What You Might Have Missed in the World of AI - May 2025
Another Landmark Month for AI
May 2025 has been another landmark month for AI. If you blinked, you might’ve missed some big plays. Here’s a quick catch-up of the top 10 that stood out to me…
1. $100bn AI Infrastructure Bet in the UAE
OpenAI and Abu Dhabi-based G42 announced “Stargate”, a gigawatt AI supercomputing cluster. The goal? To position the UAE as the world’s first fully AI-integrated society, with ChatGPT deployed across sectors.
2. OpenAI + Jony Ive’s hardware startup io
A new AI-powered consumer device is on the horizon, following OpenAI’s $6.5bn acquisition of io, led by former Apple design legend Jony Ive. Think post-smartphone. It’s hush-hush now, but it could redefine how we interact with AI daily.
3. Google’s Gemini 2.5 Pro + AI Mode
With “AI Mode”, Google now offers Gemini-powered search that feels more like a chat than a query. Combined with Gemini 2.5 Pro’s reasoning upgrades, this marks a serious step toward AI assistants for everyone.
4. Anthropic debuts Claude 4
Claude 4 (Opus and Sonnet) ups the ante in reasoning, context retention, and safe usage. It introduced AI Safety Level 3, a framework to manage risk with increasingly agentic models. These are models that act, not just answer.
5. Google’s Veo and Imagen 4 take on Sora
Google entered the generative media race with Veo 3 (video) and Imagen 4 (image). The results are strikingly realistic. Veo now rivals OpenAI’s Sora. Expect creative tools to go fully multimodal by year-end.
6. AI-powered Search hits mobile
Behind the scenes, Apple is rumoured to license Google’s Gemini for on-device AI, marking a rare alignment. This means Gemini could power hundreds of millions of iPhones; game-changing scale for AI inference.
7. Anthropic launches ‘AI for Science’ program
In a welcome move for researchers, Anthropic is giving away free API credits for use in biology, genomics, and life sciences. Expect a wave of AI-native research papers by year-end.
8. OpenAI introduces o4-mini
OpenAI launched new lightweight models (o4-mini) and quietly made progress on standardising tool use with MCP, enabling smoother integration of external tools and agents across AI systems.
9. Google launches AI Ultra Subscription Plan
Gemini users can now opt into “AI Ultra”, a premium service tier for early access to experimental features and larger model contexts - akin to OpenAI’s ChatGPT Plus Pro tier.
10. Meta (quietly) updates Llama 3 family
While less splashy, Meta continued its steady open-source push with Llama 3.1 optimisations, boosting local inference and multimodal support. Not disruptive - but it quietly powers thousands of open models globally.
May wasn’t just about bigger models. It was about scale, safety, infrastructure, and interfaces. The convergence of AI + hardware, and the normalisation of agentic behaviours, is pointing toward an always-on AI layer that’s more accessible, multimodal, and mission-critical than ever.