SpaceX Officially Agrees to Acquire Anysphere, Maker of Cursor, for $60 Billion
SpaceX has officially agreed to acquire Anysphere, the company behind the AI coding tool "Cursor". The deal values Cursor at $60 billion (approximately ¥9 trillion), and Cursor's investors will receive rights to SpaceX stock based on that valuation. The acquisition is expected to close in Q3 2026.
SpaceX had secured the option to acquire the company back in April, and this formal agreement marks the full-scale launch of Elon Musk's comeback in the AI coding space. Cursor has been growing rapidly as an AI coding assistant for developers, and with competition intensifying against GitHub Copilot, AWS's Q Developer, and others, this large-scale acquisition is expected to reshape the market structure.
DeepSeek Raises First External Funding at a $50 Billion Valuation
Chinese AI startup DeepSeek has raised external funding for the first time. The company raised 50 billion yuan (approximately $7.4 billion), at an estimated valuation of $50 billion (approximately ¥7.5 trillion). DeepSeek had previously been funded by its parent company's fund, and its decision to turn to outside investors signals a new phase for China's AI industry.
DeepSeek has built a strong track record with open-weight models, and the funding is likely aimed at securing resources for future model development and infrastructure expansion. As US-China tensions deepen, the fundraising moves of Chinese AI companies are something to watch closely.
OpenAI's Annual Spending Reaches $34 Billion
It has come to light that OpenAI spent $34 billion (approximately ¥5 trillion) over the past year. This marks a sharp increase from the previous year, highlighting how the costs of training AI models and running inference are ballooning along with the scale of the business.
While revenue growth has also been reported, costs may be outpacing it, adding to market concerns about the profitability of AI companies. As models continue to grow larger, monetizing them efficiently has become a shared challenge across the industry.
Trump Administration Orders Anthropic to Restrict Foreign Access to Its Models
The Trump administration has ordered Anthropic to deny foreign-national users access to its latest AI models. These export controls exercise unprecedentedly broad authority, sending ripples even among allied nations.
The US's tightening grip on AI technology could have implications for countries around the world, including Japan. Leaders of allied nations are voicing concerns about the impact on their own technological competitiveness, once again raising questions about the shape of global AI governance.
Anthropic Withdraws Planned Billing Changes for the Claude Agent SDK
Anthropic has withdrawn, at the last minute, planned changes to the billing system for the Claude Agent SDK. The company had planned to introduce a separate credit pool, but the SDK and third-party apps will instead continue to use the existing subscription allowance as before.
With price competition against OpenAI intensifying, consideration for the developer ecosystem is believed to be behind the decision. As the race to cut API costs accelerates, maintaining a strong developer experience is emerging as a key differentiator.
Alibaba Unveils AI Models for Robots, Pivoting from Chatbots to Agents
Alibaba has announced a suite of AI models for robot control. The company is clearly positioning itself to move beyond chatbot-style dialogue toward agentic AI that operates in physical environments.
The robotics and autonomous agent space is drawing attention as the field transitions from the first generation of text-generation-centric AI to its next stage. Chinese companies have been announcing investments in this area one after another, and the race to deploy these technologies in manufacturing and logistics looks set to accelerate.
Enterprises Confront the "Tokenomics" Challenge as Token Consumption Optimization Becomes Urgent
According to WIRED, software and e-commerce companies in Silicon Valley are struggling with "tokenomics" — managing token consumption and costs — in their AI usage. There are reports of real-world workloads consuming far more tokens than anticipated, making ROI difficult to calculate in some cases.
As enterprise AI adoption moves into full swing, this offers a useful example of companies shifting from simply evaluating adoption to a phase of continuous cost management and usage optimization.
---