Since the start of the year, markets have fretted that AI giants would inundate and devour traditional software. The logic was that businesses could build their own tools at lower costs, and that seat-based subscription models—where firms pay per employee—would face extinction as AI Agents take over software interactions. This bearish thesis, dubbed the "SaaSpocalypse," has loomed over software stocks like a sword of Damocles for the past two to three quarters. Yet now, that sword is being gently sheathed by a string of blowout earnings from U.S. software behemoths like Salesforce and CrowdStrike.
Salesforce' latest results have confirmed that enterprise-grade AI is already generating revenue: AI-related income surged 210% year-over-year, and earnings per share smashed expectations by 80%. Software isn't being killed by AI—it's becoming the indispensable infrastructure for AI's commercial deployment. As the "AI devours software" narrative crumbles, investors are re-evaluating three key questions: Will AI truly consume software? If Agents operate software on behalf of humans, how will software firms generate revenue? And has a pricing revolution—shifting from per-seat to per-outcome—already quietly begun? Adding to the mix, China's first national policy targeting "AI application service providers" could trigger a re-rating of the domestic A-share software sector. This article will unpack these seismic shifts with one earnings report, one pricing revolution, and one policy initiative.
Section 1: Narrative Reversal – From Zero-Sum Game to Symbiotic Growth
On the evening of August 26th (U.S. Eastern Time), Salesforce released a stellar set of fiscal Q2 2027 results, defying the pessimistic outlook on AI disrupting traditional enterprise software. The company posted revenue of $11.35 billion, up 11% year-over-year, while adjusted earnings per share hit $5.90—a staggering 103% increase and an 80.4% beat over consensus estimates. The most narrative-shifting metric was the annualized recurring revenue (ARR) from AI and data products, which neared $3.9 billion, soaring more than 210% year-over-year. Notably, ARR for its AI agent platform, Agentforce, topped $1.5 billion, marking a 240% surge. These numbers not only highlight enterprise software's resilience in the AI era but also prove that platform-based software firms—armed with proprietary customer data, workflow orchestration, and deep industry moats—are prime beneficiaries of AI commercialization.
It's worth emphasizing that Salesforce is a trailblazing customer relationship management (CRM) software provider. It delivers cloud-based customer engagement platforms to B2B enterprises, covering the entire customer lifecycle from acquisition to sales, conversion, after-sales support, and data analytics. Its roster includes numerous Fortune 500 companies. Ranking third globally in the software industry—behind only Microsoft and Oracle—Salesforce' earnings serve as a critical barometer for the entire global software sector. The market's previous anxiety was that as large language models become more powerful, companies might train their own or use open-source models to replace expensive SaaS subscriptions. But in reality, the essential elements of business operations—customer data, access controls, compliance frameworks, business processes, and audit trails—are all embedded within software systems like CRM, ERP, and collaboration tools. No matter how advanced the large model, it still requires a trustworthy data foundation and mature workflows to function effectively. The launch of Claudeforce, a collaboration between Salesforce and Anthropic, perfectly illustrates this principle: Claude's reasoning capabilities are deeply woven into Salesforce's enterprise data and business process architecture. The large model handles thinking and generation; the software manages execution and implementation. They aren't competitors—they're partners up and down the value chain.
Section 2: The Pricing Revolution – Moving from Per-User Fees to Pay-Per-Outcome
An even more profound shift is unfolding in pricing models. Twenty-five years ago, Salesforce spearheaded the industry's transition from one-time perpetual licenses to per-seat subscriptions. Now, it's at the vanguard of change again. The emergence of AI agents has drastically reduced the frequency of direct human interaction with software interfaces—many tasks are now handled by agents on behalf of employees. This erodes the very foundation of user-count-based subscriptions. Salesforce' answer is to charge based on business outcomes—focusing not merely on task completion but on a model where, for example, "we take $2 for every $20 we help you earn." This approach signals Salesforce's intent to align its own revenues with its customers' actual business results, rather than just metering task execution.
Why is the partnership with Anthropic to create Claudeforce so pivotal? Salesforce envisions a revenue mechanism where every time a third-party AI agent calls upon data within Salesforce's applications, Salesforce benefits financially. This means that even if users don't directly open the Salesforce interface, the mere act of agents invoking its data turns what could be a customer attrition risk into a new monetization channel. Notably, other key players like OpenAI and Sierra are also adopting "pay-on-completion" models. The software industry's pricing power is decisively shifting from "seat counts" to "value creation."
