Treasury Market Pricing Higher-for-Longer Under Warsh Resets Corporate Cost-of-Capital Assumptions

The Treasury market has already priced Warsh

The Treasury market is not waiting for Kevin Warsh to chair his first FOMC meeting on June 16. It has already priced him. The 2-year yield has run 79 basis points since end-February to 4.17%, the highest since February 2025, and now sits 54 basis points above the funds rate [1]. The second-order effect is the one every CFO will spend the next two quarters arguing about internally: the discount rate in every model built since 2023 is wrong, and the AI capex cycle is the largest single position exposed to the revision.

The market has done Warsh's first hike for him

Bloomberg Economics puts the tightening-equivalent of the recent yield rise at roughly 75 basis points [2]. CME FedWatch had the probability of at least one hike by year-end at about 70% after the June 6 payrolls print, which came in at 172,000 with upward revisions to prior months [3]. The 6-month bill at 3.80% says the first hike lands in 2026, not 2027 [4]. The 30-year crossed back above 5% on June 6 and hit 5.19% intra-period on May 19 [5].

What the curve is saying matters more than the level. Front-end and long-end are both selling off, which is not what you see when the market expects a single hawkish chair to overshoot and then capitulate. It is what you see when the market believes the reaction function has changed and the neutral rate sits higher than Powell's 3.5% guidance implied [6]. Jack McIntyre at Brandywine put it plainly: "Treasury yields are going to be biased higher until something breaks" [7].

For a CFO, the practical consequence is that the gap between what your treasury team is paying today and what your strategy team is using in its hurdle rate has just widened by enough to kill marginal projects. A 100-basis-point error in the discount rate carried through a 10-year cash-flow model moves present value by roughly 7-9%. Most internal models still anchor on a 2024 cost-of-capital memo. Those memos need rewriting before the autumn budget round, not after.

The AI capex cycle is the largest position exposed to the reset

This is where the second-order effect concentrates. Technology issuers have raised $250 billion in global debt markets year-to-date, per Morgan Stanley, to fund AI infrastructure [8]. Annualised AI capex sits at $750-850 billion and is projected to approach $1 trillion next year [9]. AI-related corporate issuance now supplies about 15% of the duration that total Treasury issuance does, per the Dallas Fed's Srini Ramaswamy [10]. AI-driven swap activity added the equivalent of $50 billion of 10-year Treasury supply in Q4 2025 alone, against $540 billion of 10-year notes the Treasury itself sold over the past year [11].

The mechanics matter. Data centres mix assets with very different useful lives: chips replaced every few years, buildings and substations lasting 20 to 30 [12]. That gives hyperscalers strong incentive to term out fixed-rate debt at the long end, which is precisely the part of the curve the market is now repricing. Oracle's five-year credit default swap cost has gone from about 30 basis points to 150 over the past year as the market digests its debt load [13]. That is no longer a curiosity. It is a credit signal about a category, not a name.

The reflexivity here is what most boards have not yet absorbed. AI issuance is itself a meaningful contributor to the supply pressure raising long yields. Higher yields raise the financing cost of the next tranche. Sage Advisory's Thomas Urano described the capex spend as economically comparable to "a federal stimulus package or some kind of infrastructure spend at a federal level" [14]. Stimulus crowds out. So does this.

Brad Conger at Hirtle & Co. has identified the single condition under which the position unwinds rapidly: anything that "disrupts the AI narrative" [15]. The narrative does not need to break. It needs only to wobble enough that marginal hyperscaler capex commitments get phased rather than front-loaded. If that happens, the duration supply pressure eases, but so does the earnings story underpinning a meaningful share of S&P index gains since 2023.

The Fed is not actually restrictive

CFOs who assume the Fed has been tight for two years and is about to pivot have inherited a framing that the data no longer supports. CPI and PCE are both expected above 4% for May, double target, and inflation has run above target for over five years [16]. Payrolls have accelerated, not slowed [17]. The Fed balance sheet still exceeds $6.7 trillion, and Warsh has signalled he wants it smaller [18]. Dallas Fed President Lorie Logan said on June 5 she is "increasingly concerned that higher interest rates could be necessary later this year to fully restore price stability" [19]. Jason Thomas of the Carlyle Group was blunter: "It's long past time to abandon the endemic easing bias that's characterized policy for the past two years" [20].

Layer on the fiscal side. Federal debt is growing at over $2 trillion a year, which puts a floor under term premium regardless of where the funds rate settles [21]. CFOs who have not separated "Fed policy rate" from "long-end corporate borrowing cost" in their planning assumptions have a more painful conversation ahead than they realise. The two have decoupled, and balance-sheet runoff under Warsh would widen the gap further.

The counter-case: the market has overshot the man

The strongest counter-argument is not that Warsh is dovish. It is that the Treasury market has historically over-extrapolated from incoming Fed chairs, and that the Greenspan analogy Warsh and the White House favour is weaker than it looks. Jason Thomas himself, despite his hawkish framing, notes that real rates under Greenspan were materially higher than today, meaning Warsh has less policy runway to let the economy run hot than the template implies [22]. Governor Bowman has warned that the current price spike is energy-driven and may fade [23]. Oil is up roughly 60% year-to-date on the Iran conflict; a Hormuz reopening could pull headline inflation back below 3% within two prints [24].

Warsh also does not command a clean majority. Waller has publicly challenged his framework on inflation expectations, Musalem on AI as a disinflationary force, Logan on trimmed-mean reliability (the Dallas trimmed-mean reading was 2.3% against 3.8% headline), and Barr on balance-sheet reduction [25]. A chair who loses two or three votes at the June 16-17 meeting on either rate guidance or runoff pace would prompt an immediate rally.

The case is real but does not dismantle the thesis. Even in the dovish scenario, the floor on cost of capital is higher than the 2023-24 framework assumed, because the fiscal trajectory and the AI duration supply are independent of who chairs the Fed. The cheap-capital window does not reopen. It at most stops closing further.

What to watch

1. The June 16-17 FOMC dot plot dispersion. If the longer-run dot moves up by 25 basis points or more, or if the 2026 dot shows two or more hikes with fewer than three dissents, the market signal is confirmed at the institutional level. If three or more dots dissent against Warsh's preferred path, the counter-case strengthens materially.

2. Hyperscaler debt issuance calendar through end-Q3. Watch whether Meta, Oracle, Alphabet and Microsoft accelerate long-dated issuance into July-August to lock in current yields, or pause. Acceleration confirms hyperscaler CFOs share the higher-for-longer view and tightens the reflexive loop. A pause signals either capex phasing or a bet on rates falling, either of which has equity-market implications.

3. The Oracle five-year credit default swap level and the spread between investment-grade tech and broader IG. Oracle at 150 basis points is the canary. If it moves through 200 without a name-specific catalyst, the market is repricing AI infrastructure credit as a category, which would force ratings agencies to act and tighten the funding window for second-tier players first.