Time for a catch-up!

If you feel like it’s getting harder to keep up with AI developments, you’re not alone. This week has brought another avalanche of updates, from new models and upgraded co-pilots, to billion-dollar investments and long-term policy planning.

What’s most striking isn’t just the pace of development—it’s the scale. We’re now seeing infrastructure projects, legislative reforms, and multi-billion-dollar ventures stacking up alongside algorithm tweaks and model comparisons. AI is no longer the future. It’s the infrastructure of the now.

Big Sharks vs. small fish

OpenAI continued to lead headlines firstly with the finalization of a $40 billion funding round. That level of investment not only reshapes OpenAI’s future—it’s reshaping the entire market, with U.S. venture capital reaching $91.5 billion in Q1 alone. But the really big news was OpenAI’s introduction of a groundbreaking image generation feature in ChatGPT, greatly improving realism, editing capabilities, and style application. Popularity surged with "Studio Ghibli-style" images appearing everywhere, leading to 1 million new subscribers in a day, but this led to OpenAI’s servers almost melting down, forcing them to restrict usage whilst they brought more capacity online.

Google answered back with Gemini 2.5, their smartest model yet, with outstanding benchmarking in logic, science, and mathematics, and a 1 million-token context window, suggesting a race to corner long-form reasoning. Microsoft wasn’t far behind, adding enhanced Co-Pilot capabilities to Office 365, including Analyst, which can search across all of the other Microsoft apps and the web, integrating Python execution and deeper reasoning tools.

Meanwhile, Amazon launched 'Project Rainier' - a $100 billion plan to build the largest AI datacenter cluster, anchored by its own Trainium 2 chips. This is primarily about building an infrastructure to support Anthropic, as the battle for dominance of the AI space continues. Not all has gone smoothly though, with both Amazon and Apple delaying the upgrades of Siri and Alexa.

Meta was also in the news, with the release of new open-weight models, but more significantly a plan for undersea cable expansion, as they prepare for a globally distributed AI race. This is now yet another area where they are competing against Amazon, Google and Microsoft, all of whom are also investing in deep-sea cabling.

Whilst the big players dominated the headlines there was still lots of other news and releases, particularly in relation to image creation, with updates from Midjourney and Ideogram, and a new company named Reve appearing, which scores higher than others on the standard benchmarks. There were also continuing developments in China, with Alibaba releasing two new models and DeepSeek updating their underlying V3 LLM, reducing its dependency on GPUs.

Legislation, policy and other news

In terms of policy and legislation Google DeepMind released a comprehensive 145-page paper stressing the imminent potential of achieving Artificial General Intelligence (AGI) and the critical need for long-term safety planning. The document outlines primary risk areas associated with AGI and proposes mitigation strategies involving developer interventions, societal adjustments, and policy reforms.

Also, OpenAI’s image generation software again opened up the issue of intellectual property rights around the “copying” of the Studio Ghibli style. What many may have missed however is that there was also a privacy issue, as millions of people uploaded photos of themselves and others, which are now held by OpenAI for future use.

It’s not all end-of-the-World scenarios though, as on the medical front there was more positive news, with research demonstrating how AI-enhanced deep brain stimulation can assist those with Parkinson’s.

Also, amongst all of the noise, OpenAI also announced that they would be adopting Model Code Protocols (MCPs), which signifies an endorsement of the idea created by Anthropic.

What’s becoming clear is that the AI race is not just about computing power, but also the underlying infrastructure necessary to sustain it.

From Studio Ghibli-style image generators to automated medical diagnostics, we’re in the middle of the most consequential tech transformation since the internet.

Stay informed. Stay critical. And wherever possible—stay ahead.

Regards

Tom Carter