The Information Applied AI newsletter

Oct 1, 2026 Applied AI By Jessica E. Lessin While interviewing Google’s Koray Kavukcuoglu last week, I had a flash of deja vu. In his first sit-down since taking over as CEO of Google DeepMind, Kavukcuoglu offered some of his most detailed comments yet on where the company's robotics ambitions are headed and how they differ from the rest of the industry. Rather than chase a single flagship humanoid the way OpenAI-backed Figure or Elon Musk‘s Tesla have, Kavukcuoglu described Google’s approach as fundamentally a software and intelligence play. I pressed him on whether Google plans to build its own humanoid. He didn't rule it out, but he was clear that Google sees its edge as the model layer, not the chassis. I had an instant flashback to when I asked Google co-founder Sergey Brin in 2007 or 2008 whether Google would ever build a phone. At the time, the company was partnering with T-Mobile and HTC to build an Android device. But would Google build its own? While rollerblading across the Google cafeteria (and here I do not exaggerate), Brin told me only if the company had to, cracking the door open for Google to make its own phone, which it did, first in partnership with HTC and then eventually more independently with Pixel. If I were a betting woman, I would bet that eventually (and eventually could be a very long time), Google verticalizes and makes its own humanoid-like robots too. Today, Google is using the first part of the Android playbook: partnering. The company has built Gemini Robotics, a version of its core Gemini models adapted specifically for physical control.Its strategy is to get that intelligence running across many different robots built by other companies like Boston Dynamics rather than betting everything on one hardware platform of its own. “We try to work with other robotics companies for them to use this, so that we put the intelligence into all the robots...in a good and safe way,” Kavukcuoglu said at The Information’s AI Agenda Live last week. (You can watch the wide-ranging interview here.) Kavukcuoglu also connected the robotics push to Google's broader thesis around artificial general intelligence: the same multimodal Gemini architecture that handles text, vision and audio is supposed to generalize into the physical world too. That means pedestrian detection, translation and now robot control all running through one system rather than bespoke models built for each task. And he name-checked Waymo as the clearest proof point so far, calling autonomous driving “a really, really safe way” to show what physical AI can do at scale. Speaking of Gemini, Google on Wednesday announced the launch of Gemini 4 Argon to a select group of customers interested in using it for cyber defense, ahead of a broader release. That means we’ll soon know whether Google has narrowed the gap between itself and the two leading AI firms, Anthropic and OpenAI, in terms of general model capabilities. New From Our Reporters AI Agenda Why OpenAI’s DevDay Looked Like Catch-Up By Stephanie Palazzolo Exclusive Cracks Emerge in AI’s Debt-Fueled Data Center Boom By Dakin Campbell Exclusive How SpaceX’s AI Unit Turned Itself Into an AI Cloud Firm By Grace Kay Exclusive Google Is Paying About 100 Digital Publishers for AI Overviews By Alix Coutures and Catherine Perloff Upcoming Events Thursday, October 8 — AI’s Power Challenge: Building the Infrastructure for the Next Generation of AI Can the grid keep pace with AI? Steve LeVine moderates a live discussion with William Mao, Allan Schurr, and Brian Janous on scaling data centers for the AI era. Register now. More details Tuesday, October 13 — The AI Value Gap As enterprises move beyond AI deployment, where is real value landing? Aaron Holmes moderates this live discussion unpacking AI ROI with executives from Celonis and National Life Group. Register today. More details Tuesday, October 27 – Wednesday, October 28 — The Information’s 2026 WTF Summit The Informati