The Long End Breaks as the 30-Year Hits Its Highest Since 2007. The Iran Deadline Expires and Brent Reaches $91.
Tuesday, August 18, 2026  ·  Before Bell
Morning Note  ·  Equity Research

The Long End Breaks as the 30-Year Hits Its Highest Since 2007. The Iran Deadline Expires and Brent Reaches $91.

August 18, 2026
Fund X
Steven Yuan
30Y Treasury 5.327%, highest since June 2007
Japan 10Y JGB 2.95%, highest since September 1996
Eurozone 10Y above 3.2%, a fifteen-year high
Brent $91, deadline expired, no talks scheduled
Diesel crack record $102/bbl, SPR under 300M bbl
Nasdaq 100 futures -1.27% vs Dow futures -0.07%
Nvidia Ohio deal $105B guarantee, 8GW, exclusive supply
Baidu GPU cloud +283%, ads -19%, stock -7%
Home Depot $47.86B vs $47.27B, comps +1.7%, reaffirms
BofA cash level 3.5%, sixth lowest since 1998

Tuesday, August 18, 2026, written at 8:20am Eastern, ninety minutes before the bell, with Home Depot already out and the July FOMC minutes twenty-nine hours away. Two prices did the damage overnight and neither of them is an equity. The 30-year Treasury printed 5.327%, the highest since June 2007. Brent went to $91 after the 60-day window for a US-Iran settlement expired with nothing to replace it. Everything in the tape this morning is downstream of those two lines.

Top Call: The Discount Rate Is Now Global and the Growth Complex Pays First

Long-dated government bonds sold off everywhere at once. The US 30-year reached 5.327%, up nearly forty basis points since the end of June. Japan’s 10-year hit 2.95%, a level last seen in September 1996, with the 30-year JGB at 4.12%. UK 30-year gilts sit near 5.82%. Eurozone 10-year yields cleared 3.2% for the first time in fifteen years. The US 2-year, meanwhile, moved three basis points to 4.21%. That combination is the whole story: this is not the market repricing central bank policy, it is the market repricing the cost of holding duration in a world of heavy issuance, sticky inflation and shrinking foreign sponsorship. Treasury holdings by the UK, China and Japan all fell in June.

Our take: a term premium shock is more dangerous for the AI complex than a policy shock, and this morning’s futures composition proves the market already knows it. Nasdaq 100 futures are down 1.27% while Dow futures are down 0.07% and the Russell is down 0.35%. When the front end moves, cyclical earnings get repriced. When the back end moves, terminal value gets repriced, and the AI infrastructure trade is almost entirely terminal value financed by long-dated debt. Friday’s note argued that the long end had become the binding input for these equities. Today is the second consecutive session where that link priced through cleanly, and with oil adding a fresh inflation impulse on top, the two shocks are pointing the same way rather than offsetting. We are not calling the top of the buildout. We are saying the buildout now trades as a rates product, and anyone still modelling it off demand headlines is watching the wrong screen.

The Tape

  This Morning Prior Close (Aug 17)
S&P 500 fut / cash 7,735.00 (-0.43%) 7,745 (-0.52%)
Nasdaq 100 fut 29,712.50 (-1.27%) Comp 26,645 (-0.32%)
Dow fut / cash 53,504 (-0.07%) 53,460 (-0.51%)
Brent / WTI $91 / $84.34 (+0.72%) Brent settled above $90
30Y / 10Y / 2Y 5.327% / 4.74% / 4.21% 2s30s steepest since 2021
Gold / VIX $4,450 (-0.53%) / 15.75 (+3.68%) volatility still cheap

The internals matter more than the index level. A 120 basis point spread between Nasdaq and Dow futures on a morning with no company-specific tech catastrophe is a duration trade, not a sentiment trade. The VIX at 15.75 is up almost 4% and remains close to its 2026 low of 14.39, which tells you the market is repricing value rather than buying protection. That is a coherent response to a rates move and a poor one if the Iran situation deteriorates further.

Iran, Hormuz and the Expired Deadline

The 60-day negotiating window closed Monday with no accord and no schedule for talks. President Trump said he intends to inflict further economic pain on Iran, threatened to bomb Oman if it interferes with US plans for the Strait of Hormuz, and repeated that he expects to declare the strait a US territory once the war ends. Iran’s foreign ministry said it had reached its own understanding with Oman on shipping routes. Energy Secretary Chris Wright described the American approach as a long game, which is the closest thing to official guidance that de-escalation is not imminent. The war is in its twenty-fourth week.

The physical market has moved past the headlines. War risk premiums for Hormuz transits are running at 7.5% to 10% of hull value against roughly 0.25% before the conflict, which puts $7.5 million to $10 million of insurance cost on a $100 million tanker for a single passage. The Strategic Petroleum Reserve is below 300 million barrels, the lowest since 1982, so the traditional American release valve is empty. Diesel cracks hit a record near $102 a barrel. That last number is the one to carry around, because it converts directly into freight, agriculture and industrial input costs with a lag of weeks rather than quarters.

