Deep Dive · June 9, 2026
Higher for Longer, from Both Ends
An energy shock and firm US jobs data have flipped the next likely Fed move from a cut to a hike, but the more durable pressure sits at the long end — debt at full employment, the retreat of the price-insensitive central-bank buyer, and the capital intensity of the AI buildout. Why higher-for-longer sorts equities rather than hitting the market as a block.
Markets entered the year expecting the inflation problem to keep fading into the background. The debate was not whether the Fed would cut, but when and how far. That assumption no longer holds.
A renewed energy shock and three firm US jobs reports have moved markets from pricing cuts to pricing the risk of another hike. June is still expected to be a hold, but the next move in the funds rate now looks more likely to be up than down. That turns the Fed meeting into a test of how hawkish the new chair is willing to sound. The oil shock has also split the major central banks. The ECB and the Bank of Japan are likely to raise this month, while net energy exporters are still holding back.
The more important move is at the long end. Yields have risen together across economies that agree on little else, held up by forces that outlast the war: public debt still growing at full employment, the retreat of the price-insensitive central-bank buyer, and the capital intensity of the AI buildout. The front end is being pushed up by the energy shock. The long end is being pushed up by structure. The second force matters more because it lasts longer.
Brent sits near $93, roughly a third above its pre-war level, even as flows through the Strait of Hormuz remain far below normal. Demand destruction across Asia and a record release of strategic reserves have held the price down, but both buffers are draining toward operational limits that analysts place between June and September.
Equities have been more resilient than bonds. The S&P remains near its record, but higher-for-longer does not hit the market as a block. It sorts companies by balance-sheet quality, duration, refinancing need, and pricing power. That dispersion is the subject of the note's final section.
The June meetings: two hikes, four holds
Six major central banks meet inside ten days this June, all of them reacting to the same oil shock, and they will move in opposite directions. A supply shock raises inflation and weighs on growth at once, so it does not in itself dictate the response. What decides the response is whether an economy imports or exports its energy and what its currency is doing. On that split, June points to two hikes and four holds, and the holds now anchor a more hawkish stance.
Exhibit 1. The June meetings
| Bank | Policy rate now | Meeting | Likely outcome | Driver |
|---|---|---|---|---|
| Bank of Canada | 2.25% | 10 Jun | Hold | Net oil exporter; higher oil lifts income |
| ECB (deposit) | 2.00% | 11 Jun | Hike to 2.25% | Importer; May core rose to ~2.5% |
| Bank of Japan | 0.75% | 15–16 Jun | Hike to 1.00% | Importer; weak yen amplifies the bill |
| Federal Reserve (upper) | 3.75% | 16–17 Jun | Hold, hawkish | Sticky core; strong labour market |
| Bank of England | 3.75% | 18 Jun | Hold | Exposed importer; growth too weak to hike |
| Swiss National Bank | 0.00% | June | Hold | Safe-haven franc imports disinflation |
Oil: a contained price resting on two depleting buffers
The split across the June meetings rests on one market. Every bank reacting this month is reading the same oil shock, and the shock is stranger than the headline price suggests. Brent trades around $93, roughly a third above its pre-conflict level, even though the IEA puts cumulative supply losses near 13 million barrels per day over a Hormuz closure approaching 100 days, which it has described as the largest supply disruption on record. Throughput ran near 5% of the pre-war average across April on Kpler data, the US says clearing the mines it believes Iran laid will take six months, and ADNOC's chief executive has said full flows may not return before 2027. Two buffers explain the gap between a contained price and an extreme physical shock, and both are wasting assets.
The first is demand destruction, concentrated in Asia. The IEA has swung its 2026 global demand outlook from growth of about 730,000 barrels per day before the war to an outright contraction, with the sharpest cuts in the Middle East and Asia Pacific. India, the third-largest importer, has had its 2026 demand growth forecast cut by roughly 40% to 78,000 barrels per day, the weakest since the pandemic, per Kpler. The response is partly policy-driven: the Philippines moved to a four-day work week, Thailand imposed cooling and dress-code rules, Prime Minister Modi urged Indians to limit travel and work from home, and Chinese seaborne crude imports fell to a near ten-year low in May. This is genuine demand destruction, and it is the main reason Brent never reached the $150–$200 range some analysts modeled at the closure.
The second buffer is an unprecedented drawdown of inventories. The IEA is coordinating a record release of about 400 million barrels, of which the US is contributing 172 million. US crude stocks including the Strategic Petroleum Reserve fell to 791 million barrels in the week to 29 May, the lowest since February 2024 and an eighth consecutive weekly draw per EIA data, with the SPR itself near its lowest in four decades. The IMF projects global inventories will fall to about 7.5 billion barrels by July, a five-year low.
