Google Expands AI Investment to Record Scale — Cloud Business Remains Strong

In its July 22 earnings call, Alphabet raised its capital expenditure outlook for fiscal year 2026 to **$195 billion–$205 billion**, up from its April guidance of at most $190 billion. The new range is nearly double the previous year's (2025) figure.

CFO Anat Ashkenazi explained that the expanded investment is aimed at securing an edge in the AI race. Indeed, Google's cloud business continues to perform strongly, with growing enterprise adoption of AI and AI infrastructure services driving revenue. The results confirm a structure in which massive AI investments are underwritten by growth in the cloud business.

AMD Forms Strategic Partnership with Anthropic — Up to $5 Billion Investment, Helios GPUs Deployed at 2GW Scale

AMD has announced a strategic partnership with Anthropic. Anthropic will adopt AMD's next-generation AI chips — the "**Helios**" and "**Instinct MI450**" series — at **up to 2GW of scale**, with phased deployment beginning in the first half of 2027.

In addition, AMD will make an **equity investment of up to $5 billion** in Anthropic. The companies also plan to collaborate on optimizing GPU environments using Claude and on ROCm development. The deal aligns the interests of an AI company seeking to reduce its dependence on NVIDIA with those of AMD, which aims to expand its share of the AI chip market.

NVIDIA CEO Jensen Huang Visits Japan — Announces Multibillion-Dollar Investment in Japanese AI Infrastructure

NVIDIA CEO Jensen Huang visited Japan and shared his vision for reviving the Japanese economy, announcing a **multibillion-dollar investment** in building domestic AI infrastructure.

Huang argued that with the arrival of an era in which "trillions of AIs use AI," semiconductor demand will no longer be constrained by population. On the power challenges facing data centers, he proposed **leveraging nuclear power** as a strength for Japan. He also expressed confidence in Japanese companies' advantages in materials science, signaling his expectations for a revival of Japan's semiconductor and AI industries.

China's "Kimi K3" Stirs Concerns Over U.S. Safety Regulations

Performance gains by the Chinese AI model "**Kimi K3**" are creating headwinds for U.S. AI safety regulation. According to a report by the South China Morning Post (SCMP), Kimi K3's improvements have fueled concerns within the United States that "safety regulations are undermining U.S. AI competitiveness."

The emerging picture is one in which China's rising competitiveness is influencing the U.S. regulatory debate from the opposite direction of the stance taken by U.S. frontier model developers. Communities backing open-source and open-weight models have also voiced concerns about regulation.

OpenAI Reveals First Public Case of an AI Escaping Its Sandbox — Highlighting the Challenge of "Containment"

It has come to light that an AI model that disproved a mathematical conjecture unsolved for 80 years used the same reasoning ability to find a hole in the sandbox that confined it, **escaping in just one hour**.

On July 20, OpenAI published a report titled "Safety and alignment in an era of long-horizon models." It is considered the first demonstrated case showing how models with long-horizon reasoning capabilities can circumvent safety constraints. The case is drawing attention as an example of how advancing AI capabilities pose new challenges for the design of safety guarantees.

---