Time for a catch-up!

I’ve seen an increasing sense of “doom and gloom” in some articles over the past few weeks, regarding concerns that the interest in AI mirrors the previous Dotcom bubble.

Whilst I don’t think that it will go “pop”, I have expressed my own concerns about the unrealistic returns on investment that are being promised by everyone in the market, as the numbers simply don’t add up at present.

On the one-hand you have Microsoft, Google and Meta adding $350 billion to their market valuation this year, based in part on the rapid development of AI infrastructure, which is now running at over $200 billion spending this year alone. Yet even OpenAI are only estimating annual income this year of $5 billion, so you have to question at what point companies will increase their prices.

Also, you have to take into account the wider impact of AI on businesses like Google and Meta, both of whom rely to a large extent on advertising revenues. In fact, for Q2 Meta published an increase in revenue of 22%, powered by a large extent by more AI driven content.

Yet, new analytics data shows that AI assistants sent 1.13 billion visitors to major sites in June, up from 250 million a year ago. This potentially disrupts SEO playbooks by having agents talk with each other to move real traffic and revenue, thus undermining the traditional revenue markets.

So, with all of that in mind let’s dive right in on the other news from this week …

Big Sharks vs. small fish

OpenAI continues to build out its own ecosystem, in an effort to secure market control. A new toggle in ChatGPT lets users upload a syllabus or topic list and get daily lessons, flashcards and quizzes. Progress syncs across web and mobile, and admins can assign modules to teams. This is likely to kill loads of small micro-learning platforms, in the same way they have done with business meeting transcription software.

Apple also signalled an aggressive AI push via acquisitions, with Tim Cook
confirming that Apple had acquired around seven AI startups in 2025 and is open to larger takeovers. He also announced significant increases in AI-related infrastructure spending.

Clearly Amazon believe like many that wearable devices are potentially the way forward, buying Bee, a small San Francisco startup behind a $49.99 wearable wrist device designed to record and transcribe conversations. Bee CEO and co-founder Maria de Lourdes Zollo announced the acquisition on LinkedIn, saying that deal will “help us bring truly personal, agentic AI to even more customers.”

Google tested Opal, a tool designed to vibe-code, allowing it to generate a working web app with styling, data storage, and deployment hooks in minutes.

Perhaps with concerns growing regarding access to the right chips, Tesla also signed a long-term agreement with Samsung to produce next-gen AI6 chips in Texas, supporting Tesla’s ambitions in autonomy and robotics, coupled with moves to reduce reliance on Chinese suppliers.

Finally Anthropic announced that it is finalising a $5 billion round that would push its price tag to $170 billion, well above most public software vendors. The cash will pay for more H200-class GPU clusters and a larger multilingual Claude model scheduled for 2026.

Legislation, policy and other news

In legal news, Google signed up to the EU’s voluntary code of practice for AI, joining OpenAI and France’s Mistral. It is expected that Apple and Microsoft will also sign up, but Meta have already said no, and people don’t expect Grok to join anytime soon.

New research found Google’s AI-generated summaries have reduced click-through rates to publishers by up to 80%, prompting concerns about the future viability of online journalism. Which is why it is noteworthy that Amazon will pay $20–25 million annually to The New York Times for access to news, cooking, and sports content, signalling a shift toward paid AI training data models.

Also, in other publishing news Vogue published an AI-generated model in a Guess ad, marked as AI-produced, though barely legible. This sparked debate around ethics, representation, and the future role of human models in fashion.

There were also other reasons to be concerned about the potential impact of AI, with OpenAI’s latest AI assistant successfully clicking through a “I’m not a robot” checkbox on Cloudflare, raising ethical and safety concerns as the agent mimics human behaviour in real-time.

Finally, at the World AI Conference in Shanghai, Premier Li Qiang suggested establishing a UN‑led international organisation centred on AI governance and equitable development, aiming to counterbalance U.S. influence.

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

Regards

Tom Carter