AI News Report — 2026-06-30 (Tue)

AI industry trend report | June 30, 2026

Overview

Anthropic has entered into a major partnership with the California state government, accelerating the introduction of AI for government. On the other hand, Google restricted Meta's Gemini access due to a lack of computing resources, highlighting the structural problem of AI infrastructure supply constraints. DeepSeek has taken a new step in the race to improve inference efficiency by releasing DSpark, an open source framework that speeds up inference by up to 85%.

Anthropic / Claude trends

California Governor Gavin Newsom has entered into an exclusive partnership with Anthropic to provide AI tools for the state government. The contract allows state agencies to use Claude at half the normal price, with the aim of streamlining government services. The deal, reported as "first report" by Politico, is notable as a precedent for government AI implementation at the state level rather than the federal government. Separately, it has been reported that Anthropic has resumed providing its most powerful model to some companies in response to a request from the US government, and it is clear that the company is cooperating with regulatory authorities. Furthermore, it is reported that the company has poached a top Gemini engineer from Google, making it aggressive in the competition to acquire human resources.

OpenAI trends

Cybersecurity Dive reported that OpenAI has voluntarily restricted some functions of its new AI model in response to a request from the government. HP prepares to deploy OpenAI's Agentic platform, advancing commercialization of enterprise AI agents. The Verge reports a teaser of new hardware for Codex. On the personnel front, the head of hardware for Apple Vision Pro has reportedly been transferred to OpenAI, and Sam Altman's AI device concept is taking shape. New lawsuits related to the FSU shooting have also been filed, adding to the legal risks.

Google AI trends

Meta was secretly using Gemini for customer service, advertising tools, and content moderation, but it has been revealed that Google has restricted and terminated access due to a lack of computing resources. This is a symbolic example of a lack of AI computing directly impacting power relations between companies. Meanwhile, the company continues to expand its services for consumers, such as Gemini's personalized image generation service being made available for free to US users, and Google Meet's "Take notes for me" feature being rolled out to AI Pro/Ultra plans. CNET's report that Chrome is downloading AI without your knowledge may cause a backlash from a privacy perspective.

Trends of other AI companies

Meta: Google's termination of access to Gemini exposes the risk of external dependence on AI infrastructure. Zuckerberg emphasized the need for AI smart glasses that are both fashionable and functional. The Meta Fury AI glasses received a sarcastic review from Gizmodo, who said, "The worst company has made the best smart glasses."

xAI / Grok: Grok's voice API is integrated into Vercel AI Gateway and supports Realtime, TTS, and STT. Musk said he would bring in top engineers from SpaceX and Tesla to develop Grok. Grok 4.5 has entered private beta within SpaceX and Tesla, but there are no public benchmarks. Musk personally received approval to acquire optical components company Mesh Optical.

DeepSeek: Released open source ``DSpark'', a framework that speeds up LLM inference by up to 85%. National Review and other publications have described China's AI as a "Sputnik moment," and the geopolitical impact continues.

Microsoft: Microsoft 365 Copilot rollout for SMB accelerates. A skeptical review of Mustafa Suleyman's leadership was published on GeekWire, calling him a "consumer-friendly outsider."

South Korea: Samsung and SK Hynix announce state-led measures to strengthen AI chips and seek to break away from dependence on NVIDIA.

Remarks and trends of AI celebrities

Sam Altman (OpenAI CEO): Acquisition of Apple Vision Pro hardware manager accelerates vision for AI hardware business. There were also reports about why Amazon canceled a biopic on Altman that was scheduled to be produced.

Dario Amodei (Anthropic CEO): Successfully won a major contract with the California state government. It is mentioned in an editorial on AI and jobs, and is quoted in the context of ``we are not optimistic about the impact of AI on jobs.'' Reports looking back on the 10-year feud with OpenAI were also published in Korean media.

Elon Musk (xAI / Tesla CEO): Declared that elite engineers from SpaceX and Tesla will be diverted to Grok development. Data also emerged that SpaceX's short sellers withdrew before the announcement of Grok 4.5. Personally, I received approval to acquire Mesh Optical.

Jensen Huang (NVIDIA CEO): Declared that "prompt engineering is dead" and urged the industry to fully transition to agentic AI. Palantir and NVIDIA are also working together to integrate open AI models into U.S. government systems.

