News vs. Noise: Rising Bond Yields, Oil Prices, and AI Credit Risk

Government bond yields surged to multi-decade highs across the U.S., Japan, U.K., Germany, and France last week, and it wasn’t just a Washington story.

The 10-year Treasury yield climbed to a weekly high near 4.80%, its highest level since mid-January 2025, before easing to about 4.76% by Friday [1][9]. The move wasn’t confined to the U.S.: Japan’s 10-year yield touched 3% for the first time since 1996, the U.K.’s 30-year yield hit its highest level since 1998, and yields in Germany and France reached decade-plus highs, all in the same week [2].
The common thread across these markets is fiscal: governments made spending commitments when borrowing was nearly free, and now have to finance them with inflation still elevated and oil prices climbing on renewed U.S.-Iran tensions near the Strait of Hormuz [2][4]. The U.S. Treasury has also recently doubled its planned long-end buyback program, though the larger repurchases haven’t started yet [3]. A newer complication is layering on top of that: AI-related corporate debt issuance is increasingly competing with Treasuries for investor capital [2][4].
For balanced portfolios, this matters because higher long-term yields raise the discount rate applied to future earnings and can pressure both bond and equity valuations at the same time [2]. We’re watching whether this stays an orderly repricing of growth and inflation expectations, or whether it starts to show up as stress in credit spreads and equity multiples, and we continue to favor high-quality, laddered fixed income over reaching for yield in this environment.

A renewed flare-up in the U.S.-Iran conflict pushed oil higher through last week, and diesel hit a record price on Friday.

Brent crude climbed from about $91 a barrel early in the week to a weekly high near $95, after the U.S. and Iran exchanged their largest barrage near the Strait of Hormuz since July [1][5][7]. Oil held near that level through the rest of the week, and on Friday, U.S. diesel prices hit a record $5.85 a gallon [5][6][9].
The complication for policymakers is that this is a supply shock, not a demand story, and a Fed rate move doesn’t fix it directly. No level of the federal funds rate reopens a shipping lane. That leaves the Fed weighing a cooling labor market, which argues for patience, against oil-driven inflation, which argues for vigilance [7].
We’re watching Brent crude and diesel prices as a real-time gauge of how much the Hormuz situation is feeding into broader inflation, separate from anything the Fed does with rates, and a sustained move higher in energy prices would argue for maintaining some inflation-protection exposure in client portfolios even as growth data soften.

Broadcom’s earnings validated enormous AI chip demand last week, but they also surfaced a financing and customer-concentration risk worth watching.

Broadcom Inc. (Nasdaq: AVGO), a semiconductor and infrastructure software company, reported $29.6 billion of quarterly revenue last week, with AI semiconductor revenue up 221% year-over-year to $16.7 billion, and CEO Hock Tan guided AI chip sales to roughly double to $115 billion in fiscal 2027 and double again to $230 billion in fiscal 2028 [7][8].
Underneath that growth, Broadcom disclosed that Anthropic is on track to become its largest custom-chip customer in 2027, with OpenAI close behind as its second-largest customer in 2028. Both relationships lean on a financing platform involving Apollo Global Management and Blackstone, built to help the two AI labs bridge the gap between their current cash flow and the capital AI infrastructure requires [7][10]. Broadcom separately disclosed up to $29 billion of lease obligations it could be on the hook for if data-center tenants default [4][11]. That’s a credit and concentration story sitting inside a demand story.
AI infrastructure demand looks durable, but the financing behind it is increasingly complex and concentrated among a small number of cash-flow-negative labs [7][9]. We’re watching how much of the AI capex boom is being funded with debt tied to customer contracts rather than free cash flow, since that’s the kind of structure that can unwind quickly if growth disappoints.

Reading last week’s Treasury selloff as a uniquely American credibility problem misses that every major government bond market moved the same way.

With the 10-year Treasury yield near its highest level since January 2025, some commentary framed the move as a sign that U.S. government debt was losing its footing. That read surfaced right as yields were climbing across every major developed bond market at the same time.

The reality is that Japan, the U.K., France, and Germany experienced the same or worse moves in their own government bond markets during the same week, which tells us this is a broader repricing of fiscal risk and inflation expectations across the developed world, not a story unique to U.S. credibility [2][4]. Treasuries remain the deepest and most liquid safe-haven market available, even as that whole neighborhood gets more expensive to live in [4].

This doesn’t change our approach to core fixed income allocations. We aren’t rotating away from Treasuries on the theory that they’ve lost their safe-haven status, since the global nature of the move argues against that read.

Treating the Fed’s next rate decision as the fix for the inflation problem misunderstands that a supply shock isn’t something a rate move can solve.

September hike odds didn’t move on one data point. ADP’s report on September 2 showed private employers added just 38,000 jobs in August, the slowest pace since January, and Bureau of Labor Statistics data showed hiring pulling back in professional and business services [12][13]. That softening trend, combined with Fed Governor Christopher Waller’s September 3 comments that he’d support holding rates steady if inflation data keep cooling, pushed the odds of a September hike back toward a coin flip; stocks rallied and the 10-year eased to roughly 4.76% [9][14][15].

But neither a Fed pause nor a hike reopens the Strait of Hormuz, refills the Strategic Petroleum Reserve, or brings diesel down from its September 4 record of $5.85 a gallon [6][7][9]. Monetary policy is a demand-side tool being asked to address a supply-side problem, and it can only do part of that job.

This doesn’t change how we’re positioned. We aren’t treating one week’s Fed-odds swing as a green light to add risk, because the underlying tension between a cooling labor market and elevated energy prices hasn’t actually resolved.

Blowout AI earnings from Broadcom, Dell, and Nvidia got read last week as proof the AI-spending scare was over — but that take skips over what’s building underneath the headline numbers.

Alongside Broadcom, Dell Technologies Inc. (NYSE: DELL) reported record AI-server orders of $60.9 billion and a record $95 billion backlog last week, and Nvidia Corporation (Nasdaq: NVDA) agreed to acquire Hugging Face, the open-source AI model platform, for $12.93 billion [9][16][17]. Taken together, coverage of these results treated them as an all-clear signal for AI-related stocks broadly [9].

The demand numbers are real, but they don’t resolve what’s building underneath them: financing concentrated in a handful of customers, debt-funded infrastructure buildouts, and a momentum factor that has already cracked [4][7][11]. As of August 31, the S&P 500 Momentum Index was down more than 9% since July 1, even as the S&P 500 itself gained roughly 2.8% over the same stretch [18][19]. Strong headline growth at the chip layer doesn’t automatically translate into durable returns for every stock riding the AI theme.

This doesn’t change our approach to AI-related exposure. We continue to favor established, well-capitalized names with pricing power and diversified customer bases over chasing every name tied to the AI narrative, and we aren’t reading one strong earnings week as a signal to add blind AI beta.