According to a research report from Orient Securities, the software sector, both domestically and internationally, has faced sustained pressure this year. The core reason is that the market has begun to re-evaluate the substitution risks posed by large language models and AI Agents to traditional software. Under the overarching narrative that "models will consume software," what truly deserves attention are software companies that can become AI infrastructure, data entry points, and business enablers. The main views of Orient Securities are as follows:
The Most Pessimistic Phase for Software May Have Passed
The software sector has been under pressure this year, with the primary cause being the market's reassessment of the threat of substitution from large models and AI Agents. Historically, software distribution relied on users actively opening applications and paying for seats. However, as AI Agents gradually gain capabilities in task planning, tool invocation, code generation, and process execution, software entry points may shift from standalone applications to models and agents. This could pressure the usage frequency, payment models, and profit margins of standardized, tool-based, and process-oriented software. In overseas markets, vertical software in legal, office, CRM, and data analytics has seen significant pullbacks due to AI substitution concerns. The same narrative has impacted domestic software firms, with fears that as general-purpose large models extend into vertical business plugins, they will erode the product moats and pricing power of traditional software vendors. The bank believes the earlier decline in the software sector was not merely due to earnings volatility but rather a systematic market re-evaluation of whether traditional software business models can sustain high growth, high margins, and high valuations. The peak of pessimism may now be over.
Recent Overseas AI-Related Software Rebound Spurs Sector Reassessment
Overseas software stocks have recently shown signs of recovery, with companies like ServiceNow, Snowflake, Datadog, Okta, and Palo Alto Networks performing strongly. This is not because the market completely dismisses AI disruption, but because it is starting to differentiate between "software replaced by AI" and "software benefiting from AI demand." On one hand, as enterprises deploy AI Agents, demand for foundational software like data governance, observability, cloud data platforms, security, identity management, and process orchestration actually increases. On the other hand, software companies that can embed AI capabilities into existing workflows and monetize through value-added modules, usage-based billing, or outcome-based pricing are regaining market recognition. The bank views this rebound as driven by a shift in the AI industry's pricing logic, moving from compute and model layers to the application and infrastructure software layers. The market is beginning to realize that AI will not simply eliminate all software but will shift software value from being a "tool entry point" to deeper elements like "data, processes, permissions, security, and delivery outcomes." Software companies benefiting from this shift remain worthy of attention.
AI Infrastructure and Software Firms Effectively Embracing AI Stand to Benefit Most
The bank posits that the investment focus in the software industry will shift from traditional IT and general SaaS to two more certain directions. The first is AI Infrastructure (AI Infra) software, including databases, data governance, vector search, cybersecurity, identity and access management, model operations, and agent orchestration. These form the foundational support for enterprise deployment of large models and agents. The second direction comprises software companies with established data, scenarios, and customer bases that can deeply integrate AI into their core products. These companies are not passively awaiting substitution by large models but are proactively leveraging AI to enhance product capabilities, delivery efficiency, and customer stickiness. For example, Hehe Information's products like TextIn, intelligent document parsing, and contract robots operate at the critical juncture of "transforming enterprise unstructured data into computable information assets." They are well-positioned to benefit from the rising demand for high-quality data entry points during the implementation of financial, government, enterprise knowledge base, and agent systems.
Risk Warnings
The risks include industry demand falling short of expectations and slower-than-expected progress in AI commercialization.