Sundar Pichai (Google CEO): The computational resource allocation strategy behind the decision to restrict Gemini access to Meta is attracting attention. It also includes a reminiscence from a former Google engineer who says he learned to "think crazy" from Pichai's stance on promoting Chrome.

Mark Zuckerberg (Meta CEO): Regarding AI glasses, he said, ``The balance between fashion and function is the key to continuing to wear them.'' There is an urgent need to respond to the sudden shutdown of AI infrastructure that relied on Google.

Yann LeCun (Meta Chief AI Scientist): References related to RAISE Summit 2026 in Paris. It was also mentioned in the context of commentary on Jeff Bezos' AI engineering initiatives.

Geoffrey Hinton (AI Godfather): Mentioned in editorial on Gen Z employment. It is cited in the context of the social issue of the employment impact of AI on young people.

Ilya Sutskever (SSI Co-Founder): Mentioned in Korean media retrospective coverage of the feud with OpenAI. There are no new direct statements.

Breakthroughs and innovations in AI technology

DeepSeek's DSpark achievement, which claims to improve LLM inference speed by up to 85%, directly addresses the industry-wide challenge of reducing inference costs. A Swiss laboratory reported a breakthrough in an experiment in augmenting human capabilities with AI (in the context of "super assist" rather than cyborg). A report that high school students discovered 1.5 million unknown cosmic phenomena using AI is attracting attention as an example of the democratization of AI in scientific exploration. At ASCO 2026, the fusion of precision oncology, biomarkers, and AI was positioned as the center of next-generation cancer treatment. Deployment of Google Gemma 4 for edge devices is also progressing.

AI Regulation/Policy/Social Impact

An essay questioning the legality of Frontier AI Lab's voluntary agreement to halt development was published in Lawfare, and the limits of industry self-regulation are being discussed. While the Defense Intelligence Agency (DIA) is accelerating data and AI modernization, it has been repeatedly pointed out that AI governance is insufficient in the legal field. The case of OpenAI voluntarily restricting new models at the government's request could serve as a precedent for cooperative regulation between governments and AI companies. California's Anthropic contract could have ramifications as a model for AI procurement at the state level rather than the federal government. Social backlash is beginning to influence policy debates, with increasing coverage of Gen Z's risk of job loss.

Integrated analysis: What is happening now

AI infrastructure supply constraints become a new axis of power: Google's restriction of Meta's Gemini access due to a lack of computational resources proves once again that "compute" is a strategic resource in the AI era, just like oil. Companies that rely on other companies for AI run the risk of having access suddenly cut off. This event reaffirms the importance of Meta's own LLM route (Llama).

Government AI adoption in full swing: Anthropic's California contract heralds a move by AI companies to take over the government procurement market. OpenAI has also voluntarily responded to government requests, and has entered the stage where building relationships with regulatory authorities is directly linked to competitive advantage.

Increasing competition in inference costs: DeepSeek's DSpark shows Chinese AI's obsession with "increasing inference efficiency." As reductions in cloud costs influence the speed at which commercial AI spreads, inference optimization technology is becoming more and more commoditized.

Shift to Agentic AI: Jensen Huang's statement that "prompted engineering is dead" heralds a tipping point for the entire industry to move from using one-off LLMs to autonomous looping agents. Multiple companies such as HP, Tavant, and Trust3 are announcing agent-based products one after another, and the ecosystem is rapidly forming.

Intensifying competition for talent: The flow of talent between major companies is accelerating, with Anthropic acquiring a top Gemini engineer from Google and OpenAI acquiring a hardware executive from Apple. There is a growing concentration on the top few companies in terms of both their ability to raise funds and their technological attractiveness.

Future points of interest

Grok 4.5 public benchmark: When will Grok 4.5, which is in private beta within SpaceX and Tesla, be exposed to public evaluation? The results of Musk's deployment of elite engineers will be questioned.

Spread of the California model: Will Anthropic's California contract serve as a template for government procurement in other states and countries? A scenario in which companies that cooperate with regulators monopolize the government market is becoming a reality.

