Author: Universe Wave Naruto, Deep Tide TechFlow
In February 2026, the tech stock market is experiencing a systemic crash that some media have dubbed the "SaaSpocalypse."
Salesforce's stock price has fallen nearly 40% from its 2025 high; ServiceNow plummeted over 11% in a single day after its quarterly earnings report, simply because management mentioned on the call that "AI agents are complicating visibility into seat growth"; Workday dropped over 22%; the entire S&P 500 Software and Services Index lost nearly $1 trillion in market value in the first six weeks of 2026.
The market's logic is straightforward: AI agents can already replace a lot of manual operations. If AI can do the work that previously required 100 people, then companies naturally no longer need 100 software seats. The SaaS business model, which charges per seat, is seen as having reached its historical end.
Just as this panic trading swept through the entire industry, Stephen Bersey, Head of US Technology Research at HSBC, published a research report with a highly provocative title:"Software Will Eat AI".
His core argument, in one sentence:The market's panic is a misjudgment.
A Report Against the Tide
"The market's fear that AI will replace enterprise software is wrong,"
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 AI's opponent; software is the vehicle through which AI reaches the real world.
This logic flips the entire narrative framework of the current market. The market fears "AI replacing software," while Bersey's judgment is that "software will tame AI."
He cited a historical analogy from the internet era: When the internet exploded, initial value accumulation was concentrated in physical infrastructure—servers, fiber optic cables, data centers. Vast amounts of capital flowed into hardware infrastructure, while the struggling early internet companies were the ones that ultimately won long-term value.Software was the ultimate destination of internet value.
The evolution of AI, Bersey believes, is replaying the same script. 2024 and 2025 were the infrastructure building years—computing power, models, code integration—all paving the way for the explosion of the software layer. And 2026 is the year the engine truly ignites.
"Software will be the primary mechanism for the diffusion of AI in the world's largest enterprises. We believe 2026 is the launch year for software monetization."
Why Can't Foundational Models Replace Enterprise Software?
The report's most substantial argument is a layer-by-layer deconstruction of the logic that "AI will directly disrupt software."
The critics' view seems compelling: Large language models can already write code, vibe coding (generating usable software directly through natural language descriptions) is on the rise, and AI model companies are already making more application-layer attempts. So why would enterprises still need expensive traditional software systems like Oracle, SAP, and Salesforce?
Bersey's answer unfolds across three levels.
First, foundational models have "inherent flaws."
The report clearly states that foundational models are "inherently flawed" and incapable of the task of "wholesale replacement" of core platforms for large enterprises. They perform well in narrow scenarios—image generation, small application development, text processing—but for high-fidelity, enterprise-grade core platforms, this is "not realistic."
The root cause lies in the limitations of training data. LLMs are trained on public internet data, while the proprietary architectural knowledge, business logic, and operational norms accumulated by enterprise software systems over decades—this core intellectual property is simply not on the public web. AI has no way to learn it or replicate it. The moat of Oracle and SAP systems isn't something that can be caught up with by writing code; it's built up over time and through business scenarios.
Second, the capability boundaries of Vibe Coding are severely overestimated.
The report directly names the fatal weakness of Vibe Coding: It places the entire responsibility and burden of design on the developer. You tell the AI "I want a system that can handle global supply chains," and the 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 a human.
More critically, Bersey points out that the major AI model companies have "almost no experience in creating enterprise-grade software." They are entering an extremely complex environment from scratch. Enterprise software,经过 decades of iteration, has evolved to a standard of "near-zero errors, high throughput, and high reliability," a baseline that AI newcomers cannot reach in the short term.
Third, enterprise switching costs are a real and high wall.
Even assuming AI could write code of equivalent quality, the cost for an enterprise to replace its core system remains extremely high: revenue disruption risk, productivity loss, system compatibility issues across IT environments, accumulated trust in supplier brands and service capabilities... These are real switching costs that won't disappear just because AI can write code.
Enterprise-grade software demands proven 99.999% uptime over many years, error-free operation in各种 complex IT environments. This trust is earned over time, not堆出来的 by code.
Who Will Be the Real Beneficiaries of AI Monetization?
If the first half of the report is a defensive argument, the second half is an offensive layout.
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 layer.
"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 constraints, data center bottlenecks—will persist for years to come. This scarcity恰恰 reinforces the strategic position of software platforms:Only software platforms can transform AI capabilities into scalable, repeatable commercial value.
And the specific vehicle for monetization, the report points to AI agents (agentic AI).
Bersey predicts that 2026 will see the large-scale deployment of task-oriented, workflow-embedded AI agents in Fortune 2000 companies and SMEs. But his characterization of agents is drastically different from the mainstream narrative. He does not see agents as disruptors replacing software but believes agents must operate within the parameters and permissions defined by software. It is precisely this "bounded agent" that can meet enterprises' needs for AI risk control.
In other words, enterprises don't need an all-powerful, free-running AI; they need an AI that can be governed, audited, and operated within a compliance framework. And this can only be achieved by agents deeply embedded in enterprise software systems.
"Software is the key pathway for enterprises to use AI in a controlled manner." This is the most core judgment of the entire report.
Additionally, the report predicts that inference demand will gradually surpass training demand, becoming the main driver of computing power consumption growth. This means that as agents proliferate, 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 published, the overall valuation of the software sector had fallen to historic lows. Bersey's judgment is:Low valuations coupled with the impending monetization year make this an entry opportunity, not an exit signal.
"Software valuations are at historic lows, even as the industry is on the eve of massive expansion."
In terms of specific stock 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 solely on per-seat pricing models will be the biggest beneficiaries of this wave of AI monetization.The buy-rated list includes Oracle, Microsoft, Salesforce, ServiceNow, Palantir, CrowdStrike, Alphabet, etc., covering almost all the core players in enterprise software.
It is worth noting that HSBC simultaneously downgraded IBM and Asana and placed Palo Alto Networks on "reduce." Not all software companies will safely navigate this transition; the key is whether they can become the infrastructure for AI agent deployment, rather than being bypassed as human interfaces.
Bersey's report is logically严密, timely, and its contrarian stance itself has strong传播效应.
But one question the report does not directly answer: If AI agents can indeed operate efficiently within the framework of enterprise software, will the demand for software "seats" still quietly shrink? The value of software as an AI载体 may hold, but whether the "per-seat收费" business model can support current valuations remains a question hanging in the air.
Will software eat AI, or will AI eat software? Every earnings report in 2026 will be new evidence in this debate.
