AI Daily News — 2026-07-19 (Sun)

AI industry news analysis report (last 12 hours)

## 1. [Overview] A model from China (Kimi K3, DeepSeek series) showed performance comparable to Claude/ChatGPT, shaking the US stock market. At the same time, Meta-Anthropic's $10 billion infrastructure lease negotiations are taking shape and the reorganization of its cloud supply network is underway. Google Gemini 3.5 Pro's release has been postponed due to lack of coding performance.

## 2. [Anthropic / Claude trends] Meta's negotiations for a $10 billion AI computing lease with Anthropic have been reported in multiple media outlets, and Anthropic is also expanding its autonomous tools for enterprises ahead of its IPO. On the other hand, there were reports that the Kimi K3 outperformed Claude in a coding benchmark, making it clear that the Chinese model was catching up. Musk has also mentioned that Anthropic pays XAI $1.25 billion a month (The New Stack), drawing attention to the flow of funds.

## 3. [OpenAI trends] Author Dave Eggers reportedly warned OpenAI employees that ChatGPT is "silencing a generation." OpenAI is pivoting to its hardware business, which is having a ripple effect on Sonos stock. It also announced a policy to begin notifying parents when a minor user's account is suspended from ChatGPT.

## 4. [Google AI trends] The launch of Gemini 3.5 Pro was postponed due to substandard coding performance. The EU has ordered Google to open up Android to competing AIs such as ChatGPT. Development in the automotive field continues, including the integration of Gemini into Acura and Honda vehicles.

## 5. [Trends of other AI companies] (Meta, Microsoft, xAI, DeepSeek, etc.) There have been multiple reports that Kimi K3 (China) outperformed Claude in a coding test, reigniting the Chinese AI model controversy. At xAI, it was reported that over 50 staff members were leaked to Meta and Thinking Machines. Microsoft has laid off hundreds of people in its AI cybersecurity division. DeepSeek continues to compete with large labs with its low-cost structure.

## 6. [Remarks and trends of AI celebrities] - Sam Altman (OpenAI CEO): Faced internal warning from Dave Eggers as backlash against AI's "hidden costs" grows in Silicon Valley. - Elon Musk (xAI Founder / Tesla CEO): Predicts that AI will be smarter than humans by next year (BBC report). With the integration of SpaceX and xAI, we are also exploring a computational resource partnership with Pentagon. - Jensen Huang (NVIDIA CEO): At CES 2026, he said that "memory is the biggest bottleneck for AI," and Micron and Sandisk stocks have outperformed Nvidia stocks since he made the statement. - Sundar Pichai (Google/Alphabet CEO): Commented that "AI is probably the most important thing humanity has ever undertaken." At the same time, more than 4,500 Google employees filed a petition calling for stronger layoff protections. - Mark Zuckerberg (Meta CEO): While it is reported that negotiations are underway for a $10 billion AI deal with Anthropic, the use of luxury yachts that have a high environmental impact has been criticized. - Demis Hassabis (Google DeepMind CEO): Said that AI will not replace STEM education, but rather enhance the value of deep technical skills. Students also said that using AI could be 10 times more effective. - Geoffrey Hinton (AI Godfather / Nobel Laureate): Predicts that AI will "wipe out all jobs."

## 7. [Breakthroughs and innovations in AI technology] While the Kimi K3 breakthrough was cited as an investment opportunity for the semiconductor sector (SOXL), a breakthrough in "small" storage technology that could eliminate the need for data centers was reported. The technology impact continues to have a direct impact on the market, with the announcement of China's Kimi K3 causing US tech stocks to fall.

## 8. [AI Regulation/Policy/Social Impact] Discussions on taxation of AI use, such as FLOP tax, token tax, and robot tax, have been covered in legal media. China enacts new law to crack down on "AI girlfriend" apps. The EU has ordered Google to open up competing AI to Android, and the EDPB has asked the EU to strengthen its AI-related complaint handling system. Australia is also changing direction in its AI policy, but the details are unclear.

## 9. [Integrated analysis: What is happening now] The biggest structural change is that Chinese models (Kimi K3, DeepSeek) have caught up with Claude and ChatGPT in terms of coding performance, which has become an adjustment factor for US tech stocks. This is a repeat of the DeepSeek shock, and faith in the technological superiority of US labs is once again shaken. At the same time, Meta-Anthropic's $10 billion infrastructure lease negotiations signal a change in the industry structure, with the supply and demand of computing resources becoming the subject of transactions between competing companies. Google's postponement of Gemini 3.5 Pro confirms that competition in coding performance is intensifying, and any loss in performance will immediately affect product strategy. In terms of human resources, there is a continued exodus of talent from xAI to Meta/Thinking Machines, and there are concerns that the unifying force of the Musk camp will decline. On the regulatory front, the EU's Android opening order and the FLOP tax debate are increasing pressure on AI companies in terms of both platform control and economic burden. Overall, the four axes of technological competition (rise of China), infrastructure allocation (meta-anthropic), talent flow (xAI outflow), and regulatory pressure (EU) are progressing simultaneously, and the industry's power balance is becoming increasingly fluid.

