Goldman Sachs interprets this week's market focus: "The AI-SaaS debate"

```

Goldman Sachs Technology Research Team points out that the core challenges currently facing software stocks are centered around the new competitive landscape at the application software layer, as well as concerns about the return on capital expenditures for infrastructure companies.

WallstreetCN mentioned that a routine occurrence from Anthropic severely hit software stocks this week; despite a rebound on Friday, they remain in turbulent low positions.

Goldman's Gabriela Borges team expects that an improvement in investor sentiment will require 2–3 quarters of stable fundamentals, with the software sector facing valuation pressures in the short term.

Goldman emphasizes that the onset of a new technological cycle, the surge of competitors nurtured by the financing boom in 2020–2021, and the need for system architecture redesign are all intensifying competition at the application software layer.

Borges outlines a checklist for investors, listing seven key metrics that might signal the stabilization of the AI-SaaS industry. This framework aims to help the market determine whether application software companies can withstand the impact of new competitors, and when the massive capital spending by infrastructure providers may translate into substantial returns.

Two Key Debates in the AI-SaaS Field

Goldman Sachs Technology Research Team points out that there are two major debates in the AI-SaaS sector.

The first point of contention is whether software companies will be replaced.

Goldman believes that every new technological cycle brings chaos. Currently, traditional SaaS giants, high-end custom software vendors like Palantir, and those startups with truly unique datasets and products (not just companies with an AI veneer) are all vying for customers in the same battlefield.

There are three reasons for this melee. First, when new technology emerges, enterprises always want to try it themselves, giving newcomers a window of opportunity.

Second, companies fueled by hot money around 2021 may not have as deep a moat as imagined.

Most crucially, software was previously centered on human users, but in the future, software must serve both humans and AI agents. This means the entire software stack needs to be rebuilt; the burden of legacy system transformation for established players is precisely the opportunity for new entrants.

Thus, choices become stringent. When investing, one should look critically at application software companies that are rapidly restructuring themselves to support “human-machine collaboration”.

The second focus is, who will ultimately receive the money?

Goldman has found a consensus among clients: in the future, every software vendor will offer AI agents. So, where will differentiation and excess profits come from?

The key is “orchestration”—how to efficiently and reliably integrate base computing power, AI models, enterprise data, business processes, and security rules to deliver usable agents. This spans infrastructure, platform, and application layers; technical complexity and industry expertise are both essential.

This leads to the ability of vertical integration becoming more important than in the cloud era. For giants like Microsoft, dominating all layers will be more advantageous.

At the company level, Goldman’s analysts believe application software vendors will find value converge in the “orchestration layer”. The race to transform from SaaS to “SaaS+AI” is brutal, and everyone has a different starting point.

For infrastructure providers offering computing power, the winning move lies in diversifying chip sources to increase profits and in providing valuable platform services atop basic computing power.

Seven Key Signals of Recovery

These changes are gradual. Goldman’s analysts list seven specific signals that can be tracked quarterly. If these appear successively, it suggests industry fundamentals are starting to stabilize.

First is a quiet shift in revenue structure. If this year, enterprises’ overall traditional software budgets don’t increase, but the leading software companies’ total revenue stops declining and rises, it indicates AI orders are meaningfully compensating for or even exceeding the weakness in traditional business.

Second, “self-developed” projects return. If a company like ServiceNow begins to complain that clients are switching from self-developed AI projects to mature products, that’s a positive signal. It means packaged software is establishing enough advantages in functionality, security, and compliance control, and the enterprise market’s game rules are taking hold.

Third is the courage to raise prices. Currently, most AI features are still in the promotion phase. In the future, when vendors try to charge separately or raise prices for agent features, and clients accept, it proves AI has created recognized business value. This is similar to what Palantir did one or two years ago: “let customers use it first, then talk about money”.

Next, track whether industry expertise becomes more valuable than models. The market needs clearer cases to understand why, in certain scenarios, agents from Salesforce or HubShop with deep industry experience outperform generic Copilot products. This will determine pricing power and market space for different companies’ agent products.

Fifth, watch for unexpected mergers and acquisitions. If an AI platform or large model company suddenly acquires a vertical SaaS company, don’t be surprised. This is the most direct means for AI players to quickly acquire industry knowledge, sales channels, and “enterprise-grade” credibility.

Sixth, solving talent issues with money. AI brings fierce competition for expensive technical and sales talent. Investors should focus on how companies manage risks of equity dilution and employee turnover. Whether it’s buying back shares more aggressively or decisively adjusting compensation structures to attract new talent, this reflects management’s determination.

Lastly, understand the cloud giants’ bottom line. For giants like Microsoft and Oracle that provide AI computing power, the market needs clearer visibility on when their new chip capacity will be online, and how much will be reserved for internal business versus external customers. This directly affects their cloud business growth and profitability.

Goldman finally points out that it is expected to take 2–3 quarters of stable fundamentals for investor sentiment to improve. Even so, there is a possibility that bearish scenarios could be postponed to future years.

Risk Warning and DisclaimerThe market involves risk, investment must be cautious. This article does not constitute personalized investment advice and does not take into account the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable to their particular circumstances. Invest accordingly, at your own risk. ```