The headline announcement at Google I/O 2025 — the company’s annual developer conference which took place today — wasn’t a futuristic demo or another chatbot update. It was a price tag. Google’s new AI Ultra subscription costs $249.99 per month, offering early access to its most powerful Gemini models, tools and a growing suite of automation features.
That price makes it one of the most expensive flat-rate AI subscriptions on the market. But it’s not without precedent. Both OpenAI’s Pro plan and Anthropic’s Claude Max tier plans are $200 per month. What sets Google apart, however, is the bundling.
The Ultra AI plan includes the company’s best Gemini model, access to experimental tools, 30TB of cloud storage and YouTube Premium — all wrapped in a publicly advertised plan aimed at developers and power users. It also marks a shift in how Google wants to position Gemini — not just as a behind-the-scenes engine powering search and Workspace, but as a standalone, monetizable product in its own right.
Elsewhere, Google revealed a mixture of AI-focused and hardware announcements:
Android XR Brings Competition To Meta
Google I/O 2025 was light on hardware news, but the company did show off its Android XR platform. XR, or extended reality, is a catch-all term that includes augmented reality (AR), virtual reality (VR) and mixed reality (MR) — technologies that blend digital content with the real world through headsets or smart glasses.
The new Android XR platform is designed to support this spectrum of experiences, starting with Samsung’s upcoming Project Moohan headset. Built in collaboration with Google and Qualcomm, Moohan is a fully immersive device — similar in form factor to a VR headset — designed for entertainment, productivity and AI-assisted interaction.
Slated for release later this year, it will run Android XR and include deep integration with Gemini for navigation, guidance and real-time help in virtual environments.
Google also previewed a series of Android XR smart glasses built in collaboration with Gentle Monster, Xreal and Warby Parker. These glasses integrate microphones, cameras and displays to enable features like live translation, object recognition and visual overlays — all powered by Gemini. Unlike earlier Glass prototypes, the focus here is on consumer-ready styling, with fashion-forward designs that aim to bring AI wearables into everyday life.
These XR announcements position Google more squarely in competition with Meta, which already has a consumer presence through its Ray-Ban Meta smart glasses — offering features like hands-free image capture, audio playback and integration with Meta AI. Meta is also developing a more advanced AR headset, codenamed Orion, which is expected to push further into immersive spatial computing.
Imagen 4: Advanced AI Image Generation
AI image generation tools are one of the most accessible, mainstream applications of AI, as demonstrated by Chat GPT’s AI image generator updates. At I/O 2025, Google introduced Imagen 4, its most advanced text-to-image model to date, promising significant improvements in image quality, like rendering intricate details such as fabrics, water droplets and animal fur with remarkable realism.
Another standout feature is its enhanced ability to generate accurate and legible text within images, addressing a common limitation in previous models. This advancement makes Imagen 4 particularly useful for creating visuals like posters, greeting cards and slides that require precise typography.
Project Astra
Project Astra, a prototype multimodal agent shown live at I/O, was arguably the most futuristic part of the keynote. Point your phone at a cluttered desk and Astra can describe what it sees, pull relevant information from your environment and maintain an ongoing voice conversation.
The demo recalled earlier hype around AR assistants and computer vision, but Astra’s real-time understanding and contextual awareness set it apart. It wasn’t just identifying objects — it was reasoning about them, combining visual input with memory and audio to act more like a persistent assistant than a reactive chatbot.
Astra isn’t ready for widespread release, but parts of it — including enhanced visual understanding — will roll out to Gemini apps and Android later this year.
AI Mode In Search is Live
The new AI Mode in Google Search is now live for all users in the US. Now, instead of 10 blue links, users are presented with AI-written summaries, contextual explanations and citations — a move clearly designed to stem the rise of people turning to ChatGPT or Perplexity for instant answers.
But this shift also risks Google’s lucrative business. With Search ads still accounting for a large share of Alphabet’s revenue, replacing link-based queries with AI answers reduces the number of clickable ad slots. Google is experimenting with new ad formats inside these AI summaries, but the balancing act is delicate.
Google, it seems, is hoping that AI will expand search usage, not cannibalize it. The bet is that users will ask more complex, high-value questions when the answers are richer and more interactive — and that advertisers will follow.
Gemini 2.5
Google’s DeepMind team claims that Gemini 2.5 Pro is its most advanced AI yet, with strong performance on reasoning tasks and a clear emphasis on code generation and debugging. In internal benchmarks, it outperformed OpenAI’s GPT-4 on several software development tests and introduces an experimental Deep Think mode that lets the model reason through complex problems step-by-step.
Efficiency has been improved as well. A lightweight Gemini variant runs 22% faster than the previous release, making it more viable for integration into mobile and edge applications. This is a key development, considering that Google’s entire OS and hardware stack — from Android to Pixel to ChromeOS — is being reworked around embedded AI assistants.
Google Beam
Google also showed off Beam, a high-end 3D video chat system built from its ongoing Project Starline. The system uses depth sensors and AI-driven rendering to recreate virtual face-to-face interactions. It’s initially targeted at enterprise use, with HP as a hardware partner and slots into Google’s broader push to reimagine remote work through AI-enhanced collaboration.
Beam highlights how Google sees AI not just as a productivity tool, but as a means of making virtual interactions feel more human. It also signals that Google hasn’t given up on ambitious hardware, even as its messaging around Pixel devices was notably absent this year.
Google I/O 2025: The Future
Overall, Google I/O 2025 saw a smattering of hardware announcements (some more solidified than others), but the real focus was, unsurprisingly, on AI. The keynote solidified Gemini’s presence beyond a mere test bed — t’s now embedded in products, packaged into subscriptions and tied into how Google intends to make money going forward.
There are still big unknowns around usage limits, developer adoption and whether consumers will stick with one assistant over another. But Google has moved decisively and time will tell how its strategy plays out against other established players.
