AI News Report — 2026-07-01 (Wed)

[Overview] Anthropic officially launches "Claude Science" for researchers and the pharmaceutical industry, and also begins its own drug discovery program. While Google announced ``Nano Banana 2 Lite'' and ``Gemini Omni Flash,'' it restricted the provision of Gemini to Meta due to the tightness of computing resources. Around NVIDIA, improvements in the cost efficiency of DeepSeek models and the rapid growth of emerging chip company Etched are progressing in parallel.

[Anthropic / Claude trends] Anthropic announced Claude Science, an AI workbench for scientists. TechCrunch clearly points out this point with its strategy of differentiating itself on the workflow side rather than on new models. At the same time, a unique drug discovery program in the biopharmaceutical field was launched, as reported by CNBC, Reuters, and Endpoints News. Data from OfficeChai shows that Anthropic, led by Dario Amodei, is more vocal than OpenAI about regulation and AI safety.

[Trends in OpenAI] The company has announced a new benchmark, "GeneBench-Pro," and is working to build an evaluation axis in the bio/science field. According to The Information, they have also discovered a new method that reduces inference costs by half. Meanwhile, a lawsuit was reported regarding the victims of the FSU shooting incident, calling into question the hasty introduction of chatbots to the market. MediaPost also reports that Anthropic and OpenAI account for the majority of startup AI revenue.

[Google AI trends] The company announced "Nano Banana 2 Lite" and "Gemini Omni Flash" to promote low-cost media generation. On the other hand, multiple media outlets have reported moves to restrict the provision of Gemini to Meta due to the tightness of AI computing resources, and the resource allocation decision is being made while facing a $46 billion backlog. Gemini's remote control feature (for Mac) was also featured by Forbes.

[Trends of other AI companies] Meta launches Meta AI with Hebrew support in Israel, including WhatsApp integration. In response to Google's supply restrictions, Mark Zuckerberg has said that he is not buying the company's concerns about job losses, while Yahoo Finance analyzes that the company is moving back to its own AI strategy. Microsoft has made 365 Copilot's "Service Agent" GA (Generally Available) and has also integrated Copilot Autofix into Azure DevOps. Regarding DeepSeek, it was reported that NVIDIA reduced the inference cost of DeepSeek v4 by up to one-fifth just by tuning the software for Blackwell, and Meituan learned LongCat-2.0 using Chinese chips and made it open source.

[AI celebrities' statements and trends] Neon is expected to acquire the movie "Artificial" (directed by Luca Guadagnino), which is based on Sam Altman (OpenAI CEO), and multiple media reported that this is a move that follows Amazon's cancellation of its release. Elon Musk (xAI Founder/Tesla CEO) is reported to have once again regained his status as a trillionaire (with assets of over $1 trillion) while facing a lawsuit challenging his authority to enforce the Clean Air Act over a huge AI data center. Jensen Huang (NVIDIA CEO) said, ``AI supercomputers can become commonplace in homes,'' suggesting that it will spread to the consumer computing market. Fortune analyzes that Dario Amodei (Anthropic CEO) is sticking to his own path despite friction with Washington. Mark Zuckerberg (Meta CEO) expressed skepticism about excessive fear of job loss due to AI.

[Breakthroughs and innovations in AI technology] AI chip startup Etched has come out of stealth to reveal a working chip, $800 million in funding, more than $1 billion in customer contracts, and a valuation of $5 billion. It is attracting attention as a rival axis to NVIDIA. 3 billion dollars of VC funds are flowing into the World Models field, a move that is in line with the direction advocated by Yann LeCun (Meta Chief AI Scientist).

[AI regulation/policy/social impact] While the development of AI regulatory trackers is progressing in countries including the United States, some essays have been published pointing out the need for a local policy perspective, saying that ``global regulations cannot be simply copied and pasted''. Law.com also reported that while law firms are taking the lead in introducing generative AI workflows, governance arrangements have not kept pace. AWS is investing $1 billion in a new AI division to accelerate AI implementation in the public sector.

[Integrated analysis: What is happening now] The focus of the picture is shifting to the dynamics surrounding the ``right to allocate computational resources.'' The fact that Google has reduced the supply of Gemini to Meta indicates that we have entered a stage where the scarcity of TPU/GPU infrastructure directly influences relationships between companies. The approaches of the top two companies are diverging, with Anthropic choosing a strategy to differentiate itself through workflow (Claude Science) rather than announcing new models, and OpenAI focusing on cost optimization and benchmark development. On the chip side, specialized ASIC startups like Etched have begun to penetrate the NVIDIA-dominated structure, and Chinese companies such as DeepSeek/Meituan are competing with software optimization and learning from domestically produced chips. VC funds are starting to move toward the next technology bet called the global model, and we can see a shift in investment themes after the generative AI cycle ends.

[Future points of interest] 1. Will Gemini supply restrictions between Google and Meta be prolonged, and how far will Meta's return to its own AI progress? 2. Whether the practical use of Claude Science will expand in the pharmaceutical industry, and the progress of Anthropic's original drug discovery program 3. How much will emerging AI chip companies, including Etched, cut into NVIDIA's market share?

AI industry trend report (last 12 hours)

## 1. [Overview] The biggest move was that Anthropic consulted with the US administration and the Trump administration lifted export restrictions on Claude Fable 5 and Mythos 5. DeepSeek announces $7 billion in funding, a sign of change in China's AI industry. Google has reached the point where it is restricting external provision of services such as Meta due to the tight cloud capacity of Gemini.

