Can AI-generated code disrupt the SaaS industry? HSBC's analysis suggests the opposite is true.
On February 24th, Stephen Bersey, Head of US Technology Research at HSBC Holdings PLC, and his team released a report titled "Software Will Eat AI Stocks," presenting a contrarian perspective to the prevailing market sentiment.
Amidst widespread talk of "AI disruption and panic trading," HSBC Holdings PLC clearly states that software is not facing obsolescence. Instead, it is the key channel through which the world's largest enterprises can harness AI in a controlled manner.
The bank summarized its view with a contrasting statement: "As good as Hardware/Semi has been, Software will be better." The core logic is that enterprises genuinely need not just "talking models," but system capabilities that are controllable, auditable, and capable of running reliably—precisely the strengths of software platforms.
Enterprise software is not threatened by AI; on the contrary, AI will be embedded into these software platforms. Enterprise software vendors have already done the heavy lifting, including design, intuitive programming, and testing embedded intelligent agents. Furthermore, valuation levels for the software sector are at historical lows, even as the industry stands poised for a significant expansion cycle.
A major current market concern is that AI writing code will drastically lower software development barriers, allowing startups to easily disrupt established SaaS giants. HSBC Holdings PLC firmly refutes this. The report points out inherent technical limitations in foundational AI models. AI is inherently non-deterministic, potentially providing different answers or even making mistakes when faced with the same problem.
This is a critical flaw for enterprise-grade applications. "The world is accustomed to software platforms that are repeatable, auditable, and error-free in daily operations—attributes that foundational models lack," HSBC Holdings PLC emphasizes. Expecting a complete "lift-and-shift" replacement with AI for high-fidelity enterprise platforms is unrealistic.
Moreover, enterprise software has evolved over decades to achieve extremely high throughput and reliability. It contains vast amounts of critical and proprietary intellectual property (IP), data that simply cannot be used to train AI models on the public internet. "Perceptive programming is almost useless if you don't know what code you are writing," HSBC Holdings PLC states bluntly. This is analogous to a pharmaceutical company not designing its own chips or smelting its own steel. Enterprises abandoned writing their own core IT systems decades ago because it defied basic economic sense.
They quickly realized that developing, maintaining, and staffing these systems internally is prohibitively expensive. Spending vast sums to build massive platforms, only to amortize the cost over a single use case, is highly inefficient. Conversely, purchasing from software vendors, who possess the expertise for development, maintenance, and staffing, is far more economical as these costs are spread across thousands of customers.
If startups and large model providers lack the experience to build complex "enterprise-grade" architectures, then who is best positioned to use AI to generate superior software? HSBC Holdings PLC provides a definitive answer: "The software vendors themselves."
The logic is clear: established software giants possess deep domain expertise, robust sales channels, and strong customer trust. Crucially, they are already using the same AI programming tools to embed refined intelligent agents within their extensive platforms. AI's role is being simplified and "domesticated." HSBC Holdings PLC offers an analogy: AI is responsible for the creative analysis and production of intelligent data, but this data must then be handed over to deterministic software technology stacks for processing, storage, inspection, and execution.
"The vast majority of enterprise software is not threatened by AI; instead, AI will be 'domesticated' via agents within the application tech stack, creating significant value in the process," the report concludes.
From an investor's perspective, the technological logic must translate into performance guidance and market potential. HSBC Holdings PLC provides a clear timeline: major software giants began the heavy work of designing and beta-testing embedded AI agents in 2024. The technology is now mature and is being rolled out to major global clients.
"We believe 2026 will be the kick-off for monetization," HSBC Holdings PLC notes, identifying this as the primary mechanism for large enterprises to consume AI, which will drive exponential growth in AI inference demand.
Regarding investment strategy, HSBC Holdings PLC delivers a strong conclusion: "As good as Hardware/Semi has been, Software will be better." The bank views AI as a technology, but "enterprises rarely buy technology; they buy solutions to business problems," and these solutions can only come from infinitely flexible software technology stacks. Within the ecosystem generating over $100 trillion in global GDP, traditional software giants are the core beneficiaries poised to unlock AI's value potential.
Currently, the software industry's Total Addressable Market (TAM) is on the verge of a substantial 5-10 year expansion cycle. However, a market disconnect has led to software sector valuations sitting at historical lows. HSBC Holdings PLC suggests that now is an opportune time to establish or increase positions in the software sector ahead of a potential valuation reassessment.
Comments