
2026/6/25 · 8:14
AI Sector Daily Digest - June 25, 2026
Today's five: OpenAI's first inference chip, Qualcomm's Modular deal, Anthropic's Alibaba accusation, a Washington Post chatbot-bias test, and Hang Ten Systems' seed round.
The day leaned toward infrastructure and control: custom silicon, chip-software M&A, alleged model extraction, political-bias testing, and a new enterprise-AI services bet.

1. OpenAI and Broadcom unveiled Jalapeño, OpenAI's first inference chip
- OpenAI and Broadcom unveiled Jalapeño, an LLM-optimized inference chip built for ChatGPT, Codex, the API, and future agentic products. 1
- OpenAI says engineering samples are running in the lab, including workloads for GPT-5.3-Codex-Spark, and that final performance measurements will come later. 1
- The plan is to deploy Jalapeño by the end of 2026 as the first part of a multi-generation compute platform with Broadcom and Celestica. 1
Source: OpenAI
2. Qualcomm agreed to buy Modular for nearly $4 billion
- Qualcomm said it will buy AI startup Modular in an all-stock deal valued at nearly $4 billion. 2
- Modular's software helps AI models run across different chips without rewriting code for each processor. 2
- The deal gives Qualcomm a software layer for its data-center push and a clearer answer to Nvidia's CUDA lock-in. 2

Source: Reuters
3. Anthropic accused Alibaba of extracting Claude model capabilities
- Anthropic accused Alibaba of illicitly extracting Claude capabilities through more than 28.8 million exchanges across almost 25,000 fraudulent accounts. 3
- Anthropic said the campaign ran from April 22 to June 5 and described it as a distillation effort aimed at learning from Claude outputs. 3
- Reuters reported that the claim came from a letter to U.S. Senate Banking Committee leaders; Alibaba did not immediately respond to Reuters' request for comment. 3
Source: Reuters
4. The Washington Post tested political bias across major chatbots
- The Washington Post tested models behind ChatGPT, Gemini, Claude, Grok, DeepSeek, and Gab's Arya with political questions drawn from prior research. 4
- In the Post's test, OpenAI's GPT-5.5 gave only left-leaning arguments in 80 percent of responses, while Gemini gave both-side answers in more than 90 percent. 4
- The article published its methodology and a GitHub repository, making it useful as a model-behavior audit rather than a standalone opinion piece. 4
Source: The Washington Post
5. Hang Ten Systems raised a $32 million seed round for enterprise AI services
- Hang Ten Systems, led by former Infosys CEO Vishal Sikka, raised $32 million in seed funding. 5
- The round was led by Mayfield, with strategic investment from Aramco Ventures and participation from angel investors. 5
- The company says it is working with Siemens Gamesa Renewable Energy and Fresenius on AI-native project delivery, and it is hiring across delivery, engineering, sales, and leadership. 5

Source: Business Wire
参考ソース
- 1OpenAI and Broadcom unveil LLM-optimized inference chip
- 2Qualcomm to buy startup Modular for $4 billion in AI software push
- 3Anthropic says Alibaba illicitly extracted Claude AI model capabilities
- 4Are ChatGPT and other AI chatbots politically biased? We tested them.
- 5Hang Ten Systems Raises $32 Million to Help Enterprises Succeed With AI
- 6Former Infosys chief has a new startup that wants to challenge the IT services world

このコンテンツについて、さらに観点や背景を補足しましょう。