areas such as CRM, HR and finance, investors are still trying to understand what happens when AI agents can perform tasks that historically required humans to interact directly with these applications.
Does that reduce the need for software or simply change how software is priced, particularly under traditional seat-based models? The bearish case was that AI would disintermediate SaaS entirely. The more constructive view is that AI becomes another layer on top of the existing software stack, making applications more intelligent, automating workflows and ultimately expanding what software can do. The opportunity for software may not be smaller; it may be larger, but it could look very different from the market we've known over the past decade.
What determines AI winners and losers?
Rishi Jaluria: When we decide AI winners and losers, it's important to think about both magnitude and duration. Ultimately, the most important moat is innovation. The companies that have a strong track record of innovating organically are likely in the best position to benefit from AI long-term.
We expect M&A activity to accelerate, driven by both private equity and large platforms acquiring AI-native companies. We view these deals as outsourced R&D, with execution determining value creation.
We're also watching for business model transitions. SaaS has long relied on predictable, seat-based pricing, but AI is driving a shift toward consumption and outcome-based models. If AI makes a salesperson five times more productive, vendors need new ways to monetize that value beyond charging more per seat. Our framework is a 60/30/10 mix: 60% subscription, 30% consumption and 10% outcome-based pricing. While potentially disruptive in the near term, these shifts could strengthen business models over time.
"The companies that have a strong track record of innovating organically are likely in the best position to benefit from AI."
Rishi Jaluria, U.S. Software Analyst, RBC Capital Markets
What are the cybersecurity implications of recent AI incidents?
Hedberg: As AI becomes more pervasive, we're seeing the increase in cyber threats that many expected.
Thomas Wolfe called the recent Mythos AI cyber attack on Hugging Face a wake-up call, and IBM reported a 56% increase in AI-driven attacks. We think recent breaches are just the start of a multi-year tailwind around security modernization.
Ultimately, we think this trend largely benefits cybersecurity consolidators versus point-based vendors.
What stood out from the software and AI bus tour?
Jaluria: This year’s software and AI bus tour reflected how deeply embedded AI has become across the software ecosystem. We met with 15 public software companies, as well as several VCs and private companies, and a few themes stood out.
First, software moats remain intact. Data infrastructure and systems of record continue to be powerful competitive advantages. Enterprises are buying more than outcomes; they are paying for governance, security, compliance, reliability and trusted implementation.
Second, incumbents still have advantages. Large software vendors benefit from established customer bases, proprietary data and strong cash flow. Those strengths matter, but maintaining leadership will require continued investment and innovation.
Third, AI is expanding the software stack, not replacing it. Frontier models are unlikely to own every software vertical. Distribution, domain expertise and specialized workflows will continue to support a broad software ecosystem. As AI and agentic tools move from experimentation to deployment, however, companies increasingly want pricing tied to measurable value creation. The challenge is determining what constitutes an outcome and how that value should be attributed across the software, the AI and the human user.
"Software moats remain intact. Data infrastructure and systems of record continue to be powerful competitive advantages."
Rishi Jaluria, U.S. Software Analyst, RBC Capital Markets
What does the DIY framework suggest about buy versus build?
Hedberg: We built a comprehensive framework to determine whether it's cheaper to buy or build software from vendors. We've applied this model to 14 companies in our coverage and found the DIY approach doesn't always save money.
Jaluria: The model we built looks at total cost of ownership (TCO), including significant hidden costs like maintenance, security risks, and the quality gap. The model shows that when you embed all of these costs of DIY systems, they very rarely come out meaningfully cheaper than buying directly from vendors.
Now there are some caveats here. If a company can use AI to build a better product than the incumbents, then maybe it's actually worth the extra cost if that unique solution is a competitive differentiator.
But we don't believe that most companies have that level of expertise, especially when we're talking about horizontal solutions like a CRM or an HR system. Creating and maintaining custom applications still requires talent, oversight and ongoing investment, which is why many organizations continue to rely on packaged software for core business functions.
This model doesn't capture opportunity costs, which is probably the biggest hidden cost. If a company is using its best and brightest AI talent to replace existing software and save costs as a line item on a budget, should that talent instead be used in areas where the company has real domain and industry expertise?
We also acknowledge DIY isn't all or nothing. Companies could use AI in certain areas to replace specific licenses or use it as a bargaining chip on renewals rather than a total rip-and-replace exercise.