Section 3: Policy Catalyst – China's First National Initiative for AI Application
Turning to domestic developments, AI applications have received a major policy boost. On August 31st, China's Ministry of Industry and Information Technology issued the "Notice on the Special Action for Cultivating AI Application Service Providers," marking the nation's first national-level policy dedicated to this sector. AI applications generally refer to the layer of industry where large models and other AI technologies are embedded into software and services across sectors like office work, finance, healthcare, and education, providing users with intelligent solutions. As domestic large models iterate faster and computational costs become clearer, AI application is transitioning from technical verification to large-scale commercial deployment. This special action elevates the crucial delivery link—the service providers themselves—into national top-level design.
According to Guotai Junan Securities, China's AI industry is moving from technology catch-up to commercial monetization, with 2026 set to be a landmark year. The Chinese AI market is scenario-driven and not a zero-sum game. A clear division of labor is forming between big tech companies and vertical players: giants control computing power and general models, while vertical-focused firms burrow into high-barrier niches, offering comprehensive solutions. Token consumption volume has become a key metric for measuring enterprise intelligence and commercial value, while pricing power increasingly rests with vertical vendors holding irreplaceable scenario-specific data. Data shows that the Software Development ETF HuaBao (159036) tracks an index fully allocated to the software development industry, split as: 1) Vertically entrenched firms (58.2% weight in industry-specific application software): these deliver tailored software for government, finance, healthcare, etc., featuring high professionalism and barriers, such as Hithink RoyalFlush (financial IT); 2) Horizontally expansive players (41.8% weight in general software): these offer standardized software for cross-industry needs with broad appeal, such as Kingsoft Office (office software) and 360 (information security).
Where the Hardware Ends, the Software's Spring Begins
Looking across historical tech revolutions, profits consistently flow from hardware to applications. At the start of this year, concerns about "large models swallowing software" caused significant valuation discounts in the AI application space, making the software development sector a "value trough" within the AI value chain—offering an attractive risk-reward profile. While the software industry overall is in an upcycle, identifying the specific winning segment or stock is challenging. The Software Development ETF HuaBao (159036) offers a broad-based solution, encompassing 107 constituent stocks that comprehensively cover AI+Finance, AI+Healthcare, AI+Office, AI+Education, AI+Cybersecurity, and AI+Government sectors. With AI empowerment combined with the Xinchuang (domestic IT innovation) drive, the software development sector is poised for upward momentum.
The underlying index of this ETF captures several hot concepts. As of the end of August, constituent stocks related to AI applications, cloud computing, the Xinchuang industry, fintech, network security, and the HarmonyOS ecosystem held weights of 46.77%, 42.20%, 41.74%, 33.23%, 16.72%, and 14.31%, respectively. On valuation, the index's P/E ratio (TTM) stood at 182.89 times at end-August, which is lower than it has been more than 70% of the time since its listing, underscoring its valuation appeal and margin of safety.
Note: The individual stocks mentioned here—Hithink RoyalFlush (6.97% weight), Kingsoft Office (5.84% weight), and 360 (3.26% weight)—are all constituents of the Software Development ETF HuaBao (159036)'s underlying index as of end-August. Their inclusion is for illustrative purposes only and does not constitute investment advice, nor does it represent the holdings or trading activity of any fund managed by the company. Regarding fees: The Software Development ETF HuaBao does not charge a sales service fee. When subscribing or redeeming fund shares, the brokerage agent may charge a commission of up to 0.3%. On-exchange trading fees are subject to actual charges by the securities firm.
Risk Disclosure: The Software Development ETF HuaBao passively tracks the CSI All-Share Software Development Index, which has a base date of December 31, 2021, and a release date of March 29, 2023. The fund is issued and managed by HuaBao Fund, and distribution institutions do not bear responsibility for the fund's investment or redemption obligations. Investors should carefully read the Fund Contract, Prospectus, and Fund Product Information Summary to understand the fund's risk-return profile and select a product that matches their own risk tolerance. The fund manager has assessed the fund's risk level as R3 (Medium Risk), suitable for balanced (C3) and above investors. For suitability matching opinions, please refer to the sales institution. Sales institutions (including the fund manager's direct sales and other distributors) evaluate the fund's risk according to relevant laws and regulations. Investors should promptly check the suitability opinion issued by the sales institution and rely on its matching result. Suitability opinions from different sales institutions may not be consistent, and the risk level rating provided by a fund sales institution cannot be lower than that assessed by the fund manager. There may be differences between the risk-return characteristics described in the fund contract and the risk level rating because different factors are considered. Investors should understand the fund's risk and return, choose products carefully based on their own investment objectives, time horizon, experience, and risk tolerance, and bear the risks themselves. Registration with the China Securities Regulatory Commission (CSRC) does not imply that it makes a substantive judgment or guarantee regarding the fund's investment value, market prospects, or returns. Historical performance of the fund does not predict its future performance, and the performance of other funds managed by the fund manager does not constitute a guarantee of this fund's performance. Funds carry risks; invest with caution!