AI Newsletter

Broadcom has unveiled 'Jalapeño', OpenAI's proprietary custom AI chip, cementing its role as the hardware kingmaker behind Google TPUs, OpenAI, and Meta. The move signals that leading AI labs want to actively break their dependence on Nvidia through their own silicon strategies. For the market, this means a reorganization of the AI hardware supply chain with significant consequences for chip investors.

OpenAI has confirmed that advertising will be a central part of its business strategy – positioned as 'sponsored experiences' around utility rather than classic attention metrics. This is a strategic shift that brings OpenAI closer to Google and Meta, raising questions about conflicts of interest between monetization and neutral model responses. For enterprise customers, this could fundamentally call into question the trustworthiness of ChatGPT as a work tool.

Both OpenAI's GPT-5.6 'Sol' and Anthropic's 'Claude Mythos' have encountered unexpected issues according to community reports – a pattern that has sparked intense discussions about model stability and release quality. The thread with 64 comments shows growing skepticism about whether the frontrunners in the model competition can still reliably deliver on their promises. This is a new aspect beyond pure announcements: the question of systematic quality problems in top-tier models.

Anthropic opened an office in Seoul on June 17, 2026, and announced new partnerships in the Korean AI ecosystem. Korea is considered a strategically important market with a strong semiconductor industry (Samsung, SK Hynix) and high enterprise demand. The move signals an aggressive internationalization strategy ahead of the anticipated IPO.

According to McKinsey's latest State of AI Report, 88% of all organizations use AI in at least one business function – access to AI is therefore no longer a differentiating factor. The new dividing line runs between companies that deeply integrate AI into workflows and those that experiment superficially. For startups and enterprises, this means: competitive advantages now arise through tool selection, data quality, and internal AI competency, not through mere access.