DeepMind's Veo rollout
What happened
- DeepMind has started rolling out Veo, a text-to-video system, to creators and enterprise customers. - The announcement frames Veo as part of DeepMind's push into generative video for production workflows. - The rollout ties into bigger cloud and hardware narratives about scaling video generation and inference workloads (x.com).
Why it matters
Google DeepMind has started putting Veo, its text-to-video model, into the hands of creators, developers and enterprise customers through Google’s own apps and cloud tools. (deepmind.google) Veo is Google DeepMind’s video generator: users type a prompt or upload an image, and the system returns short clips with camera motion, scene details and, in Veo 3.1, native audio. Google says Veo 3.1 is available in Flow and the Gemini app, while developers can use it through the Gemini API and Vertex AI. (deepmind.google) (ai.google.dev) (docs.cloud.google.com) The rollout has widened in stages. Google introduced Flow in May 2025 as an AI filmmaking tool built around Veo, Imagen and Gemini, then said in October 2025 that Veo 3.1 had reached Flow, the Gemini app, the Gemini API and Vertex AI with richer audio and more narrative control. (blog.google) (developers.googleblog.com) Google has also been threading Veo into work software, not just creator tools. In January 2026, the company said updated Veo 3.1 features would be available in the Gemini app, YouTube Shorts, Flow, the Gemini API, Vertex AI and Google Vids, and in February it said Business and Enterprise Workspace plans had promotional access to Veo 3.1 video generation in Vids, Flow and the Gemini app. (blog.google) (workspaceupdates.googleblog.com) That distribution matters because Google is selling Veo as production software, not a lab demo. Vertex AI documentation now lists text-to-video, image-to-video, first-and-last-frame generation, object insertion, object removal and video extension as product features for enterprise users. (docs.cloud.google.com) Google has been publishing usage numbers to show demand. In October 2025, the company said creators had generated more than 275 million videos in Flow in the first five months after launch. (blog.google) The hardware story sits underneath the software push. At Cloud Next 2025, Google introduced Ironwood, its seventh-generation Tensor Processing Unit, and said it was the first TPU designed specifically for inference, the step where a trained model actually produces an answer or a video. (blog.google) Google has tied the next phase to even newer chips. At Cloud Next 2026 on April 22, Google said its roadmap now stretches “to our eighth-generation Tensor Processing Units and beyond,” extending the same cloud-and-hardware pitch that supports Gemini and Veo-style media workloads. (blog.google) DeepMind has also added more formal documentation around the model family as access expands. Its model-card index shows Veo 3 was updated in January 2026 and Veo 3.1 Lite was added in April 2026, a sign that Google is segmenting video models by speed, cost or use case as it broadens distribution. (deepmind.google 1) (deepmind.google 2) The immediate next step is not a single consumer launch but a wider spread across Google’s stack. Veo now sits in the Gemini app, Flow, Google Vids, the Gemini API and Vertex AI, which makes the rollout look less like a one-off release and more like a new Google product layer for making video. (deepmind.google) (ai.google.dev)
Key numbers
- (deepmind.google) Veo is Google DeepMind’s video generator: users type a prompt or upload an image, and the system returns short clips with camera motion, scene details and, in Veo 3.1, native audio.
- Google says Veo 3.1 is available in Flow and the Gemini app, while developers can use it through the Gemini API and Vertex AI.
- Google introduced Flow in May 2025 as an AI filmmaking tool built around Veo, Imagen and Gemini, then said in October 2025 that Veo 3.1 had reached Flow, the Gemini app, the Gemini API and Vertex AI with richer audio and more narrative control.
- In October 2025, the company said creators had generated more than 275 million videos in Flow in the first five months after launch.
What happens next
- Google introduced Flow in May 2025 as an AI filmmaking tool built around Veo, Imagen and Gemini, then said in October 2025 that Veo 3.1 had reached Flow, the Gemini app, the Gemini API and Vertex AI with richer audio and more narrative control.
- In October 2025, the company said creators had generated more than 275 million videos in Flow in the first five months after launch.
- At Cloud Next 2025, Google introduced Ironwood, its seventh-generation Tensor Processing Unit, and said it was the first TPU designed specifically for inference, the step where a trained model actually produces an answer or a video.
Quick answers
What happened in DeepMind's Veo rollout?
DeepMind has started rolling out Veo, a text-to-video system, to creators and enterprise customers. The announcement frames Veo as part of DeepMind's push into generative video for production workflows. The rollout ties into bigger cloud and hardware narratives about scaling video generation and inference workloads (x.com).
Why does DeepMind's Veo rollout matter?
Google DeepMind has started putting Veo, its text-to-video model, into the hands of creators, developers and enterprise customers through Google’s own apps and cloud tools. (deepmind.google) Veo is Google DeepMind’s video generator: users type a prompt or upload an image, and the system returns short clips with camera motion, scene details and, in Veo 3.1, native audio. Google says Veo 3.1 is available in Flow and the Gemini app, while developers can use it through the Gemini API and Vertex AI. (deepmind.google) (ai.google.dev) (docs.cloud.google.com) The rollout has widened in stages. Google introduced Flow in May 2025 as an AI filmmaking tool built around Veo, Imagen and Gemini, then said in October 2025 that Veo 3.1 had reached Flow, the Gemini app, the Gemini API and Vertex AI with richer audio and more narrative control. (blog.google) (developers.googleblog.com) Google has also been threading Veo into work software, not just creator tools. In January 2026, the company said updated Veo 3.1 features would be available in the Gemini app, YouTube Shorts, Flow, the Gemini API, Vertex AI and Google Vids, and in February it said Business and Enterprise Workspace plans had promotional access to Veo 3.1 video generation in Vids, Flow and the Gemini app. (blog.google) (workspaceupdates.googleblog.com) That distribution matters because Google is selling Veo as production software, not a lab demo. Vertex AI documentation now lists text-to-video, image-to-video, first-and-last-frame generation, object insertion, object removal and video extension as product features for enterprise users. (docs.cloud.google.com) Google has been publishing usage numbers to show demand. In October 2025, the company said creators had generated more than 275 million videos in Flow in the first five months after launch. (blog.google) The hardware story sits underneath the software push. At Cloud Next 2025, Google introduced Ironwood, its seventh-generation Tensor Processing Unit, and said it was the first TPU designed specifically for inference, the step where a trained model actually produces an answer or a video. (blog.google) Google has tied the next phase to even newer chips. At Cloud Next 2026 on April 22, Google said its roadmap now stretches “to our eighth-generation Tensor Processing Units and beyond,” extending the same cloud-and-hardware pitch that supports Gemini and Veo-style media workloads. (blog.google) DeepMind has also added more formal documentation around the model family as access expands. Its model-card index shows Veo 3 was updated in January 2026 and Veo 3.1 Lite was added in April 2026, a sign that Google is segmenting video models by speed, cost or use case as it broadens distribution. (deepmind.google 1) (deepmind.google 2) The immediate next step is not a single consumer launch but a wider spread across Google’s stack. Veo now sits in the Gemini app, Flow, Google Vids, the Gemini API and Vertex AI, which makes the rollout look less like a one-off release and more like a new Google product layer for making video. (deepmind.google) (ai.google.dev)
