When everyone is selling software stocks, HSBC says you are wrong

Written by: Cosmic Wave Naruto, Deep Tide TechFlow

In February 2026, the tech stock market is experiencing a systemic collapse referred to by some media as the "SaaSpocalypse."

Salesforce's stock price has dropped nearly 40% from its peak in 2025; ServiceNow's stock plummeted over 11% in a single day after its quarterly earnings report, simply because management mentioned on a conference call that "AI agents are complicating the visibility of seat growth"; Workday fell over 22%; the entire S&P 500 Software and Services Index evaporated nearly $1 trillion in market value within the first six weeks of 2026.

The market logic is straightforward: AI agents are now capable of replacing a large number of manual operations. Companies using AI to accomplish tasks that previously required 100 people naturally no longer need 100 software seats. The seat-based charging SaaS business model is believed to have reached the end of its historical trajectory.

As this panic trading swept through the industry, Stephen Bersey, head of U.S. tech research at HSBC, released a provocatively titled research report: "Software Will Eat AI."

His core argument can be summarized in one sentence: The market's panic is a misjudgment.

A Contrarian Report

"The market fears that AI will replace enterprise software, and this fear is misguided."

He wrote at the beginning of the report. In his view, AI will not eliminate software but will be absorbed by it, becoming a capability layer embedded within enterprise software platforms. Software is not the enemy of AI; software is the vehicle for AI to reach the real world.

This logic flips the entire narrative framework of the current market. The market's fear is "AI replacing software," while Bersey's judgment is "software will tame AI."

He cited a historical analogy from the internet era: when the internet exploded, the initial value accumulation was concentrated in physical infrastructure—servers, fiber optic cables, data centers. A large amount of capital flowed into hardware infrastructure, while those struggling early internet companies ultimately became the ones to win long-term value. Software is the endpoint of internet value.

Bersey believes that the evolution of AI is replaying the same script. The years 2024 and 2025 are the construction period for infrastructure—computing power, models, code integration—everything is paving the way for an explosion at the software layer. And 2026 is the year when the engine truly ignites.

"Software will be the primary mechanism for AI to spread in the world's largest enterprises. We believe 2026 will be the year when software monetization begins."

Why Can't Foundation Models Replace Enterprise Software?

The most substantial argument in the report is a layered dissection of the logic that "AI will directly disrupt software."

Critics' viewpoints seem persuasive: large language models can already write code, vibe coding (directly generating usable software through natural language descriptions) is on the rise, and AI model companies are making more attempts at the application layer. So why do enterprises still need traditional software systems like Oracle, SAP, and Salesforce, which are costly?

Bersey's answer unfolds from three levels.

First, foundation models have "inherent flaws."

The report clearly states that foundation models "have intrinsic defects" and cannot competently perform the task of "wholesale replacement" of core platforms in large enterprises. They perform well in narrow scenarios—image generation, small application development, text processing—but for high fidelity, enterprise-level core platforms, this is "not realistic."

The fundamental reason lies in the limitations of training data. LLMs are trained on publicly available internet data, while the proprietary architectural knowledge, business logic, and operational standards accumulated by enterprise software systems over decades—these core intellectual properties are not available on the public internet, making it impossible for AI to learn or replicate. The moat of Oracle and SAP's systems cannot be matched by writing code; it is built over time and through business scenarios.

Second, the capabilities of Vibe Coding are severely overestimated.

The report directly points out the fatal weakness of Vibe Coding: it places the entire burden of design on the developers. You tell AI, "I want a system that can handle global supply chains," and AI can generate code, but "how to define the architecture of this system, how to handle exceptions, how to ensure it doesn't crash under extreme pressure"—these judgments still require human input.

More critically, Bersey points out that the major AI model companies "have almost no experience in creating enterprise-level software." They are entering an extremely complex environment from scratch. In contrast, enterprise software has evolved over decades to achieve levels of "almost zero errors, high throughput, and high reliability," which is a benchmark that new AI players cannot reach in the short term.

Third, the switching costs for enterprises are a real barrier.

Even assuming AI can indeed write code of comparable quality, the cost of replacing core systems for enterprises remains extremely high, including risks of revenue interruption, productivity loss, compatibility issues across IT environments, and the trust accumulated in supplier brands and service capabilities... these are real switching costs that will not disappear just because AI can write code.

Enterprise software requires a proven 99.999% uptime over the years, functioning without errors in various complex IT environments. This trust is earned over time, not built from piles of code.

Who Will Be the Real Beneficiaries of AI Monetization?

If the first half of the report is a defensive argument, the latter half is an offensive strategy.

Bersey's core judgment is: the largest share of the AI value chain will ultimately flow to the software layer, not the hardware and chip layers.

"We believe AI is the primary source of value creation in the software stack, and the largest share of long-term value will belong to software, not hardware."

He also points out that hardware scarcity, GPU shortages, power limitations, and data center bottlenecks will persist in the coming years. This scarcity reinforces the strategic position of software platforms: only software platforms can convert AI capabilities into scalable and repeatable business value.

As for the specific monetization vehicle, the report points to AI agents.

Bersey predicts that in 2026, we will see large-scale deployment of task-oriented, workflow-embedded AI agents in Fortune 2000 companies and SMEs. However, his qualitative assessment of agents is starkly different from the mainstream narrative in the market; he does not see agents as disruptors replacing software but believes they must operate within the parameters and permissions defined by software. It is precisely this "bounded agent" that can meet enterprises' needs for AI risk management.

In other words, enterprises do not need an all-powerful, free-running AI; what they need is an AI that can be governed, audited, and operate within a compliance framework. This is something that only agents deeply embedded in enterprise software systems can achieve.

"Software is the key way for enterprises to use AI in a controlled manner." This is the most critical judgment in the entire report.

At the same time, the report predicts that inference demand will gradually exceed training demand, becoming the main driver of computing power consumption growth. This means that as agents become more widespread, computing power consumption will not shrink but will continue to grow, further supporting the entire software and infrastructure ecosystem.

Opportunity or Trap?

When the report was released, the overall valuation of the software sector had already fallen to historic lows. Bersey's judgment is: low valuations combined with the upcoming monetization year present an entry opportunity, not a signal to exit.

"Software valuations are at historic lows, even though the industry is on the brink of massive expansion."

In terms of specific recommendations, HSBC's logic is clear: those software companies that have established deep data moats, possess the capability to embed AI agents, and do not rely on purely headcount-based billing models will be the biggest beneficiaries of this wave of AI monetization. The buy rating list includes Oracle, Microsoft, Salesforce, ServiceNow, Palantir, CrowdStrike, Alphabet, etc., covering almost all core players in enterprise software.

It is worth noting that HSBC also downgraded IBM and Asana's ratings and listed Palo Alto Networks as "underweight." Not all software companies can safely navigate this challenge; the key lies in whether they can become the infrastructure for AI agents to land, rather than being bypassed by agents as mere human interfaces.

Bersey's report is logically rigorous, timely, and its contrarian stance itself has a strong viral effect.

However, there is one question the report does not directly address: if AI agents can indeed operate efficiently within the framework of enterprise software, will the demand for software "seats" quietly shrink? The value of software as a carrier for AI may hold, but whether the "per head billing" business model can sustain current valuations remains an open question.

Will software consume AI, or will AI consume software? This debate will find new evidence in every financial report of 2026.