Time for a catch-up!

Well, it’s been another crazy week, where the expenditure of the key players has hit new levels. Whether it’s paying hundreds of millions of dollars to hire key staff, or spending tens of billions on infrastructure, the race for dominance of the AI eco-system continues unabated.

That was also reflected with Nvidia’s valuation surpassing $4 Trillion USD, underlining Nvidia’s dominance in AI hardware. New research also estimated the value of the global AI market at $391 billion, growing at a 35.9% compound annual growth rate. Also, 87% of organisations reported that AI gives them a competitive edge, although this contradicts to a large degree the Gartner report conclusions from last week.

Regardless of who is correct, one has to wonder how and when this will start to provide a return on all of the considerable investment made to date, and what safeguards (if any) will be put in place.

So, let’s dive right in …

Big Sharks vs. small fish

Well, Meta have continued to dominate the headlines as more news emerges of Mark Zuckerberg’s aggressive campaign of buying up companies, poaching staff, or both, in order to create his planned Team Super-Intelligence. Yet another person signed up to a $100 million signing up bonus, this time Ruoming Pang, the former chief of Apple’s AI development programme.

With a joining bonus of $100 million, an annual salary of $100 million, plus a huge META stock grant, tied to loyalty & performance, it would clearly be difficult to say no. Rumours are that this offer has been made to a total of 11 “targeted” individuals. Unfortunately, I was missed of the list of people to be hired 🤣🤣🤣

The infrastructure side of things also continued apace, with CoreWeave announcing a $9 billion acquisition of Core Scientific to consolidate its AI-focused data-centre infrastructure, although investor confidence appears cautious. Groq also partnered with Equinix to launch a Helsinki AI inference hub and European AI Data Centre, reducing latency for EU customers and boosting on-­shore compute capacity. 

Microsoft also highlighted its planned investment in AI infrastructure via a $80 billion capex plan although the downside is that it will cut roughly 9,100 employees (about 4% of its workforce) as a result.

Additionally, Google’s AI policy head confirmed the company is investing heavily in AI infrastructure, with $75 billion allocated this year alone for data-centre expansion and compute capacity. There is clearly confidence in what Google are doing, as one major market trader raised Alphabet’s stock target from $200 to $220 after noting strong user adoption of Google’s conversational AI Mode, with 60% finding it more helpful than ChatGPT and 75% of paid users preferred it.

So, it’s not surprising that OpenAI plans to launch an AI-enhanced web browser in the coming weeks. Built on Chromium, it will feature agentic capabilities, such as booking and form-filling and aims to challenge Google Chrome’s dominance by offering deeper integration with user data and browsing behaviour. 

Amazon are also getting into the battle, announcing today (11th July) that customers as soon as next week, will be able to buy and deploy pre-built agents (call centres, finance bots, code fixers) directly inside the Management Console. Launch partners include Anthropic, which will sell Claude-powered agents under AWS credits. Agents inherit virtual private cloud networking and identity and access management roles out of the box, so no extra security work is required.

As I’ve mentioned many times already, there is a battle for market dominance, and the bigger players are looking to form partnerships with Governments in order to achieve that dominance. They are also offering lots of things for free or low-cost, on the basis that once you’ve signed into their infrastructure, who will not want to change providers constantly. As an example of this, Google Cloud launched a 12-week accelerator for UK AI startups and introduced improved data residency options under its Gemini for UK scheme.

Trump’s newly enacted "One, Big, Beautiful Bill" provides manufacturing tax credits of up to 35%, favouring chipmakers like Intel, Micron, Nvidia, and TSMC, so one could expect a further boost to US domestic AI-tech infrastructure. However, his proposal to impose 50% tariffs on copper imports (effective August 2025) threatens to increase costs and slow AI data centre development, as copper is essential for server infrastructure.

AI energy also made the news, with Elon Musk's xAI announcing plans to import a dedicated power plant, most likely powered by natural gas, in order to support its massive AI data centre, projected to draw up to 1.96 GW of power. Politico also warned of exploding electricity demand from AI data centres in the UK, underlining the urgent need for grid upgrades before 2030.

The good news though, is that energy efficiency is increasing by about 40%, with Stanford’s 2025 AI Index reporting that inference costs dropped by a factor of 280 times since 2022. Also of note was the finding that smaller open-source models are closing the gap with proprietary systems.

There’s also been some new models released, with Samsung introducing the Z Fold7 and Z Flip7 containing AI features powered by Google Gemini, including Drawing & Writing Assist and Outfit Suggestions, along with the AI‑enhanced Galaxy Watch8 featuring vascular load and antioxidant metrics.

xAI also launched two models named Grok 4 and Grok 4 Heavy. The second one is the company’s “multi-agent version” that offers increased performance but comes in at a hefty $300 USD per month. Musk claimed that Grok 4 Heavy creates multiple agents to work on a problem simultaneously, and then they all compare their work “like a study group” to find the best answer. Musk also confirmed Grok integration in Teslas this coming week.

Legislation, policy and other news

Grok was also in the news for less positive reasons this last week, with some of its responses and engagements with the public raising concerns due to the political and discriminatory nature of some of the posts, something that Musk has previously tried to position as merely underlining his strong anti-censorship stance.

It will therefore be interesting to see how the EU react, particularly following the release of The General-Purpose AI (GPAI) Code of Practice, in the lead up to the EU AI Act starting to come into force next month. Yes, it’s voluntary (for now), but I believe that’s it is only a matter of time before there is a major showdown between the Commission and xAI, particularly given the indications of antisemitism in some of the posts.

The shift was also illustrated by the coalition of EU publishers which filed an antitrust complaint against Google, alleging AI-generated summaries in search results are diverting traffic from content creators and harming their viability. 

Not wanting to be outdone, the leaders from the BRICS countries (Brazil, Russia, India, China, and South Africa) also urged stronger safeguards against unauthorised AI use at the Rio summit, calling for data-usage rules and fair compensation for content creators.

From the industry side, Anthropic released both an open-source AI transparency framework and a full research paper with code, aiming to foster trust and accountability in advanced AI development.

Finally, to end on a positive note again, Australian researchers have used AI to create in seconds custom proteins that kill superbugs like E. coli and can be used for treating some cancers, signalling a rapid advance in AI-powered biomedical discovery.

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

Regards

Tom Carter