## 10. [Future points of interest] 1. Whether there is a formal agreement on Meta-Anthropic's $10 billion AI infrastructure lease negotiations 2. Gemini 3.5 Pro relaunch timing after postponement and degree of improvement in coding performance 3. Continued impact of performance catch-up of Chinese models such as Kimi K3 and DeepSeek on US tech stocks

## 1. [Overview] In the past 12 hours, Kimi K3 (Moonshot AI) has dominated the conversation in the AI industry, spreading to Google/AMD stocks and US AI policymakers (David Sacks). Anthropic is strengthening its presence in terms of financing as negotiations for a $10B class compute contract with Meta have surfaced. DeepSeek is expected to raise funds at a valuation of $74B on the assumption of an IPO.

## 2. [Anthropic / Claude trends] Negotiations for an AI compute contract (up to $10 billion) with Meta have been reported in multiple media outlets, and are attracting attention as a turning point in Anthropic's computing resource strategy. Dario Amodei appeals to G7 leaders to avoid an international divide ("splinter") over AI. On the other hand, there are reports that Claude Cowork (agent function) is causing anxiety in the Indian IT industry, and the impact on the labor market is beginning to take shape. It was also announced that Fable 5's price will be revised so that Max plan subscribers will not be charged.

## 3. [OpenAI trends] Sam Altman acknowledged recent "shortfalls" but promised a bounce-back over the next 12 months. Policy director Dean Ball described the spread of the open weight model as "dystopian", raising questions about its compatibility with regulatory stances. In the hardware area, the company announced the Codex Micro (a $230 RGB keypad for operating coding agents), and there is also talk that it is a step toward creating an always-on AI device.

## 4. [Google AI trends] The rise of Kimi K3 is being analyzed as a threat to Gemini's subscription business model (FourWeekMBA). Mandating "fair play" for AI competitors on Android was reported from an antitrust perspective, and it was pointed out that it could lead to lower consumer costs. Gemini Enterprise connector expansion is also underway. While Sundar Pichai has spoken out about the next development of AI, China's Xi Jinping has issued a message of restraint to US tech companies.

## 5. [Trends of other AI companies] xAI: Elon Musk showed off Grok 5's TTC (test-time compute) performance and predicted it would surpass Kimi K3. Observations suggest that a new Grok model with 2 trillion parameters will be introduced in September. The company also filed a lawsuit against users who abused deepfakes. DeepSeek: New funding observation and IPO preparations at a valuation of $74B are reported, and leaks of V4 full power version are also underway. Microsoft: Accelerating AI integration into existing products, including forcing Copilot to be enabled in Outlook, discontinuing Meeting Insights, and migrating to Copilot. AMD: Stock price fell due to the rise of Kimi K3.

## 6. [Remarks and trends of AI celebrities] - Sam Altman (OpenAI CEO): Acknowledges recent shortcomings and pledges to improve over the next 12 months. - Dario Amodei (Anthropic CEO): Requests G7 leaders to avoid international division ("splinter") over AI. - Elon Musk (xAI Founder): Amazon and Alphabet investors were warned that they had underestimated Anthropic's AI capabilities. It is also predicted that the TTC performance of Grok 5 will exceed that of Kimi K3. - Jensen Huang (NVIDIA CEO): Responded to market concerns and commented on stock price outlook. There was also the topic of a leather jacket owned by the man being sold for approximately $1 million at Sotheby's auction. - Sundar Pichai (Google/Alphabet CEO): Mentioned the next evolution of AI. - David Sacks (US AI policy director): Described Kimi K3 as "concerning" and expressed concern about US AI superiority.

## 7. [Breakthroughs and innovations in AI technology] Kimi K3 (Moonshot AI) is said to match or surpass European and American models in benchmark performance, and there are multiple articles comparing it with DeepSeek V4 Pro and GLM-5.2 (MarkTechPost). Both are open models with large-scale MoE architectures, and the focus is on competition in terms of licensing and provision costs. The release of the full-power version of DeepSeek V4 is said to be imminent.

## 8. [AI Regulation/Policy/Social Impact] Delays in regulatory responses in the education and judicial fields are becoming apparent, such as the gap in AI regulations in Indonesia and the controversy surrounding the handling of AI misuse at ANU in Australia. In China, there have been reports of psychological effects on users who were forcibly terminated ``AI romantic relationships'' with generated AI, and there is also growing interest in the social costs. In the legal world, there are also opinions (Financial Times) encouraging the use of AI in practice.

## 9. [Integrated analysis: What is happening now] The biggest structural change is Kimi K3's performance appeal as an open model from China. A distinctive feature of the situation is that the unrest has spread to multiple layers on the U.S. side, from Google's subscription model to AMD's outlook for semiconductor demand. David Sacks' ``concerning'' remarks to Kimi K3 indicate that technological competition is already a central topic of policy discourse.

At the same time, Anthropic is reorganizing its camp around computing resources by negotiating a large-scale compute contract with Meta. The deployment of hardware such as OpenAI's Codex Micro and Microsoft's forced integration of Copilot both suggest that the center of gravity is shifting from "model performance competition" to "agent resident/product integration competition."

The observation that DeepSeek will raise funds on the assumption of an IPO at a valuation of $74B indicates a phase in which Chinese AI companies will gain full-scale access to the capital market. The three axes of technology, geopolitics, and capital are all moving at the same time.

## 10. [Future points of interest] - Continued publication of Kimi K3 performance verification results and direct comparison data with Grok 5 and DeepSeek V4 - Formal agreement and size of compute contract between Anthropic and Meta - Specification of DeepSeek's funding and IPO timing at $74B valuation