Software: Selectivity Required

Technology have welcomed AI as a new growth engine, with both the NASDAQ Composite Index and the S&P Information Technology Index reaching fresh all-time highs, driven by technology giants, semiconductor manufacturers, and cloud service providers.

However, many software companies — particularly those in application software — have experienced a very different trajectory, influenced by AI. While the decline in IT spending amid global uncertainty has been a contributing factor, concerns over AI disrupting their business models have been an arguably unjustified force weighing on share prices.

The question now: has the sell-off overshot fundamentals?

The Disruptor Gets Disrupted

Now the great disruptor is getting disrupted. The emergence of artificial intelligence has inspired similar euphoria, and some investors are bracing for a significant portion of the software industry to become the ones being eaten — sparking the phrase, “AI is eating software companies.” These firms could face a fate similar to Dell Technologies Inc. (DELL) in the 2010s, when the PC maker’s on-premise hardware business was overtaken by cloud infrastructure. The market is already separating winners from losers, with shares of software giants such as Salesforce Inc., Adobe Inc., and ServiceNow Inc. among the worst performers in the S&P 500 this year, each down at least 17%.

If you take out the three largest components of the software index — Microsoft Inc, Palantir Technologies Inc and Oracle Corp, all of whom are now seen as AI icons rather than software firms — the rest of the software subsector index would be roughly flat year to date compared with the broader market, which is up 10%.

Figure 1: Software has been the drag

Why software as a service (SaaS)?

We believe that software companies are the early innings of this AI disruption wave. When software as a service (SaaS) first emerged 25 years ago, it revolutionized software by moving it to the cloud and speeding up feature delivery. Now, a fresh discontinuity is at hand. Generative and agentic AI—tools that can reason, decide, and act—are already:

·       drafting code in Cursor's AI code editor;

·       handling support tickets in ServiceNow;

·       preparing journal entries in Workday Financial Management; and

·       writing ad copy in Adobe’s Experience Cloud.

In the future, any routine, rules-based digital task could shift from a “human plus app” model to an “AI agent plus application programming interface (API)” model. While many SaaS companies are rolling out AI-enabled features — such as Salesforce’s AgentForce, Adobe’s Firefly and Intuit’s TurboTax — these efforts have not been sufficient to reassure investors. AI risks undermining the core value proposition of these firms, which is to provide customers with premium-priced digital tools that enhance productivity. If cost-conscious clients, such as banks or retailers, can access virtually the same services at a fraction of the cost through generative AI, entire business models could collapse. This helps explain why companies like Adobe have seen their share prices react negatively despite reporting strong quarterly earnings.

This gives rise to the two main challenges facing the industry: cost and competition — both of which present significant threats to long-term profitability. The greatest concern is that intense competition will spark a race to the bottom on pricing, particularly within the Software-as-a-Service (SaaS) sector. As AI reduces the need for large teams of software developers to operate a business, it lowers the barriers to entry, allowing more players to enter the market and intensifying the battle for market share. This competitive pressure to cut prices could erode margins, making it increasingly difficult for companies to sustain growth and invest in innovation. Such a price war could ultimately commoditise SaaS offerings, forcing firms to differentiate themselves through value-added features, superior customer service, or ecosystem integration rather than price alone. Those who fail to adapt risk being left behind.

Figure 2: Scenario of potential AI impact

Source: Bain Analysis.

AI disruption is uneven

The Rule of 40 is a financial metric primarily used for SaaS and other high-growth technology companies, serving as a measure of business growth efficiency. At present, close to 25% of companies within the S&P expanded software subsector pass the Rule of 40 — a group largely concentrated among larger-cap players. Many mid- and small-cap software companies fail the test, primarily due to weak EBITDA margins.

Looking ahead, we expect generative AI to have varying levels of impact across the sector. All software categories are likely to experience some degree of value creation, switching, and value erosion, which could, in turn, affect how companies perform against the Rule of 40. However, the extent of these impacts will differ by category.

For example, as shown in Table 1, while AI could significantly disrupt customer service software through automation, categories such as CRM and Enterprise Resource Management (ERM) — which are deeply embedded in core enterprise workflows and house substantial enterprise or industry data — may experience a milder impact. These segments could even benefit from AI through workflow streamlining and the ability to leverage proprietary data and insights.

That said, while we expect the overall impact across the software industry to be uneven — with some subsectors, such as CRM/ERM and application development, facing lesser disruption — all of this circles back to the central question the market has yet to answer: cost. Traditional B2B SaaS companies typically enjoy gross margins of 70–90%, whereas AI SaaS businesses currently operate at gross margins closer to 30–50%. Even Anthropic’s Claude, one of the most advanced AI models, achieves gross margins of only around 55%. This raises the key question: even if AI can replicate only part of the existing business model, how many customers will be willing to pay a premium for the additional features?