Increasing AI compute shortage: How will the supply situation of computing resources, which is so tight that Google restricts access to Meta, affect the strategies of which companies in the future? The focus is on whether the reinforcement of AI chips by South Korean companies (Samsung and SK Hynix) can serve as a short-term mitigation measure.

AI industry trend report (June 30, 2026)

Overview

Three movements are progressing simultaneously: Anthropic's lifting of the ban on military/cyber-use AI, OpenAI's delay in releasing GPT-5.6 Preview, and Google's restriction on Gemini usage for Meta, and the AI industry is shifting from model competition to an "access control/security" phase. The axis of competition is expanding to cost and supply management, with Musk's xAI declaring a monthly model release and DeepSeek introducing peak hour billing for APIs. Job cuts and capital investment are accelerating simultaneously across the AI industry, accelerating economic structural transformation.

Anthropic / Claude trends

The U.S. administration has authorized the limited release of Anthropic's cyber-enabled AI model, making its deployment for military and security applications a reality. Bloomberg details the cyber risks of the Anthropic model through the Mythos saga. Meanwhile, there are reports that Amazon is considering switching to OpenAI and Amazon's own "Nova" model in response to Anthropic's rising costs, causing friction in the Amazon-Anthropic relationship. Anthropic has released an "AI for Science" briefing and is actively promoting its application in research and science fields.

OpenAI trends

GPT-5.6 Preview was announced, but its public release has been delayed due to a review request from the US government. As the relationship with security deepens, business for corporations and governments is accelerating, including a national security AI contract with Booz Allen Hamilton and an expanded enterprise partnership with HP. An IPO is expected by analysts to occur before 2027, with head of advertising David Dugan calling the introduction of third-party measurement a "natural step" to clarify the company's entry into the advertising ecosystem.

Google AI trends

It was reported that Google was restricting Meta's use of Gemini AI due to computing capacity constraints. This is also referred to as the "New AI Cold War," and it represents a serious rift in the cooperative relationship between Google and Meta. On the other hand, Google is strengthening its consumer rollout by opening Gemini's personalized image generation function to free users and adding AI shopping tools. Coupled with Google's inclusion in the Dow Jones index (63% increase YoY in Q1 Cloud), business expansion continues under the Sundar Pichai regime.

Trends of other AI companies

xAI/Grok Musk announces monthly model releases starting with Grok 4.5. Grok 4.5 is a base model with 1.5 trillion parameters, which is said to have been further trained using Cursor data, and is in preliminary beta with SpaceX and Tesla. X launches Hosted MCP, providing real-time data access for Grok and AI agents. British MP Jess Asato has even filed a lawsuit against xAI over sexual images.

Microsoft The Copilot system will be revamped, new lead personnel will be appointed, financial analysis functions will be strengthened in Excel, and Dataverse will become an AI agent platform. Regulatory risks are emerging in Europe as Italian authorities are investigating Microsoft 365 price hikes tied to Copilot.

Meta The company has developed its own chip technology, Vistara, and is working to improve infrastructure efficiency by reusing memory from older generation servers. Google's access restrictions may accelerate the trend toward increasing reliance on proprietary AI infrastructure.

DeepSeek Going against China's AI price competition, we introduced peak hour billing for APIs. Released open source DSpark in collaboration with Peking University, improving single-user inference speed by 60-85%. There are also reports of Indian developers moving to Chinese-made AI.

Remarks and trends of AI celebrities

Sam Altman (OpenAI CEO) Opposed to Anthropic's Attack Ads in an official interview. Amid reports of friction with the business community over the "AI Bill Costs," the OpenAI lawsuit with Elon Musk is focusing on whether a jury trial will be possible.

Elon Musk (xAI/Tesla CEO) While going on the offensive by declaring monthly Grok releases, they receive harsh criticism of their AI strategy from SoftBank CEO Son. FTC approves acquisition of Mesh Optical. Musk's influence as a symbol continues, with singer Nicki Minaj praising him as a "defender of free speech."

Jensen Huang (NVIDIA CEO) Announcement of AI partnership in South Korea. Proposed ``Loop Engineering'' as the next frontier after ``Prompt Engineering'' and suggested a shift in the AI development paradigm. In an interview with 36Kr, he expressed his recognition that "Physical AI" is the second phase of the industry.

