- Long-dated Treasury yields have reached their highest levels in nearly two decades even as Fed policy rates are 175 basis points below their peak — a disconnect that Citadel Securities’s Nohshad Shah says reflects a market view that “policymakers, both the Fed and fiscal authorities, tend to take the easier route when faced with difficult choices,” creating a persistent risk for markets more broadly.
- The 30-year Treasury yield climbed to a 19-year high above 5.28% on Monday after bond markets lowered expectations for a Fed rate cut in September, following data suggesting both inflation and consumer demand are easing — but Shah cautions against reading this as an all-clear, noting more than 55% of core goods prices are still rising and the September Fed meeting is a “line-ball call.”
- On AI, Shah offers a contrarian near-term thesis: he sees the investment case shifting toward cloud infrastructure — hyperscalers like Microsoft and Google, which will monetize AI through computing capacity, inference, and distribution — and away from frontier AI developers like OpenAI and Anthropic, which offer less visible return trajectories despite their dramatic revenue growth.
- The combination of fiscal profligacy, above-target inflation, and an AI corporate borrowing surge competing with Treasuries for capital creates a structural environment where long-end yields may stay elevated well beyond what traditional Fed rate cycle models would predict — a regime change that investors positioned for eventual rate normalization may be underpricing.
What Happened?
Nohshad Shah, Citadel Securities’s head of EMEA fixed-income sales, published a client note Monday arguing that spiking bond yields reflect a deeper problem: a market that has lost confidence that either the Fed or fiscal authorities will make the hard choices necessary to bring long-run inflation expectations under control. Even as the Fed has kept policy rates 175 basis points below their peak, long-dated Treasury yields are at their highest levels in nearly two decades — a structural disconnect that Shah attributes to the market pricing in policy makers’ revealed preference for avoiding pain. The 30-year yield topped 5.28% Monday — a 19-year high — while Shah noted that more than 55% of core goods prices are still rising, meaning the recent softening in headline data is not yet a clean inflation victory.
Why It Matters?
Citadel Securities operates at the center of global fixed-income markets as one of the largest market-makers in Treasuries, giving Shah’s views on yield dynamics significant weight. His framing — that elevated yields reflect a loss of faith in policymaker discipline rather than purely mechanical supply/demand factors — is more politically charged than typical rate analysis, but also more durable: supply problems can be addressed through auction tweaks, but a credibility deficit takes years to rebuild. The addendum on AI investing is notable: Shah explicitly says the case is shifting toward cloud infrastructure over frontier model developers — a view that, coming from one of the most plugged-in fixed-income salespeople on Wall Street, will attract attention from institutional allocators deciding between buying Anthropic or OpenAI pre-IPO stakes versus holding Microsoft or Alphabet equity.
What’s Next?
The September Fed meeting is the next critical inflection point for rates. Shah calls it a “line-ball call” — close enough that the July FOMC minutes (due Wednesday) could tip market expectations materially in either direction. If the minutes reveal the July hold was a genuinely close decision rather than a consensus outcome, markets may price in more easing, relieving some long-end pressure. But the structural forces Shah identifies — policy credibility deficit, AI corporate supply, fiscal profligacy — will persist regardless of any single meeting outcome, suggesting that long-end yields will remain elevated on any medium-term horizon. For investors, the Citadel note’s AI rotation thesis (cloud over frontier models) may be the most actionable takeaway as IPO season for Anthropic and OpenAI approaches.
Source: Bloomberg
