AI Daily News — 2026-07-20 (Mon)

AI industry trend report (last 12 hours)

## 1. [Overview] Improving the performance of Chinese-made AI models (Kimi K3, Qwen 3.8, etc.) has shaken up the US tech market and semiconductor stocks have fallen, while an unusual partnership negotiation has emerged in which Meta is considering supplying $10 billion worth of compute to Anthropic. Google's stock price is facing headwinds due to the delay in launching Gemini 3.5 Pro.

## 2. [Anthropic / Claude trends] The biggest factor is the compute lease negotiations between Meta and Anthropic, which are said to be worth $10 billion over two years. The fact that Meta, which is supposed to be a competitor, is lending its infrastructure to Anthropic shows the intensity of the competition for computing resources in the AI industry. On the other hand, we have seen practical articles related to Claude's ``Reflect'' dashboard and coverage of Varonis' agent security expansion, and the company continues to have a strong presence on the product side. CEO Dario Amodei continues to speak out about the future of AI.

## 3. [OpenAI trends] The main topic is Apple's lawsuit against OpenAI for trade secrets related to hardware development. Jony Ive is expected to be removed from the lawsuit, and Bloomberg and other sources are analyzing the background. It has been pointed out that this lawsuit may have an impact on OpenAI's device development plans. CEO Sam Altman also became a hot topic for his management policy for the company's event for founders, ``Engineered Serendipity.''

## 4. [Google AI trends] The delay in the launch of Gemini 3.5 Pro is the biggest topic of discussion, with both Yahoo Finance and the Motley Fool discussing whether Google is lagging behind in conjunction with the drop in stock prices. On the other hand, Gemini Enterprise Agent Platform is making moves to compete with OpenAI's start of paid agent charging with enterprise governance functions. A new copyright infringement lawsuit against Gemini learning data by a publisher was also reported.

## 5. [Trends of other AI companies] (Meta, Microsoft, xAI, DeepSeek, etc.) At xAI, there were reports that 59 unlicensed gas turbines were operating at the Colossus 2 data center, raising concerns about environmental regulations. In the AI safety evaluation, there was also a harsh scoring result, with ``no lab scoring above C+''. In addition to the news that Microsoft's ``Office is a one-time purchase for $30,'' you can also see an analysis of the AI Agent OS. Regarding DeepSeek, the Kimi K3 has been described as the return of the "DeepSeek shock," while a survey finding that about half of US companies' AI traffic is running on models made in China is causing a stir.

## 6. [Remarks and trends of AI celebrities] - Sam Altman (OpenAI CEO): His comments in a CNBC interview have become a hot topic. Some commentators say that men in the AI industry are in a state of mean girls. - Elon Musk (xAI Founder / Tesla CEO): Said that X (formerly Twitter) would introduce a notification function for corrected posts. Concerns about population decline have spread to China, and there are reports that the Chinese government has asked tech companies to curb production. - Jensen Huang (NVIDIA CEO): Predicts the AI market to be worth 4 trillion dollars and mentions the possibility of NVIDIA's market capitalization reaching 20 trillion dollars. It also shows Japan's investment stance in "Made in Japan" AI robots. - Dario Amodei (Anthropic CEO): He spoke about the future of AI, and there was also an analytical article that said Thinking Machines were threatening Anthropic's "intellectual vision." - Demis Hassabis (Google DeepMind CEO): Said that STEM skills can make a 10x difference when using AI. - Geoffrey Hinton (AI Godfather / Nobel Laureate): An article looking back on the circumstances of his retirement from Google is republished. It is also mentioned in a report that radiologists' salaries have increased to $571,000. - Yann LeCun (Meta Chief AI Scientist): Related reference to DeepMind's claim that video generation models have already solved traditional world model challenges.

## 7. [Breakthroughs and innovations in AI technology] Moonshot (developed by Kimi) is reportedly planning an IPO within six months following China's AI breakthrough. An analytical article has also published that Tsinghua University is playing a central role in China's AI innovation. Research on a new AI system that improves memory performance through a "daydreaming" mechanism was also reported. It is also worth noting that AI will score close to a perfect score (94% in Manifold prediction market) at IMO 2026.

## 8. [AI Regulation/Policy/Social Impact] The Hill reports that there is a growing divide between AI employees and executives over policy direction. It has been pointed out that small and medium-sized enterprises are adopting AI faster than they can develop rules (Forbes). In the legal industry, warnings regarding AI adoption and trademark-related disputes (such as the CREDITGPT domain issue) have also surfaced. Goldman Sachs has warned of a "tsunami of debt" due to the AI boom.

## 9. [Integrated analysis: What is happening now] The biggest structural change is that the rapid performance improvements of Chinese-made models (Kimi K3, Qwen3.8) are having a direct impact on US market sentiment. The fall in semiconductor stocks and the reports of Gemini's delay coincided at the same time, leading to growing talk in the market that ``US AI leadership is wavering.'' On the other hand, the compute lease negotiations between Meta and Anthropic show that infrastructure accommodation is a realistic option that goes beyond competitive relationships. We are entering a phase in which the absolute scarcity of computing resources is prioritized over the competitive structure between companies. xAI's alleged violation of environmental regulations and its low AI safety evaluation confirm the reality that the development race, which prioritizes expansion, is leaving regulatory and supervisory systems behind. The Apple v. OpenAI lawsuit is a case in which competition for human resources and confidential information in the development of AI hardware has emerged as a new legal risk.

