AFS Markets Blog: Midday 30/07/2026

• Meanwhile in markets, yield curves are steepening following a disappointing performance by Fed Chair Kevin Warsh at yesterday's FOMC meeting. He failed to back up his prior hawkish rhetoric with a rate hike – and his unconvincing hawkish spin at the presser has given punters reason to react;

• US Treasury yields pushed another 2 bps higher at the long end this morning, with the 30-year yield now at its highest level since 2007. The Bund curve mirrored this steepening, with 2-year yields falling 2bps while 30-year yields rose 4bps. The dip in short-end Bund yields is particularly striking given that Brent crude has risen back above the $90 handle following new tit-for-tat strikes in the Middle East;

• The conflict between the US and Iran actually expanded geographically yesterday. Iran reportedly struck two US LNG vessels off the coast of Egypt – the first time this happened. According to the New York Times, Iranian officials intended the strike as a signal that shipping lanes across the globe could become targets if the conflict escalates further;

• Over in equities, the Stoxx 50 is up half a percent this morning while S&P 500 futures are up a quarter of a percent. While benchmark indices may be up, the AI trade continues to deflate. Our own AI index – containing hyperscalers and large chip companies – has now fallen over 11 percent in the past week and is now in bear market territory, 23 percent below last month's all-time high;

• Amid the decline in AI stocks, it is worth highlighting the Silicon Data LLM Token Expenditure Index – a closely watched usage-weighted index tracking the average price paid per million tokens across established large language models. Tokens – the actual input and output of LLMs – are the product AI labs actually sell to the end users of their models. Over the past two months, this index has fallen by nearly a third;

• The decline can stem from a combination of factors: end-users shifting toward cheaper or open-source models, the use of models that requires fewer tokens or less compute, or outright token price cuts. Either way, it signals a lower willingness to pay for extra computing capacity – part of the reason behind mounting doubts over the future profitability of AI labs. And of course, the question over when – or even whether – investors in AI data center infrastructure will earn their money back;

• Not all tech names are suffering amid the pullback. Apple – one of the few major tech giants that avoided going all-in on massive AI and data center capex – is up 4 percent over the past week. Apple now comfortably outperforms the other Magnificent 7 names – up 25 percent year-to-date. The only other two Mag7 names in the green are Google (up 7 percent) and Nvidia (up 2 percent);

• At the same time, it is worth highlighting the so-called AI losers – companies whose business models face disruption from large language models. Back in January, this group took center stage in equities as it suffered severe sell-offs following new AI model releases. Our AI Losers Index (comprising European software-as-a-service providers, publishers and staffing agencies) was down nearly 25 percent YTD at its March trough. But as market anxieties over AI resurface this index has surged nearly 20 percent over the past week – reaching its highest level since late January;

• Looking ahead, the calendar is full of data releases this afternoon, including the US PCE inflation reading for June. Eyes will also be on the Bank of England's interest rate decision – where rates are expected to be held steady.

‍