## 2. [Anthropic / Claude trends] The US Department of Commerce has lifted export restrictions on Claude Fable 5 and Mythos 5, allowing Anthropic to expand its high-performance models overseas on the condition that it continues to cooperate with the US government. Major media outlets such as NYT, WSJ, CBS, and Al Jazeera are reporting on the incident, and it is attracting a lot of political attention. Dario Amodei has a presence at science-related events (including Claude Science presentations), and Anthropic is reportedly aiming to "write the rulebook for frontier AI." At the same time, there have been reports of tensions with the Department of Defense over AI ethics, highlighting the dual nature of deregulation and government cooperation.

## 3. [OpenAI trends] It was reported that the company is diversifying its advertising business and is developing image, video, and conversational ad formats. The company is making notable moves in the infrastructure and hardware field, such as developing a custom chip "Jalapeño" with Broadcom and entering into a strategic partnership with HP. After Amazon postponed the release of the film ``Artificial'' based on Sam Altman, Neon acquired the distribution rights and took a new turn.

## 4. [Google AI trends] While the product is progressing, such as implementing Gemini Spark into the Mac version of Gemini App and expanding the "Ask Gemini" function to Google Drive, new access to Gemini has been restricted due to cloud capacity constraints. The Financial Times reported that the new macOS AI agent will be limited to Ultra subscribers for $99.99 per month, and the Gemini offer for Meta will also be restricted. There are speculations that the release of Gemini 3.5 Pro will be pushed back to July, and there are reports that Alphabet's stock has suffered a $22.5 billion market capitalization decline.

## 5. [Trends of other AI companies] (Meta, Microsoft, xAI, DeepSeek, etc.) Meta has caused a stir by announcing layoffs for 8,000 people at 4:00 a.m., and while Zuckerberg has refuted concerns about job losses, it is reported that anxiety and frustration are spreading within the company. A new AI strategy called “Muse Spark” has also emerged. xAI begins rolling out Grok 4.5 in private beta for Tesla and SpaceX. Microsoft is adding native cost management functionality to Visual Studio and changing the billing system for GitHub Copilot. DeepSeek plans to raise $7 billion in funding to invest in computing infrastructure in China, and DeepSeek V4 aims to break away from price competition by introducing a pay-as-you-go pricing model.

## 6. [Remarks and trends of AI celebrities] - Dario Amodei (Anthropic CEO): Following the lifting of export restrictions, he has made a series of statements aimed at strengthening Anthropic's position. Ethical tensions with the Department of Defense are also unfolding at the same time. - Mark Zuckerberg (Meta CEO): Immediately after mass layoffs, Barron's reports that there is widespread unrest within the company as he refutes concerns about job losses due to AI. - Jensen Huang (NVIDIA CEO): "AI agents will outnumber humans." Also involved in the UN's "AI for Good" committee. On the other hand, development of the Rubin Ultra 4-die GPU was discontinued due to packaging constraints, and the performance forecast for 2027 was halved. - Sundar Pichai (Google/Alphabet CEO): Mentioned by a former Google employee who points out similarities with Larry Page and Sergey Brin. Amidst the Gemini supply limit issue. - Elon Musk (xAI Founder / Tesla CEO): ``Reshaping the world'' (The Conversation), likening it to Henry Ford. Tesla has invited a former Intel executive to lead the Terafab project. - Geoffrey Hinton (AI Godfather): Mentioned as an investor in AI chip startup Etched, which reached a valuation of $5 billion.

## 7. [Breakthroughs and innovations in AI technology] The development of an AI model that detects pancreatic cancer earlier has been reported, and its practical application in the medical field is progressing. Willow wins “Autonomous Decision-Making Platform of the Year”. Sergey Edunov (former head of Meta Llama) points out that the most important application of diffusion models is moving to the field of drug discovery.

## 8. [AI Regulation/Policy/Social Impact] The U.S. Department of Commerce is currently expanding its export control framework and granting exceptions to "trusted partners," and Anthropic's case appears to be the first. Bloomberg Law points out that cooperation between companies regarding AI safety raises antitrust concerns. Reuters also features a governance discussion on the use of AI in board meetings. The US Senate is currently soliciting opinions on the pros and cons of legislation regarding the AI agent market.

## 9. [Integrated analysis: What is happening now] For Anthropic, the lifting of export restrictions is not just a commercial lifting of the ban, but also a political transaction in exchange for a cooperative relationship with the U.S. government. As Politico points out, the Trump administration's AI policy "oscillates between tightening and loosening regulations," and a tug-of-war continues between the conflicting demands of ensuring the risk of technology leakage to China and ensuring the international competitiveness of domestic AI companies.

It is noteworthy that the Google camp has taken a defensive stance by restricting supply to external providers (including Meta) as the strain on cloud computing resources has come to light. This is a sign that infrastructure constraints are starting to dictate corporate behavior before technology roadmaps, and combined with the discontinued development of NVIDIA Rubin Ultra (packaging limitations), it suggests that the entire AI industry is facing a hardware ceiling for scaling.

DeepSeek's $7 billion raise and switch to a pay-as-you-go model signifies the end of the price competition phase and the transition to the monetization phase. The scale of China's AI investment is expanding in parallel with the tightening of export controls in the United States, and geopolitical polarization is progressing in terms of both infrastructure investment.

Meta's mass layoffs and Zuckerberg's rebuttals symbolize the internal contradictions within companies where AI investment is expanding and staff reductions are occurring at the same time, and social concerns about the impact on the labor market are likely to become even more apparent in the future.

## 10. [Future points of interest] 1. Will the lifting of Anthropic's export restrictions set a precedent for loosening restrictions on other US AI companies (OpenAI, Google)? 2. How will the prolonged strain on Google's cloud capacity affect Gemini 3.5 Pro's release delay? 3. How will DeepSeek's $7 billion raise affect domestic AI infrastructure investment in China and international competition for computing resources?