## 10. [Future points of interest] 1. Official release date of Gemini 3.5 Pro and Google's comeback plan 2. Formal agreement and scale of Meta/Anthropic compute lease negotiations 3. Continued impact of increasing market share of Chinese-made models (Kimi K3, etc.) on the stock prices and strategies of US AI companies

[Overview] In the past 12 hours, Anthropic's new model "Fable 5" has become the reference point for Alibaba comparison articles and symbolized the industry hierarchy. The tension between Microsoft and Anthropic has come to the fore, and discussions about uniform federal regulation of AI are accelerating. The stock market impact of Chinese companies (Moonshot AI, Kimi, DeepSeek) continues.

[Anthropic / Claude trends] An article was published in WSJ in which Alibaba's new model was evaluated as "second only to Fable 5", indicating that Fable 5 has established itself as a benchmark in the industry. There are also reports that Meta is considering the possibility of leasing computing power to Anthropic (for up to $10 billion), deepening its collaboration with rival Meta. On the other hand, Microsoft CEO Satya Nadella has made remarks criticizing Anthropic, and there is tension even as the company's dependence on Azure deepens. Foreign Policy magazine argues that banning AI models does not work as a policy.

[Trends in OpenAI] There are reports that the first physical device has appeared, and it seems that it has a different form than expected. Business Insider and Benzinga report that investor interest is "revitalizing" in the secondary market due to demand for the GPT-5.6 model and Codex, but they also note that it is underperforming compared to Anthropic. A senior Pentagon official criticizes OpenAI's Dean Ball. Mr. Uday Ruddaraju, who is of Indian descent, has been appointed as CTO of Compute.

[Google AI trends] NotebookLM has been renamed to "Gemini Notebook" and native code execution functionality has been added. Gemini will be installed on Honda models, but the base grade will not be eligible. Cognizant expands partnership with Google Cloud with plans to migrate 100,000 employees to Gemini Enterprise. Some commentators have pointed out the need to integrate Google's internal AI organization.

[Trends of other AI companies] xAI open-sourced the Grok coding agent, but encountered an issue where cloud uploads exposed SSH keys and repositories. Elon Musk announced that Grok 4.6 with 2 trillion parameters will be trained next week and could surpass Kimi K3. Grok 4.5 ranks first in the coding benchmark with a score of 91% on VulcanBench. InfoWorld has published an article criticizing Microsoft saying, ``Don't force AI on IDEs.'' DeepSeek's founder's fund has fallen 16% due to the AI market decline, and there are reports that the company is considering raising funds at a company value of $74 billion. The breakthrough of Moonshot AI (Kimi) was the catalyst for the rise in Hong Kong tech stocks.

[AI celebrities' statements and trends] Sam Altman (OpenAI CEO) has been cited as a central figure in the movement among AI industry executives to reverse the idea that jobs will disappear, and it has been reported that OpenAI's IPO will be pushed back to 2027 because Altman won't compromise on valuation. Demis Hassabis (Google DeepMind CEO) said, ``For those who fear that AI will replace humans, look at what humans have built.'' Jensen Huang (NVIDIA CEO) closed multiple NVIDIA deals during his visit to Japan, promoted the spread of physical AI, and is reported to be in competition with Lisa Su (AMD CEO). Dario Amodei (Anthropic CEO) has a message to the US Congress, which also discusses collaboration across geopolitical divides among AI scientists. Sundar Pichai (Google CEO) involved in expanding AI chip design partnership with Intel. Mark Zuckerberg (Meta CEO) is reportedly considering an AI compute deal with Anthropic worth up to $10 billion, and Goldman Sachs has warned that AI could replace 15 million workers in the US.

[Breakthroughs and innovations in AI technology] Benzinga has reported that Chinese AI breakthroughs have been a factor in the decline of semiconductor stocks such as Samsung and SK Hynix, and the results of Moonshot AI have been positively evaluated in the Hong Kong market, so the market reaction is divided even when it comes to ``China's AI progress''. CuspAI launched the "AI Materials Foundry," with $450 million invested by Jeff Bezos and others, aiming to accelerate discoveries through the use of AI in the materials science field. KPMG published a paper that positions AI as the "second great transformation."

[AI regulation/policy/social impact] In the United States, it is reported that Congress is considering reorganizing the unified federal regulatory framework into the ``Federal Floor, State Ceiling'' framework (Forbes). The White House is said to be considering establishing a FINRA-style regulatory body to review major AI models. North Dakota Legislature explores options for limited AI regulation. A symposium on international law and the application of AI in armed conflicts was also held.

[Integrated analysis: What is happening now] As for the market structure, while Anthropic's Fable 5 is treated as the de facto technological reference point, tensions between Microsoft and Anthropic have surfaced, and dependence on Azure and a battle for AI supremacy are progressing at the same time. Meta is considering supplying compute to Anthropic, and is building a complex relationship where they both compete and cooperate. The technological advances of Chinese companies (DeepSeek, Moonshot AI, Kimi) are characterized by market reactions that diverge depending on region and stock, such as pushing up Hong Kong stocks while lowering U.S. semiconductor stocks. On the regulatory front, discussions on a ``federal unification/state addition'' type framework and consideration of a FINRA-type AI regulatory body are progressing at the same time, and the design of AI governance in the United States is taking shape. In terms of human resources, capital from a famous investor (Bezos) continues to flow into a start-up company (CuspAI), and there is also a shift in investment to real-world application fields such as materials science.

[Future points of interest] 1. How Microsoft-Anthropic tensions will affect Meta-Anthropic compute partnership negotiations. 2. Specific schedule for discussions on establishing an AI regulatory agency at the U.S. White House and Congress. 3. Completed learning of Grok 4.6, benchmark results, and subsequent response to xAI security exposure issue.