AI Infrastructure Gets Even More Expensive

Podcast

 


Welcome to P3 Media’s AI Commerce Brief, your daily update on the AI and commerce stories shaping how companies build, sell, and grow. It’s Friday, July 24, 2026. Let’s get into it.

 

The top story: OpenAI introduced Presence, an enterprise product for deploying AI agents across voice and chat workflows. OpenAI says Presence can be used for jobs like customer support, outbound sales, billing issues, insurance claims, and internal IT service requests. It is available to eligible enterprise customers through a limited general availability program, not as a self-serve product.

 

That matters because agent adoption is moving from experimentation to governance. Companies do not just need a chatbot. They need permissions, escalation rules, simulations, quality checks, and change management when policies or customer behavior shift.

 

In another infrastructure move, AMD and Cerebras announced a technical partnership for disaggregated AI inference. Cerebras says the joint system combines AMD Helios with its Wafer-Scale Engine, splitting inference work so one part handles high-throughput prompt processing and the other focuses on fast token generation. The companies say the system is expected to deliver up to five times higher tokens per second per watt. Treat that as a company claim, but the direction is important: the next AI bottleneck is not only training. It is serving large volumes of responses quickly and cheaply.

 

Commerce pulse: The Wall Street Journal reports that Stripe is in talks to acquire OpenRouter, a startup that helps developers route work across AI models. The talks could still fall apart, and the Journal reports that a sale could value OpenRouter at around $10 billion. If it happens, it would put Stripe closer to the layer where AI usage, payments, and developer procurement meet.

 

Global model watch: Kimi’s API documentation says Kimi K3 is a 2.8 trillion parameter model with a one-million-token context window, and says full model weights will be released by July 27. The broader takeaway is that Chinese model labs are continuing to compete through scale, open access, and aggressive developer distribution.

 

Finally, OpenAI launched Health in ChatGPT for US users. OpenAI says users can connect Apple Health and supported medical records, and that connected health information and conversations using it are not used to train foundation models or target ads. This is not a commerce story directly, but it is a clear marker for where consumer AI is going: more personal, more data-connected, and more sensitive.

 

What to watch: follow whether Anthropic’s AMD deal shifts buyer confidence in non-Nvidia AI systems, whether enterprise agents start replacing narrow support software, and whether Stripe’s reported OpenRouter talks turn into a signed transaction.

 

That’s your AI Commerce Brief for today. Thanks for listening.

Share