Google has restricted Meta's access to its Gemini AI models after Meta asked for more computing capacity than Google could supply, according to the Financial Times. The fallout hit some of Meta's internal AI projects, pushed staff to ration resources, and accelerated Meta's push toward its own in-house models. This matters because it's the clearest sign yet that AI's bottleneck isn't ambition or money — it's raw physical capacity, and even the richest companies on Earth are now standing in line for it. The real story here isn't Google versus Meta. It's that the AI boom has officially collided with the limits of the world's data centres and chip supply, and every company building an AI strategy now has to plan for scarcity, not abundance.

The must read articles that are moving the business landscape today.

Amazon's 2026 Prime Day pulled in $26.4 billion, up 9.3% on last year, as deep discounts drove sales of electronics and big-ticket appliances. But analysts noted the average order value actually fell, suggesting shoppers leaned on promotions and tax refunds rather than confident spending. For retailers, the lesson is blunt: people are still buying, but only when you make it impossible to say no.

Australian firm Firmus Technologies has struck a deal with Nvidia to deploy 170,000 GPUs in Indonesia by 2028, aiming to generate up to $30 billion in revenue over six years. The pitch is simple: cheaper AI access for smaller players who can't compete with Big Tech's chequebooks. If it works, it could reshape who gets to build AI products at all — not just who gets to use them.

Austria has urged the EU to court AI company Anthropic after US restrictions limited foreign access to its top models, arguing this would protect Europe's digital sovereignty. It's a small move with a big subtext: governments are starting to treat frontier AI access as a national security issue, not just a tech trend.

Volkswagen is reportedly ending its autonomous driving partnership with Bosch after roughly €1.5 billion failed to produce technology that could compete with rivals, according to Bild. VW now plans to find a new partner by September. It's a costly reminder that even Europe's biggest automaker can't out-engineer the AI talent gap by spending alone.

A daily score of business conditions (scored out of 100), with a breakdown of what’s driving it.

The Daily Business Index — our daily 0-to-100 reading of global business health, where 50 is neutral — dipped to 49.5 today, a step below neutral. The trigger: a sharp slide in Apple's share price on Wall Street spread overnight into Asia, where South Korea's main index was halted by a circuit breaker and Japan's and China's markets tumbled too, as investors rethought how much further the AI boom can run. Working the other way, oil prices kept falling as shipping through the Strait of Hormuz normalised, and US jobless claims hit a four-week low — both signs the wider economy is holding up better than the tech sell-off suggests.

The one story from the archives worth remembering today, so you can spot the pattern before everyone else does.

A year before this date, Nike reported a fiscal fourth quarter that saw profits collapse 86%, as the company absorbed its biggest financial hit yet from a turnaround plan and warned tariffs would cost it roughly $1 billion. It read at the time like a Nike problem. It now reads like an early warning for every company that sources globally — the tariff math eventually catches up with everyone's margins. Funny how a sneaker company's earnings call ended up forecasting half of corporate America's 2026 headaches.

The one piece of business jargon demystified today, so you can nod along with confidence.

What it means: This is the raw processing power — chips, servers, data centres — needed to run AI models. It's the actual physical limit on how much AI a company can build or offer, no matter how much money it has. Think of it as electricity for an AI factory: without enough of it, nothing runs, which is exactly the wall Meta hit with Google this week.