As such, our current view remains that generative AI will disrupt the overwhelming majority of the software market, though its relative impact will vary across industry categories and individual companies.

Figure 3: Most Software companies failed to pass the rule of 40

Table 1: Impact of AI in different software categories

Look out for opportunities in large players – Don’t underestimate their strengths

While we expect disruptions across the broader software subsector, we believe there will still be winners. With that in mind, we have identified several criteria that could help potential winners weather the current storm: 1) Large-cap players with a strong presence in the industry 2) A durable moat that encourages customers to continue paying for their services 3) Proprietary databases capable of delivering more accurate and personalised results than generic AI tools.

For instance, Salesforce — one of the largest CRM companies — could see parts of its business model disrupted by AI, including basic CRM functions (such as storing contacts and tracking leads), workflow automation, analytics, and forecasting. However, the company’s key moat lies in its ecosystem: its enterprise integrations are deeply embedded in client workflows, making it challenging for AI to fully replicate or displace its offering.

In the design space, as AI capabilities extend into image generation and editing — with tools like ChatGPT’s GPT-4o and Google’s Nano Banana — incumbents such as Adobe, Figma, and Canva could face competitive pressures. While many consumers may switch to free AI tools for casual image generation, we do not expect businesses to do the same, as free tools often carry significant copyright risks. Adobe, by contrast, offers users legal protection when creating AI-generated content through its products, which could help sustain enterprise demand.

Despite ongoing challenges, we see potential tailwinds for the subsector in the remainder of the year. With global tariff uncertainties easing, US companies may resume IT spending, which had been reduced due to trade-related uncertainty — a development that could lift sentiment for software firms, many of which are exposed to mid- and small-cap businesses. Small businesses are increasingly more optimistic about the economic outlook, which should bode well with spending. Furthermore, we expect companies to begin benefiting from tax savings under the One Big Beautiful Bill Act, providing additional support to earnings growth.

Figure 5: Small Business Confidence Index

Figure 6: More businesses application should increase needs for software

The AI vs Software war is expected to continue, with future developments depend on how the companies react

Turning foes into friends. Incorporating generative AI into established products could significantly enhance the experience and productivity of traditional software. Typical applications include the introduction of natural language interfaces or agentic co-pilots to streamline existing workflows.

US software firms have spent US$33.8 billion on AI acquisitions this year. Notable deals include ServiceNow buying Moveworks which sells AI assistants for US$2.85B, NiCE acquiring conversational AI chatbot maker Cognigy for US$955M, Salesforce purchasing lead generator Bluebirds and supply chain automation provider Regrello for a combined US$1.1B, Workday acquiring recruiting assistant Paradox for US$1B, and Databricks buying Neon (US$1B) and machine-learning application start-up Tecton (US$900M).

New pricing model. One way to address the pricing premiums charged by software companies is through usage-based pricing. While subscriptions remain the dominant model today, they may not align well with AI agent deployments, as buyers increasingly seek pricing structures that reflect tasks, processes, and utilisation rather than a per-user charge. In other words, users would pay only for what they consume — light users paying less, making their costs comparable to AI models, while heavy users pay more for advanced offerings that cannot currently be replicated by general AI tools.

Workforce reduction. As subscription sales slowdown, reducing costs will likely by prompted. The lack of new employee growth in their end customers and the swath of layoffs in the tech sector will likely pressure seat growth for the industry, in addition to squeezing discretionary IT spending. In addition, we expect cost cutting to center on tech-related jobs with fewer programmers needed, given the increased adoption of coding copilots. For example, buy now, pay later firm Klarna has reported that it is employing fewer people because of its widespread use of generative AI, or Gen AI.

Neutral on Software; Subsector in a crosshair

In short, for Software as a subsector in technology, price actions have been harsh, but fundamentals/moats suggest medium-term resilience for select names.

Overall, we believe AI presents a structural challenge to the software subsector in terms of margins and competition, warranting a discount to fair PE multiples. However, the sector’s weak year-to-date performance suggests that some quality companies may now be trading at attractive valuations.

As we are Neutral on software overall where margins and competition argue for a valuation discount, investors should be selectively constructive on large-cap platforms with workflow/data moats (e.g., Salesforce, Adobe) and on vendors that embed agentic AI without over-eroding price.

As such, for investment recommendations, we continue to prefer broader digital economy sector exposure through the Invesco Nasdaq Internet ETF (NASDAQ: PNQI), which we see an upside of 35% based on current price levels.

Despite having around 20% exposure towards the software subsector, quality large-cap players like Microsoft, Salesforce and Adobe already made up close to 15%, which aligns with our view of a large-cap tilted exposure.

Table 2: Valuation Table for PNQI ETF