Title: Custom Silicon, Agents, Robots Define AI's Commercial Push

01OpenAI Unveils Custom LLM Inference Chip Jalapeño Built With Broadcom

OpenAI has announced Jalapeño, its first custom processor designed specifically for large language model inference workloads. The chip, developed in partnership with Broadcom, marks a significant strategic pivot toward vertical integration in AI infrastructure. Unlike general-purpose GPUs, Jalapeño was architected to handle the unique computational patterns of serving trained language models.

The move positions OpenAI to reduce its dependence on Nvidia, which currently dominates the AI chip market. By controlling its own silicon, OpenAI gains more control over costs, performance optimization, and supply chain constraints that have plagued the industry. This puts OpenAI in direct competition with other tech giants developing custom AI silicon, including Google, Amazon, and Microsoft.

Industry observers see the announcement as a signal that the custom silicon race among AI labs is accelerating. The ability to build purpose-built inference hardware could become a competitive moat as the industry moves beyond training-focused compute toward inference-heavy workloads.

02Google DeepMind Adds Computer Use Capabilities to Gemini 3.5 Flash

Google DeepMind has introduced computer use capabilities to Gemini 3.5 Flash, enabling the model to directly control computers and interact with software interfaces like a human user. This represents a significant step toward agentic AI, systems that can autonomously complete multi-step tasks across applications.

The feature allows Gemini 3.5 Flash to move cursors, click buttons, type text, and navigate through software environments. This could automate workflows previously requiring human intervention, from data entry to software testing. Google frames the capability as enabling models to "use" computers rather than just generating text responses.

The announcement intensifies competition in the agentic AI space, where OpenAI, Anthropic, and other labs are racing to develop models that can reason through complex, real-world tasks. Gemini 3.5 Flash's computer use joins a growing ecosystem of tools designed to bridge the gap between AI capabilities and practical workplace automation.

03Agility Robotics Plans $2.5B SPAC IPO for Humanoid Robot Commercialization

Agility Robotics has announced plans to go public through a $2.5 billion SPAC merger, with $620 million in proceeds earmarked for scaling manufacturing of its Digit humanoid robot. The company, which spun out of Oregon State University in 2015, is positioning the IPO as a milestone for commercial robotics adoption.

The proceeds will fund expansion of Agility's manufacturing capabilities as the company moves beyond pilot deployments toward broader commercial deployment. Digit is designed for logistics and warehouse operations, specifically targeting tasks like moving and handling packages. Agility has positioned the robot as a solution to labor shortages in fulfillment centers.

The SPAC structure allows Agility to access public markets while avoiding the scrutiny of a traditional IPO process. The deal underscores growing investor confidence in the commercial viability of humanoid robots, a category that has attracted significant attention following Tesla's Optimus program and Figure AI's partnerships with BMW.


Also today

ShareXLinkedIn