The inflection point is where those buffers run out. Demand can be cut only so far, and inventories have an operational floor below which pipelines, terminals and storage cannot function. JPMorgan analysts warn OECD inventories could reach operational stress levels as early as this month and operational minimums by September if the strait stays shut. Executives at Exxon and Chevron told an industry conference in late May that the market's shock absorbers are nearly depleted and that physical prices face upward pressure through June and July. The risk, in the words of one Gulf trader, is a second price shock driven by the exhaustion of buffers.
The two forces pull in opposite directions, demand destruction capping the upside and buffer exhaustion raising the risk of a renewed spike into peak summer. The point for policy is that a contained spot price understates the physical tightness, so the inflation impulse the banks are reacting to is masked, not resolved. How long it lasts depends on the strait: reopen on the mine-clearance timeline and the impulse drains, stay shut into September and the buffers give way to a second shock.
The Fed has lost its room to cut
Core PCE reached 3.3% in April and headline 3.8%, the highest since 2023, and core PCE is now running above core CPI, which leaves the gauge the Fed actually targets as the hotter of the two. The Iran war and tariff pass-through have both lifted underlying prices, so this is a domestic inflation problem on top of the energy one. The US has flipped from the economy most likely to cut to the one whose next move may be up, and the easing case has come apart since early May.
The labour market has taken away the cover to ease. May payrolls rose 172,000 against a consensus near 80,000, the third straight beat, with the prior two months revised up a combined 93,000 and unemployment steady at 4.3%. Rate-cut odds collapsed on the print, and the futures market now prices no cuts in 2026.
The chair has changed too. Kevin Warsh took office on 22 May, confirmed 54-45 in one of the narrowest votes for the role. The narrow margin shows that Congress is uncomfortable with a Fed chair so close to the White House. The administration has been open about wanting cuts, and Warsh's own hawkish record and the current data pull the other way. In our view the meeting on 17 June is still a hold, with the updated dot plot the signal that matters and the risk around it skewed to a hike. The committee is split, and the case for holding rather than hiking rests on r-star, a bet that the hot data reflects a higher neutral rate and a productivity wave yet to land.
The long end is rising on forces that outlast the war
Short rates have repriced on the oil shock. Long rates have risen together across the US, the euro area, the UK and Japan, which points to a higher resting level several economies share. The US thirty-year sits near 5.0% and the ten-year near 4.55%, the UK thirty-year reached 5.4% in late 2025 at its highest since 1998, and the Japanese thirty-year has set records near 3.5% after decades close to zero. Long yields stayed elevated as the war premium dipped and returned, so energy is not what holds the anchor up. In our view there are three primary drivers that keep longer term yields elevated.
Fiscal
The largest force is fiscal. Every major issuer is running a structural deficit at full employment, and bond markets are repricing sovereign risk across all of them.
CBO projects a US federal deficit of 5.8% of GDP in fiscal 2026, rising to 6.7% by 2036, against a fifty-year average of 3.8%. The 2025 reconciliation act added roughly $4.7tn to projected deficits over the decade, partly offset by about $3tn from higher tariff revenue. Debt held by the public rises from 101% of GDP to 120% by 2036, above any prior point in US history.
France runs a 2026 deficit near 5.7%, with debt heading toward 125% of GDP by 2030 and a sequence of governments unable to pass a consolidating budget, which has held the OAT-Bund spread around 69 bp and drawn rating downgrades. Germany has loosened its debt brake for defense and infrastructure, budgeted about €108bn for the military in 2026, and is pushing its general government deficit from 2.4% to a forecast 3.5% of GDP, so the bloc's benchmark issuer is now adding Bund supply. The UK is borrowing near £138bn in 2025/26 with consolidation deferred to the end of the decade, and its thirty-year gilt yields sit at multi-decade highs.
Japan's FY2026 budget exceeds ¥122tn against the highest debt ratio in the developed world, well above 200% of GDP. The ten-year JGB has climbed to around 2.4%, a 27-year high, and thirty-year to records near 3.5%. The average coupon on outstanding Japanese debt is roughly 0.8%, so each maturing bond reprices at a far higher yield as it rolls, lifting the interest bill against a budget already stretched.
Each year of large deficits adds a stock of bonds the market must absorb, absorbing more requires a higher yield, and the higher yield raises the government's interest bill, which widens the deficit it funds. US net interest alone doubles from $1.0tn to $2.1tn over the decade. None of these blocs has a consolidation plan in prospect, so the supply keeps coming.
Exhibit 2. Fiscal positioning, 2025
The buyer base
The second force is who buys that debt. Through the quantitative-easing era, central banks bought government bonds in every major bloc for policy reasons rather than for return, taking almost any quantity at almost any yield. That bid put a floor under demand and a ceiling on long rates.