Dario Amodei (Anthropic CEO) His name comes up in the context of Brent Sadler's statement that "Anthropic has no position" regarding setting restrictions on models. It has been mentioned in the debate over job creation vs. job extinction, and the statement that a critical phase will be reached within two years is causing a stir.

Sundar Pichai (Google/Alphabet CEO) An interview discussing "What will AI do next" has been released. The restriction on Gemini access to Meta is said to be due to a shortage of computing resources, and Pichai's management of supply constraints is being questioned.

Demis Hassabis (Google DeepMind CEO) In The Guardian and Think with Google, he has given deep philosophical thoughts on ``AI and creativity'' and ``the essential mystery of AI.'' His contributions as not only a technical leader but also a thinker are increasing.

Fei-Fei Li (Stanford HAI/World Labs) World Labs raises $1 billion to develop Spatial AI. A large-scale investment has been realized in AI fundamental technology that is responsible for 3D understanding of physical space.

Breakthroughs and innovations in AI technology

Grok 4.5's 1.5 trillion parameter scale is one of the largest models to date, and its practicality after fine-tuning using dedicated data is attracting attention. DeepSeek x Peking University's DSpark has significantly improved inference efficiency, and its release as open source is expected to have a wide impact. NVIDIA showed off AI agent workflow for the edge on RTX Spark Laptops, and edge AI implementation is in full swing. Insilico Medicine signs $2.5 billion drug development partnership, bringing AI drug discovery to industrial scale. Bloom Energy has stated that ``conventional turbines are unable to meet the power demands of the AI era,'' and is calling for a fundamental reform of energy infrastructure for AI.

AI Regulation/Policy/Social Impact

Trump's executive order referred to the regulation of frontier AI models, which became the background for the delay in the release of OpenAI's GPT-5.6 Preview. The limited release of Anthropic's cyber AI model shows the acceleration of government-led use of security AI. Privacy experts are warning of the growing use of AI by police. EU price regulation risks have become apparent, as exemplified by the investigation into Microsoft in Italy. The simultaneous acceleration of AI investment and layoffs by companies (as reported by Reuters) indicates that the economic impact of AI replacement has moved from "concern" to "reality." Discussions regarding policies for responding to AI are becoming more active in the legal and educational circles.

Integrated analysis: What is happening now

The battle for access control and supply supremacy The fact that Google has restricted Meta's use of Gemini means that we have entered the stage where AI models are managed as "strategic resources" rather than mere products. As computing capacity is a scarce resource, decisions about how much to provide and to whom are becoming central to competitive advantage.

Accelerating the fusion of security and AI The U.S. administration's lifting of the ban on Anthropic's cyber AI, the OpenAI x Booz Allen contract, and the fact that it chose SpaceX/Tesla (owned by Musk) as the destination for Grok's beta deployment show that AI is becoming institutionalized as part of military and security infrastructure. This trend is also reflected in the delay in the release of GPT-5.6 due to government review, and there is a possibility that government involvement will become the norm in future model release processes.

China's AI independence and cost structure changes The introduction of DeepSeek's peak billing and the slowdown in NVIDIA's sales in China (with the rise of Huawei, Baidu, etc.) are evidence that the Chinese AI ecosystem is weaning itself from dependence on the West. Indian developers' increasing use of Chinese AI signals that geopolitical AI competition in the Global South is in full swing.

Fei-Fei Li's Spatial AI raises $1 billion is an important signal that the evolution of AI from focusing on LLM texts to "understanding physical space" is gaining support from investors. Along with Jensen Huang's proposal of "Loop Engineering," the AI architecture debate is moving to the next phase.

Future points of interest

GPT-5.6 Preview release timing and government review results — Whether or not the precedent of requiring government approval for the release of frontier models becomes established will determine future industry standards.

What will become of the Amazon-Anthropic relationship — If Amazon's transition to OpenAI/Nova becomes a reality due to rising costs, it will have a direct impact on Anthropic's funding and valuation, and it will redefine the dynamics of cloud and AI providers.

Investment competition in Physical AI/Spatial AI — As shown by Jensen Huang's Loop Engineering theory and Fei-Fei Li's World Labs $1 billion raised, the next axis of competition is shifting to "AI that understands and manipulates physical space," and it will be interesting to see whether the fusion with robotics, autonomous driving, and XR will accelerate.