Time for a catch-up!

OK … I’ve been remiss the last 4 weeks as I’ve been distracted by a major new project and it’s been half-term, so my time has been limited.

However, the crazy world of AI continues to spin, and as per usual OpenAI has effectively dominated the headlines for most of the period. I’ll apologise now as this week’s newsletter is likely to read like an OpenAI press release, but it’s not surprising, as I will detail.

For me, the only key distraction was Nvidia becoming the first company to reach the market value of $5 Trillion. Given the fact that 5 years ago, few people outside the computer gaming world had even heard of them, it underlines their meteoric rise, particularly when you realise that its value rose 51% this year alone, and Microsoft has only just topped the $4 Trillion valuation mark.

Anyway, let’s dive right in on all of the news from last week and find out more about those moves:

Big Sharks vs. small fish

Again apologies, but this is mainly going to be about OpenAI.

So, let’s start with the big news of the last few weeks - OpenAI has officially become a public benefit corporation, changing how it handles profits and public good. As it now stands, the new OpenAI Foundation holds $130 Billion in equity and will give $25 Billion to health and AI safety projects.

The company also renegotiated its partnership with Microsoft, with Microsoft’s stake dropping from 32.5% to about 27%, with a value of around $135B after new funding, and Microsoft keeps tech rights until 2032.

For OpenAI, this ends years of legal and internal debate and allows it to focus on what it needs to do. For OpenAI CEO Sam Altman that means hitting the target of $20 Billion in annual revenue by the end of 2025.

The company has $1.4 Trillion in investment commitments for building data centres over the next eight years, underlining the huge growth in AI infrastructure and services. Altman said that he thought revenue could reach hundreds of billions by 2030.

They clearly have plans to expand beyond software, to include enterprise services, AI-powered devices, and even scientific research projects. This means that they are likely to become a major power in AI hardware, cloud services, and research, shaping how the world uses artificial intelligence in the next decade.

To underline that position, they also released a number of new tools and models, setting out the direction for the future. All of the releases were significant in their own way, but I’ll start with Atlas, their new AI-powered browser that brings ChatGPT directly into web browsing.

It helps users see, remember, and act on websites, aiming to make the internet more interactive and personalised. Atlas has a sidebar assistant and built-in Agent mode for completing online tasks, and it can remember visited sites to offer tailored answers and automate daily actions.

This marks OpenAI’s first real step into the browser world, setting up competition with other AI browsers like Comet and Dia, and hints at how browsing may soon move from passive searching to active, AI-driven interaction.

Next was the new update for Sora 2, which lets users create avatars that can be used in multiple AI-generated videos. The feature, called Character Cameo, comes with other new tools like video clip linking and a ranking board for top videos and avatars.

This means users can now build reusable avatars for their videos, which effectively means that you could create a library of characters and create your own movie, confident that the avatars will remain consistent.

And then there was their release of OpenAI has launched two open-weight versions of the gpt-oss models introduced in August. These models offer more transparency, control, and adaptability for users and are currently in research preview, allowing users to adjust them based on their own safety rules.

This is a major step toward safer AI systems, giving organisations tools to understand and explain AI decisions, reduce online abuse, and address ethical concerns about AI’s rapid growth, all of which is relevant under the EU AI Act.

All of this is thought to be in preparation for an Initial Public Offering (IPO), which could see OpenAI launch on the stock market with an initial valuation of $1 Trillion. The potential of an IPO launch may also be partially behind OpenAI’s decision to offer Indian users one year of free ChatGPT Go access starting November 4, aiming to boost adoption in its second-largest market. Investors will be interested in their long-term plans to increase revenue, and gaining a larger foothold in this market goes a long way to achieving that.

Perhaps not surprisingly, there was also other news not involving OpenAI in the last 4 weeks, the most important of which I outline below.

Microsoft rolled out an update to Windows 11, involving deeper AI integration: voice activation with “Hey Copilot”, enhanced “Copilot Vision” across markets, and a new “Copilot Actions” feature allowing the assistant to perform tasks (with user permission) such as restaurant reservations or grocery ordering. There was even a “Gaming Copilot” introduced on Xbox‑ally handheld consoles for in‑game assistance. This moves Microsoft Copilot away from being a standalone chatbot towards being embedded in the operating system and user workflow.

