AI slowdown shakes chip stocks & Apple rebuilds Siri around AI - Tech News (Sep 15, 2026)

[0:28] AI slowdown shakes chip stocks

We start with the biggest story in AI, and arguably the one with the widest ripple effects. Over the weekend, leaders from several major AI organizations, including OpenAI’s Sam Altman, Anthropic’s Dario Amodei, DeepMind cofounder Demis Hassabis, and Elon Musk, reached an informal agreement to slow the pace of frontier AI development. It is not a law, and it is not a formal treaty, but it is still one of the clearest signs yet that the industry believes raw speed may no longer be the only goal. The reason is straightforward: safety worries are becoming too large to ignore, especially around systems that could become difficult to control.

There’s also a deeper strategic layer here. Reports from code found in iOS 27 and macOS suggest Siri may be built to work much more closely with outside AI models than previously thought. In other words, Siri may eventually act less like a single assistant and more like a coordinator that can hand tasks to systems such as ChatGPT or Claude when needed. If that turns into a public feature, it would be a major shift for Apple, which has traditionally preferred tight control over core user experiences.

Apple also expanded Apple Intelligence into the home through HomeKit Secure Video, bringing smarter analysis to camera feeds while keeping its usual privacy-first message front and center. And in a separate but related note, Apple and Elon Musk’s companies have settled an antitrust dispute over AI assistant favoritism on Apple devices. That removes one legal distraction just as Apple is trying to make Siri central again.

Over at OpenAI, the company introduced GPT-Live-1 in its API, a new voice model designed for more natural back-and-forth conversation. The notable part is that it can listen and speak in a more simultaneous way, rather than forcing a rigid one-person-talks-then-the-other pattern. For developers, that could make voice apps feel less robotic and more fluid. For everyone else, it is another signal that voice is moving from a novelty feature toward a serious interface for AI products.

The AI infrastructure race, meanwhile, shows no sign of cooling even with all the safety talk. Early test results reported for Nvidia’s next-generation Rubin systems point to a major jump in efficiency over Blackwell, especially for large-scale inference workloads. That matters because inference, the act of running AI models for real users, is where costs can spiral fast. Better performance per unit of power is not just a technical win; it can directly reshape cloud economics.

And Nvidia may be thinking even bigger. Reports say the company is working with major financial firms on an effort to raise more than half a trillion dollars for AI infrastructure. If that comes together, it would rank among the largest financing pushes the tech industry has ever seen. At the same time, challengers are trying to loosen Nvidia’s grip. Cornelis Networks, an Intel spinoff, raised fresh funding and introduced a networking approach aimed at AI clusters that do not want to be locked into a single hardware ecosystem. So even as Nvidia remains the center of gravity, the rest of the market is clearly looking for alternatives.

Finally, a space story with major geopolitical weight. A senior U.S. official has publicly confirmed that the United States has placed what were described as space control weapons in orbit. Details were not provided, but the significance is hard to miss. Washington has moved from speaking cautiously about orbital conflict to openly acknowledging deployed systems in space. Supporters will call that deterrence. Critics will see it as another step toward an arms race above Earth. Either way, it is a reminder that advanced technology competition is not only about AI models and data centers. It is also about the physical systems nations are willing to place in strategic domains.