The competitive focus in the AI software sector is shifting toward actual enterprise deployment. Software platforms that can efficiently and accurately translate model capabilities into reliable business outcomes will unlock new growth opportunities.
Zhitong Finance App reports that at the highly anticipated Goldman Sachs Communacopia + Technology Conference, senior executives from leading software companies focused on B2B and B2C AI applications collectively shared their latest outlooks on future growth prospects. According to a newly released note by Goldman Sachs analysts, Ali Ghodsi, CEO of Databricks—a company specializing in big data analytics and AI engineering—stated that competition in the AI application software sector is shifting toward actual enterprise deployment. Software platforms capable of efficiently and accurately translating model capabilities into reliable business outcomes will unlock new growth opportunities.
Databricks addresses enterprise analytics needs through Genie, while Unity Catalog unifies enterprise AI and digital governance, security, and cost management. The company is also expanding into customer data platforms, security analytics, observability, and transactional databases. Based on this, Goldman Sachs emphasizes that assessing the market share growth potential of AI-related software companies over the next three years should focus on technical intellectual property, speed of innovation, and open architecture. Goldman Sachs notes that, from an investment perspective, software platforms with extensive proprietary databases and strong workflow advantages are poised for a new round of market revaluation.
Goldman Sachs analysts further stated that global capital is shifting from the first phase of AI investment—focused on GPU/HBM and AI data center hardware bottlenecks—to the second phase, where winners in the application layer convert tokens into stronger enterprise productivity, revenue, and cash flow. Future valuation divergence may become more pronounced: software companies with exclusive data, advantages in complex enterprise and consumer workflows, closed-loop agent workflow execution, and clear ROI will see revaluation, while traditional SaaS providers vulnerable to commoditization by foundational models may continue to face pressure.
AI Begins to Deliver High-Efficiency Results: Enterprise Data and Governance Determine Implementation Capability
At the Communacopia + Technology Conference, Genie demonstrated Databricks’ pathway to capturing the enterprise AI gateway. By leveraging its proprietary data indexing technology, OntoRank, and business ontologies, it organizes enterprise data into business information that models can understand, compute, and utilize, enabling the product to serve as an 'AI Agent Analyst.'
Databricks management stated that Genie has been widely adopted internally across sales, marketing, and finance, significantly transforming internal operations over the past 6–12 months. Its engineering significance lies in the need for enterprises deploying agents to uniformly understand business objects such as customers, orders, and revenue, and to integrate analytical results into daily processes. Even as model performance improves, these foundational data tasks remain decisive for accurate task completion. Management also identified potential collaboration opportunities with Palantir in complex data engineering projects.
Unity Catalog addresses the governance demands arising from the scaled deployment of agents. The Goldman Sachs analyst team pointed out that rapid model iteration increases the complexity of selection and deployment, while rising token consumption intensifies corporate scrutiny of costs and return on investment. Meanwhile, visibility into specific agent operations, data access, and security risks remains insufficient. Therefore, data governance, access permissions, security, and business semantics must be tightly integrated with the AI operating environment.
Goldman Sachs’ latest assessment of the AI application layer is that the commercial value of such platforms lies in enabling enterprises to expand deployment with confidence: clearly defining what agents can access and execute, while tracking task costs and outcomes. Consequently, customer stickiness stems more from continuous business value creation, aligning with Goldman Sachs’ emphasis on technical capabilities, innovation speed, and open architecture.
The actual product expansion of platform-based AI leaders like Databricks covers four directions: In customer data platforms, Infinite Campaigns supports a shift from segmented marketing to personalized precision marketing, with management expecting open-source models to further improve cost feasibility. In Security Information and Event Management (SIEM), new data generated by agents and the convergence of CIO and CISO responsibilities create opportunities for data platforms to enter security analytics. Observability is another expansion direction, though no product timeline has been disclosed. Lakebase is viewed by management as the largest potential new product opportunity: as AI lowers the barrier to software development, new applications require transactional databases to support daily read/write operations and business runs, thereby expanding Databricks’ market beyond analytics.
Management even suggested that the volume of software created in the next eight months could exceed the cumulative historical total. However, Goldman Sachs noted that this represents an aggressive judgment of demand and should not be regarded as a verified industry forecast.
Turning Tokens into Cash Flow: Opportunities in the Application Layer Expand, Accelerating Divergence Among Software Stocks
OpenAI’s major launch of Astra focuses on agent-based enhancements to improve computer operations, programming, and the execution of complex tasks, enabling automation for more enterprise workflows. However, competition in both “model performance” and “business implementation” will advance simultaneously. Pricing models are also becoming tiered: the Astra API continues to be priced based on input and output tokens, while AI application software vendors can charge per seat, per operation, per conversation, or based on business outcomes. For example, Intercom’s Fin charges based on defined business outcomes, starting at $0.99 per instance; Salesforce Agentforce offers billing based on operations, conversations, and user licenses.
Underlying model providers may continue to focus on selling computational resources via APIs, while upper-layer AI application providers will increasingly need to demonstrate delivered value. Outcome-based pricing aligns costs more closely with customer benefits but requires vendors to bear costs related to inference, retries, and human review. Consequently, task success rates and profit per task have become key indicators of commercialization quality.
Recent stock market trends indicate that investors are rewarding the realization of AI revenue while penalizing potential substitution risks. On September 3, Snowflake, a competitor of Databricks, saw its stock price rise approximately 17% following strong earnings results. Its product revenue for the second fiscal quarter grew 37% year-over-year to $1.49 billion, and its full-year product revenue guidance was raised from $5.84 billion to $6.07 billion, providing operational evidence that data platform consumption is expanding within the enterprise AI application layer. However, by September 8, the shares of Salesforce and Intuit fell by approximately 4%, ServiceNow dropped by about 5%, and the S&P 500 Software & Services Index declined by 1.4%. Wall Street analysts attributed part of the selling pressure to renewed concerns over large language models replacing software, triggered by the emergence of Astra.
Databricks’ software implementation path highlights the accelerating diffusion of AI benefits along the industry chain: agents executing more tasks drive consumption of computing, storage, and data services. Platforms with advantages in enterprise data connectivity, governance, and business processes are striving to convert this usage into revenue and cash flow. This constitutes a structural利好 (benefit) for software stocks, although platform expansion may squeeze the market space for some independent tools. Companies poised to benefit must deliver results in terms of new AI revenue, expanded customer spending, and profit margins. Vendors relying solely on interface or single-function advantages face intensifying competition.
The valuation anchor for the massive wave of AI investment that began in late 2022 is gradually upgrading from "scale of capital expenditure" to "efficiency of capital returns," and this process is accelerating. Specifically, the first phase of the AI investment boom focused entirely on "who dominates the allocation and benefits from building the largest GPU data centers," while the current second phase focuses on "who can convert tokens into sustainable cash flows."
The super-cycle bull market surrounding AI is shifting from “buying chip stocks” to “buying AI workflows.” The market is currently repricing the main investment theme of the AI bull run from “who benefits from continuous AI capital expenditure” to “who can fastest convert computing power into Annual Recurring Revenue (ARR), profit margins, and free cash flow.” This latest rotation favors software companies focused on AI application platforms that are embedded in critical enterprise processes, possess high renewal rates, maintain data moats, and have agent monetization capabilities. Whether AI-focused software companies can continuously integrate stronger and more cost-effective models while retaining their irreplaceable business value will determine their long-term pricing power.
Edited by Deng