Microsoft also reported stronger‑than‑expected cloud revenue, driven in part by growing demand for its AI infrastructure. However, its AI‑spend is outpacing revenue growth, raising investor concerns about the margin impact.

Microsoft announced that it has created a MAI Superintelligence Team to develop AI systems that outperform humans in specific fields, beginning with diagnostics. The team’s first goal is “medical superintelligence” within two to three years. It aims to diagnose diseases earlier to extend healthy expectancy through advanced reasoning models.

Microsoft’s approach is interesting as it focusses on avoiding self-improving AI, and focusing instead on specialist systems with minimal existential risk and clear real-world benefit, which is very different to the approaches by Meta and Safe Superintelligence Inc.

Google were also busy, announcing plans to take its AI data centres into space. The company’s new project, Suncatcher, aims to use satellite-based chips powered by solar energy to solve the growing power and resource limits of AI systems.

The goal is to build the first-ever space-based data centre, working 24/7 on clean solar power, and if successful, it might change how and where the world builds data centres. Given the concerns I’ve raised about the abilities to power all of the AI infrastructure and data centres currently being built, this may present a novel solution to the problem.

The acquisition amount rose from Google's initial $23B in 2024, which Wiz rejected, to $32B when talks resumed in early 2025.

Google also managed to finally buy Wiz, a company that provides cloud security solutions that scan and protect cloud infrastructure, making it a strategic asset for Google Cloud as enterprises migrate sensitive workloads to cloud environments requiring robust security frameworks. However, it was costly, with their initial bid of $23 Billion being rejected in 2024, before they finally closed the deal for $32 Billion this year.

This acquisition signals Google's aggressive push to compete with Microsoft and Amazon in enterprise cloud security, and for enterprise leaders, this consolidation could mean enhanced security tools within Google Cloud but potentially altered vendor landscapes as major platforms acquire independent security companies.

Legislation, policy and other news

There was also lots happening elsewhere, particularly around the issue of intellectual property.

CODA, a Japanese group that protects creative rights for studios like Studio Ghibli and Bandai Namco have asked OpenAI to stop using their copyrighted content to train its video AI model Sora 2.

I believe that OpenAI will agree to this, particularly given the fact that on 29th October, OpenAI made major changes to its product usage policy. It now restricts users from giving licensed expert advice, such as in medical or legal fields, using its tools. Users cannot provide professional advice unless a licensed expert is involved.

The new rules also ban harmful or violent content and automatic decisions in sensitive areas like jobs or finance, and OpenAI says it will keep updating its policies to balance freedom and safety. Combined with the two new models mentioned earlier, these updates demonstrate OpenAI’s desire to protect users from misuse and legal risks, while keeping AI tools safe, fair, and transparent for everyone.

Perplexity also signed a multi-year licensing agreement with Getty Images, giving it permission to display images across its AI-powered tools. The agreement follows multiple accusations of content scraping and a recent lawsuit from Reddit, giving Perplexity official permission to display Getty’s images and address challenges over its data sourcing. This partnership approach signals a growing trend toward licensed content deals to mitigate legal risks.

Despite that, Perplexity AI still received a threat from Amazon demanding that the startup block its AI agent on the Comet browser from shopping on Amazon's platform on users' behalf. Amazon repeatedly requested that Perplexity get rid of the retailer from its Comet experience, citing a "significantly degraded shopping and customer service experience" that the AI agent provides to users.

Some suspect this is less to do with the customer experience, and more about limiting competition, so that it can promote the competing tools it is developing, namely "Buy For Me" and "Rufus," suggesting the law challenge may protect market position as much as address user experience concerns.

In wider news the European Data Protection Supervisor (EDPS) released updated guidelines for generative AI used by EU institutions, including new definitions, compliance checklists, and clarifications on legal roles/responsibilities.

This comes as two surveys indicate the size of the AI market in the EU. One survey indicates that 30% of people employed in the EU (60 million), now use AI every day in work, whilst another indicates that roughly 46 % of Europe’s 23 million SMEs are already using basic AI tools (e.g., for marketing or operations) even though many lack robust digital infrastructure.

That’s all for now from me at the end of yet another busy week … so stay informed, stay critical, and wherever possible - stay ahead.

Regards

Tom Carter