According to Bloomberg, the surge in demand for AI chips has created a global parts shortage whose cost is now seeping into UK consumer prices. The scramble for enterprise GPUs and high-bandwidth memory is lifting the price of downstream goods — appliances, electronics — and is expected to help push August's CPI higher.
Every version of the boom's bill this page has tracked — the trillion-dollar valuations, the $500 billion capex plans, the price war between labs — has so far played out inside capital markets and corporate balance sheets. What is different about today's number is that the cost has surfaced at an ordinary shopper's till for the first time. Just as South Korea's memory shipments rose 277% for AI while logic fell, supply is being drawn up into compute demand, leaving less for everything else. That knock-on lands on the invoice of a buyer with no interest in AI.
The question worth asking is how far the causality has actually been established. UK inflation braids together currency, energy and tariffs, and AI cannot be isolated as the sole culprit. But the mere fact that a channel — AI capex moving real-economy prices — is now visible in the data marks a qualitative break from earlier phases. How central banks read that channel is the next argument.
According to Bloomberg, bond traders are on edge over roughly $70 billion of hidden credit backstops propping up lending to AI companies — support that lenders keep off the visible page. Part of the money flowing into data centres and capex, the argument goes, is stacking up risk outside formal balance sheets.
This is the flip side of the same coin as today's front page. A cost that surfaces in the real economy as prices, and a risk that sinks into the financial system as a chain of credit — the boom's bill is circulating along two routes. Read alongside the fund that lost $15 billion in July tied to Situational Awareness, or NVIDIA lending against its own demand, it lands on the same question: how long can the money loop hold. This is not, however, a forecast of collapse. The problem is that the backstops are not visible; the size alone does not automatically mean crisis.
Alibaba: Alibaba AI Models Hit 3 Billion Downloads, Passing Meta, Google· 21h ago
Apple: EXCLUSIVE: Apple trains its own AI model for China market with Alibaba's support, sources say· yesterday
ChatGPT lost 22 points of web share in a year· 6h ago
NVIDIA: Nvidia's $500 Billion Plan Envelops Wall Street in Its AI Frenzy· yesterday
Malaysia profits from data centre boom· 3h ago
Singapore Inc hopes AI access will stop finance high-flyers moving to Hong Kong· 5h ago
Big Manufacturers Find New Demand in Equipping AI Data Centers· 12h ago
South Korea Chip Data: Memory Up 277%, Logic Down; AI Buys Bandwidth, Not Compute· 18h ago
Is the AI boom a good bubble? The legacy that frenzied mega-investment leaves behind· 5h ago
US to warn partners they cannot join AI coalition if they sign with BeijingNEW
Safety testing was an obscure part of building AI. Then models went rogue.· 11h ago
Secondhand booksellers in UK and Ireland suspect AI firms behind 'strange' bulk orders· 23h ago
AI cheating, leaked papers and marking errors: how exam protests went globalNEW
Amazon: Amazon Can Use Your Twitch Content to Train Its AI—Unless You Opt Out· 22h ago
Exclusive | U.S. Urges Apple Not to Buy Chinese Memory Chips· yesterday
Japan to back Southeast Asian universities on AI research for agriculture and disaster responseNEW
AI video shows kamikaze pilots smiling; a former pilot: 'I don't want to see it'NEW