Software's S-Curve: Why AI is Infrastructure, Not a Disruptor

The market is in a panic. Software stocks have plunged more than 25% year to date, pushing the sector into bear market territory. The trigger is a fear of an "AI SaaSpocalypse," where agentic AI tools replace the traditional software-as-a-service model. Investors worry that AI will erode the pricing power and high margins that have fueled the sector's growth. This sell-off is happening despite many companies reporting strong results, indicating the drop is a rerating based on perceived competitive moat erosion, not fundamental deterioration.

Yet this fear represents a classic short-term misreading of the technological S-curve. AI is not a disruptor poised to replace the software rails; it is the new layer of compute and intelligence being layered onto them. As HSBC argues, software will be the primary mechanism for the diffusion of AI across enterprises. The bank contends that foundation models are "inherently flawed" for replacing major enterprise platforms, which require the reliability and complexity that incumbents have built over decades. In other words, software is already eating AI.

The setup here is a classic infrastructure layer moment. Just as the internet became the foundational layer for the web, AI is becoming the foundational layer for software. The most durable vendors are those that can embed "distilled intelligent agents" into their platforms, enhancing rather than displacing them. This is the exponential growth path: AI accelerates the value of the underlying software infrastructure. The current sell-off, therefore, may be an overreaction to a flawed narrative, creating a potential entry point for those who see the true S-curve ahead.

The Infrastructure Layer: Why Software Vendors Win

The technological and economic moats protecting established software vendors are deeper than the current panic suggests. HSBC's analysis points to a fundamental flaw in the "AI will replace software" narrative: foundation models are "inherently flawed" for the high-fidelity, mission-critical platforms that run global enterprises. These models simply aren't built for the reliability, error-free operation, and complex integration required by core business systems. As a result, a "lift-and-replace" of major software suites is "not realistic" for the majority of enterprise needs.

This creates a clear winner: the incumbents who already control the critical data layer and have decades of experience building enterprise-grade software. Legacy vendors aren't just adapting; they are actively embedding "distilled intelligent agents" into their platforms. This approach uses AI to enhance, not replace, the underlying software infrastructure, overcoming the limitations of raw foundation models in a controlled, reliable way. Oracle, for instance, is embedding machine learning across its entire product suite, leveraging its dominance in the data layer where most enterprise AI will ultimately operate.

By contrast, AI-native startups lack the essential "enterprise class" software experience. As HSBC notes, these players have "little to no experience creating 'enterprise class' software" and would be "architecting from scratch in unfamiliar highly complex areas." This forces them to rebuild the fundamental rails from the ground up in environments where incumbents have deep, hard-won moats. Even if they could develop comparable code, the burden of running core operations for global companies-accountable to shareholders-creates a formidable barrier to displacement.

The bottom line is that AI is becoming the new infrastructure layer, and the vendors with the strongest foundation are best positioned to layer it on. They control the data, understand the enterprise, and are already building the tools to harness AI's power. For investors, this isn't about betting on a software apocalypse; it's about recognizing that the most durable growth path lies with the companies already building the rails for the next paradigm.

Financial Impact and Valuation Shifts

The market is pricing in a software apocalypse, but the financials tell a different story. Salesforce's latest quarter shows a company generating immense cash and returning it to shareholders, even as it navigates cautious AI monetization. The company posted 12% year-over-year revenue growth in its fiscal fourth quarter, its fastest pace in two years. More telling is the capital allocation: CEO Marc Benioff announced a $50 billion share buyback program, explicitly citing "low prices" for the stock. This is the move of a management team confident in its long-term cash flow, even as shares have fallen about 28% so far in 2026.

Yet the stock's reaction to guidance reveals the core disconnect. Despite the strong results, Salesforce's cautious fiscal year 2027 revenue guidance-projecting 10-11% growth-sparked a decline. This reflects a market skeptical that AI investments will accelerate growth as quickly as hoped. The guidance implies a slowdown from the recent 12% quarter, tempering the exponential adoption narrative that underpins the infrastructure thesis. In other words, the market is paying less for software earnings, pricing in a slower AI integration curve.

This is where the valuation disconnect becomes a potential opportunity. HSBC's analysis, which argues software will be a major beneficiary of AI development, sees this as a mispricing. The bank's ratings reflect this view: a Buy rating on Salesforce alongside Hold and Reduce ratings on other names. It believes the sector's valuations are at historical lows despite its "strong demand momentum" and "massive" expansion potential. The bank's logic is that the most durable value capture will flow to the software vendors already building the rails, not the AI-native startups lacking enterprise experience.

The bottom line is a tension between short-term sentiment and long-term fundamentals. Salesforce's cash generation and buyback program signal confidence in its durable moat. The cautious guidance acknowledges the real friction of monetizing AI at scale. For investors, the setup is clear: the market is punishing the sector for perceived AI-related growth fears, but the underlying infrastructure thesis suggests those fears are overblown. The current low valuations may be the price of admission for a sector that is, in reality, the primary engine for AI's enterprise adoption.

Catalysts, Scenarios, and What to Watch

The infrastructure thesis is now in the testing phase. The market has priced in a software apocalypse, but the coming quarters will reveal whether AI is truly being layered onto the rails or if the rails themselves are crumbling. The key signals to watch are not flashy AI product launches, but the financial metrics that show AI driving enterprise value: improvements in software margins and customer retention rates. If AI is enhancing the core product, we should see these fundamental indicators strengthen, confirming that the technology is being used to build better software, not replace it.

A critical scenario to monitor is how AI-native players respond. The HSBC analysis argues they lack the "little to no experience creating 'enterprise class' software" and would be "architecting from scratch in unfamiliar highly complex areas". The moat thesis depends on them staying focused on their core AI models rather than attempting to build enterprise platforms. Watch for any moves by major AI model players to build or acquire enterprise software capabilities. If they do, it will be a direct test of the incumbent moats. More likely, they will partner with established vendors, a dynamic that could accelerate the embedding of AI into the software stack.

The most significant near-term risk is a faster-than-expected adoption of "vibe-coding" or low-code AI tools. HSBC acknowledges that while "vibe-coding places the burden of design on the developer," a widespread shift could accelerate the perceived timeline for disruption. If these tools make it easy for enterprises to build custom, AI-generated applications quickly and cheaply, it could pressure the traditional software model. The bank notes that even if this leads to better or free solutions, displacing incumbents that run core global operations will still be "extremely difficult". But the market may not wait for that reality to set in; sentiment could shift on the mere possibility of a faster adoption curve.

For now, the setup is clear. The catalysts are the financials, not the hype. The scenarios hinge on the behavior of AI-native players and the pace of low-code adoption. And the risk is a narrative shift driven by a technological acceleration that the market may not be prepared for. The coming quarters will separate the durable infrastructure plays from the speculative noise.