It has now reversed across all four central banks. The Fed ran quantitative tightening from mid-2022 until late 2025, shrinking its holdings by about $2.4tn, and now matches reserve growth. The ECB is letting roughly €500bn in APP and PEPP holdings run off in 2026. The Bank of England has cut its gilt portfolio by around £300bn through active sales, slowing the pace from £100bn to £70bn a year under market pressure. The Bank of Japan fixed the ten-year near zero under yield-curve control and held more than half of all JGBs; it has exited that regime, raised rates to 0.75%, and begun shrinking holdings that had reached ¥544tn. The foreign official sector has diversified reserves away from Treasuries over the same period.
The buyers that absorbed the slack are price-sensitive. Households, money market funds, and leveraged funds running the cash-futures basis trade took up much of it in the US, while UK defined-benefit pension funds, who have long been the natural home for gilts, are in secular decline. These buyers demand compensation for duration, which lifts the term premium and the yield, and they reprice when the growth or inflation outlook shifts, which raises volatility. The basis-trade positions carry their own risk, since they are repo-funded and highly leveraged and can be forced to unwind under stress, as the Treasury dysfunction of March 2020 and the UK gilt crisis of 2022 both showed. Each major long-end market is now more prone to disorderly moves on fiscal news, and the central banks that long absorbed those moves hold less of the market that would need them.
The AI buildout
The third force is the AI buildout, and it sits at the center of our mandate. It works on the bond market, on the neutral rate, and on inflation at once, and the spending behind it does not slow when rates rise. It is also a US-led force, which shapes how it reaches the other blocs.
The first channel is bond supply. The five largest hyperscalers issued about $121bn in US corporate bonds in 2025 per BofA, roughly four times their $28bn annual average over 2020 to 2024. By October, AI-tied debt had reached about $1.2tn, near 14% of the high-grade market on the JP Morgan index and the largest sector ahead of banks. Barclays expects net investment-grade issuance to rise about 30% to roughly $945bn in 2026 and traces most of that increase to hyperscaler capex, and BofA has lifted its 2026 forecast for hyperscaler borrowing to $175bn. That supply competes with government debt for the same finite pool of duration buyers. Apollo's Torsten Sløk and Dallas Fed research describe the same mechanism: hyperscaler issuance pulls the marginal buyer away from Treasuries and lifts the yield the government must offer to clear.
The second channel is the neutral rate. AI capital spending, which Goldman estimates at $7.6tn over five years, is large enough that investment demand outruns available savings and raises r-star, the real rate that holds the economy in balance. The Institute of International Finance argues a successful AI cycle should lift r-star and that markets should not expect a return to the low-rate world of the 2010s. Inside the Fed, Governor Barr has said the AI boom is unlikely to justify a cut, and Chicago's Goolsbee has warned that demand pulled forward on expected productivity could overheat the economy. A higher neutral rate makes a given policy setting less restrictive than it looks.
The third channel is electricity, already in the inflation data. PJM capacity prices rose from $28.92 per megawatt-day for the 2024/25 delivery year to $269.92 for 2025/26, an 833% increase and the largest in the market's history, then cleared at the regulatory cap of $329.17 and $333.44 in the two following auctions, the last short of the reliability requirement for the first time. PJM's independent market monitor attributed 63% of the 2025/26 increase to data centers, about $9.3bn recovered from ratepayers in one year. Retail supply rates across the thirteen-state region are up between 5 and 44% since June 2025. The buildout also bids for scarce transformers, switchgear and skilled electrical labour, raising their clearing prices.
Productivity is the counterargument. A successful buildout raises productivity and lowers costs over the medium term, which is disinflationary, and Chair Warsh has used that case to argue for keeping rates low. The capex and the power demand arrive now, while the productivity payoff is lagged and uncertain, so the near-term effect is upward pressure on rates and prices. The disagreement is live on the committee Warsh chairs, with Goolsbee, Barr and Hammack on the other side.
The euro area and Japan get this force mostly from outside. Hyperscaler issuance is dollar-denominated and the power-price pressure sits in US grids, so the other blocs import the rate impulse through correlated global yields and a strong dollar while their own neutral rates get no lift from a domestic investment boom. Europe and Japan are building out far less and later, with no hyperscaler of comparable scale and slower data-center and grid investment, so the domestic channel that lifts US r-star barely operates there. Their long-end pressure comes from the fiscal and buyer-base forces, with the AI channel arriving on top. In the US the buildout drives both yields and inflation, and it persists through higher rates because hyperscaler spending is set by competition. A firm that pauses its buildout cedes position, so the largest marginal borrower is the one least deterred by the price of borrowing.
