August 11 Financial Highlights and In-Depth Analysis

Anthropic Launches Sonnet 5.5 to Compete with DeepSeek, OpenAI Unveils GPT-5.6-Cyber

Anthropic releases Sonnet 5.5, targeting DeepSeek with enhanced reasoning capabilities and cost efficiency. On the same day, OpenAI introduces the GPT-5.6-Cyber version, emphasizing cybersecurity and code security audit capabilities. The simultaneous competition between the two models marks the AI race's transition into a multi-dimensional strategic phase.

Chief Analyst's Insight: These two moves only make sense when viewed together. Anthropic's Sonnet 5.5 targeting DeepSeek indicates that DeepSeek's cost-effective approach has become a threat that major U.S. model companies must directly address—no longer framed as "cheap Chinese alternatives" but as "we'll lose customers if we don't keep up." OpenAI's GPT-5.6-Cyber represents competition on another dimension: when foundational model capabilities converge, differentiation shifts to vertical scenarios. Cybersecurity serves as a signal that major model companies are transitioning from "tool creation" to "solution selling."

For A-share AI application stocks, the key judgment is: the more homogeneous model capabilities become, the greater the bargaining power of the application layer, as switching costs decline. However, this The premise (prerequisite) is that application companies survive the phase where "everyone loses money to acquire users."

NVIDIA Confirms Partnership with Six Giants to Establish $500 Billion AI Infrastructure Financing Platform

NVIDIA, along with Microsoft, Google, Oracle, BlackRock, Global Infrastructure Partners, and MGX, jointly announced the establishment of an AI infrastructure financing platform with a target size of $500 billion to support global AI data center and computing infrastructure development.

Chief Analyst's Insight: The $500 billion figure requires dissection. Among the six partners, BlackRock and GIP handle financing, NVIDIA supplies GPUs, and the three major cloud providers serve as tenants. This is not a charitable alliance but a sophisticated machine for financializing AI infrastructure. Essentially, NVIDIA is creating a "buyers' alliance" to insure its GPU orders: cloud providers don't need to pay upfront for data center construction, BlackRock earns long-term rental returns, and NVIDIA secures chip shipments for years to come.

For A-share implications, the most direct beneficiaries are NVIDIA supply chain components like optical modules, PCBs, and liquid cooling systems. However, the critical question is: if the $500 billion financing platform materializes, could global computing supply rapidly become surplus (surplus)? The answer to this question matters more than GPU orders themselves.

AI SaaS Application Procurement Slows as Enterprise Clients Remain in Experimental Phase, with Inference Costs Under Pressure

Enterprise AI tool procurement is cooling, with clients still in experimental stages, extended transaction cycles, and increased expectations for free trials. 66% of enterprise token budgets come from reallocations in labor or application software sectors. Core products from companies like HubSpot show downgrade trends. Except for Shopify, most AI agent user growth has yet to translate into fundamental improvements, posing incremental negative impacts on Intuit and Workday.

Chief Analyst's Insight: This may be the most critical yet overlooked news item today. After over a year of AI SaaS narratives, reality is finally hitting. Enterprise clients are trialing, experimenting, and reallocating budgets but refusing to pay at scale. The fact that 66% of token budgets come from "robbing Peter to pay Paul" indicates AI SaaS hasn't created incremental budgets but is cannibalizing existing software and labor budgets. The downgrade trend in HubSpot's core products is particularly alarming: clients aren't upgrading from traditional HubSpot products to AI—they can't afford either.

For companies like Salesforce, Workday, and Intuit that treat AI as a growth engine, if AI token revenue merely replaces their highest-margin core product revenue, this isn't growth but a zero-sum game. A-share AI application companies are far from this dilemma—their issue isn't substitution effects but the lack of paying customers altogether. However, if the U.S. AI SaaS narrative collapses, A-share AI concepts will hardly escape unscathed.

- END -

!