Each of these forces runs for years, and each eventually runs down. The balance sheet reaches its target and stops, the buildout turns from spending into the cash it was meant to throw off, and the deficits ease if consolidation ever arrives. When the forces mature the regime turns, and a cutting cycle becomes plausible, past the horizon this note forecasts. How it turns is the question. A regime that ends because the forces ran their course lets rates fall in order. A regime that ends because something breaks in the leveraged base that now holds the long end brings the cuts and the recession together.
Higher for longer reaches equities through the multiple
Higher for longer does not stop at the bond market. A higher resting level for long yields raises the discount rate on future cash flows, and that weighs most on the equities whose value sits furthest out. The AI complex is the longest-duration cohort in the market. The capital is spent now, with hyperscaler capex near $750bn for 2026 on Goldman's consensus and up more than 80%, while the return on it accrues over years. The revenue is real and compounding fast, the largest model labs moving to $25bn to $30bn annualised run-rates inside fifteen months and the infrastructure sellers already booking the buildout as earnings. What is unsettled is the return on the capital, whether spending at this pace clears its cost, and that future stream is what a higher discount rate marks down.
The consensus case for 2026 rests on a Fed that eases. Goldman's 8000 target holds the market multiple flat near 21 times forward earnings, and that flat multiple assumes Treasury yields drift lower. Morgan Stanley named a more hawkish Fed as the main risk to the bull market and judged it unlikely. The next move in the funds rate is more likely up than down, and the structural forces hold the long end where it is regardless. The yield decline the multiple was priced for is not coming, which means the multiple the consensus holds flat is more likely to compress, and a target built on it is too high.
What protects equities is earnings growth fast enough to outrun the compression. FactSet puts 2026 consensus earnings growth at 21%, with the largest names near 23%, against forward multiples in the low-to-mid 20s that already price much of that. This is what separates the present from 1999. At the 2000 peak the ten largest stocks traded near 40 times trailing earnings on far thinner profits, so rising rates hit the multiple and the missing earnings together. Today's leaders carry high multiples and real, growing earnings, so a higher discount rate compresses them. Capacity financed against multi-year contracted revenue grows into the higher cost of capital. Capacity priced on a payoff that has to land before the financing reprices is exposed on the multiple and the cash flow at once, and credit markets are already widening spreads on the issuers whose returns depended on cheap capital.
Exhibit 3. This isn't 1999
What would change the view
The note makes two bets, on the front end and the long end, and they can be wrong separately. Oil could be wrong without the structural call being wrong, and the reverse.
Start with oil. It can do three things. It can reopen, if Hormuz clears faster than the mine-clearance estimates imply, in which case flows return, the inflation impulse fades, and the hawkish holds turn back toward cuts. It can spike again, if the inventory and demand buffers run out before the strait reopens, which sends a second shock into peak summer and pushes the Fed from hold toward hike. Both of those leave the long end alone. The third is the war dragging on, and that one matters most, because an energy shock that lasts years is no longer a shock. It becomes part of the structural picture, and the clean line between the two halves of this note disappears.
The long end is harder to dislodge, but several things would do it. Governments could consolidate, cutting deficits credibly and for long enough to matter, or grow fast enough to stabilise debt without cutting. The big price-insensitive buyer could come back, with central banks buying bonds again instead of shrinking their holdings. The AI bill could fall several ways. Leaner model architectures could cut the compute it takes. The productivity payoff could arrive fast enough to bring the neutral rate and inflation down together, the case where the buildout works and rates ease because it did. Or the opposite: the buildout could fail to earn its return, capex and the issuance funding it could retrench, and rates would fall because the boom broke rather than because it paid off, taking the AI leaders down with them. If these forces burn out or reverse sooner than the next several years, the call is wrong on timing even if it is right on direction. It is wrong on substance only if long yields across these economies fall back toward their pre-pandemic levels while deficits and AI borrowing stay high, which would mean neither force was holding the level up after all.
One outcome proves the rate call right and still ruins the trade built on it. If the leveraged money now holding the long end is forced to unwind in a hurry, the way it was in the Treasury market in 2020 and in gilts in 2022, yields drop fast, but they drop into a recession with earnings falling, and that buries the long-duration names rather than rewarding the ones that survived.
The equity call can fail three ways, none of which touches the rates view. It fails if the AI names move as a pack, rising or falling together with the strong and weak balance sheets indistinguishable, because then there is nothing to select. It fails if yields stay high and the multiples do not compress, because then the discount rate was never the channel. And it fails if the leaders stop growing fast enough to outrun the compression and sink to the level of the also-rans, because then picking among them earns nothing.