Our take: the crude price is no longer the right thing to watch. Refined product spreads are where the shortage actually lives, and a record diesel crack against an SPR at four-decade lows is a supply structure with no shock absorber left. This is also the mechanism by which the oil story becomes the bond story, since a durable diesel-led inflation impulse is exactly what forces term premium higher regardless of what the Fed does in September.

Nvidia and Ohio: What the Guarantee Actually Bought

Yesterday we read the reduction of Nvidia’s Ohio backstop from $250 billion to under $120 billion as a straightforward de-risking of the supplier’s balance sheet. The fuller version arrived today and it is more interesting. The guarantee settled at up to $105 billion against an 8-gigawatt campus, and the consideration flowing back to Nvidia is exclusive chip supply to the site. That is not a retreat. That is a supplier converting contingent credit into a locked order book, and it comes a week after Nvidia arranged $500 billion of chip financing capacity with Wall Street firms.

Our take: we were half right and the half we missed matters. The de-risking is real, since $105 billion of contingent obligation is materially better than $250 billion. But the framing that vendors are stepping back from financing customers is wrong. They are repricing what the financing buys. Nvidia now holds $30.2 billion of marketable equity securities against $12.9 billion a year ago, and it is using balance sheet as a competitive weapon rather than as a subsidy. The strategic read is that the moat is migrating from silicon to capital, which is durable while credit is available and fragile the moment the long end makes that capital expensive. Note the date on that risk: the 30-year is at a nineteen-year high this morning.

Baidu: AI Growth and AI Disruption in One Income Statement

Baidu Q2 2026 Consensus Actual Read
Revenue RMB 31.95B RMB 31.33B -4% y/y, missed
EPS RMB 9.84 RMB 7.22 large miss
Core advertising n/a RMB 13.1B -19% y/y, fourth straight decline
AI cloud infrastructure n/a RMB 7.3B +50% y/y
GPU cloud n/a n/a +283% y/y, from +184%
Total AI revenue n/a RMB 12.5B +25% y/y

Shares fell roughly 7% pre-market. The headline is a miss and the composition is the most instructive print of the morning. GPU cloud revenue nearly quadrupled and accelerated from the prior quarter’s 184%, which is direct evidence that Chinese AI compute demand is being met and monetised despite export restrictions. In the same three months the advertising business lost 19% of its revenue, the fourth consecutive quarter of declines in the 15% to 21% range, because users are migrating from search to chatbots and e-commerce platforms are redirecting budget from traffic acquisition to subsidies.

Our take: this is the clearest single-company demonstration of a pattern the sector has been arguing about abstractly. The AI revenue is real and compounding fast. It is also arriving too slowly, at too low a base, to offset the destruction of the legacy business it is replacing, and the destruction is happening inside the same company that owns the growth. RMB 7.3 billion of AI cloud infrastructure does not repair a RMB 13.1 billion advertising line falling 19%. For anyone holding incumbent internet platforms on the theory that AI investment protects the franchise, Baidu is the counter-example, and the market treated it correctly.

Home Depot: The Consumer Is Not the Problem This Morning

Home Depot FQ2 Consensus Actual Read
Revenue $47.27B $47.86B +5.7% y/y, beat
Comparable sales ~+1.0% +1.7% US comps +1.3%
Adjusted EPS n/a $4.92 $4.68 a year ago
FY26 guidance n/a reaffirmed comps flat to +2%

Shares rose about 1.7% pre-market. Four sessions ago July retail sales fell 0.6% for the worst month since May 2025 and this note said the consumer was cracking. Home Depot’s comp acceleration is a partial rebuttal, and an honest one: big-ticket housing-linked demand held up in a quarter when aggregate retail spending did not.

Our take: we would not upgrade the consumer on this. Home Depot’s customer skews toward homeowners with equity and toward professional contractors, which is the most insulated part of the household sector, and management reaffirmed rather than raised full-year guidance despite the beat. The genuine tension is that this company’s demand curve is priced off the same 30-year yield that broke to a nineteen-year high overnight. Mortgage rates follow the long end, housing turnover follows mortgage rates, and renovation follows turnover with a lag. A good quarter reported into a rising term premium is a good quarter with a shortening shelf life. Target and Lowe’s tomorrow and Walmart Thursday remain the real tests.

Positioning: Bank of America Finds Everyone Already In

The August Global Fund Manager Survey landed this morning and it reads as the mirror image of the bond market. Cash fell to 3.5% of assets, the sixth lowest reading since the survey began in 1998, which keeps the firm’s cash rule on a sell signal. Global equity allocation is a 56% net overweight, the highest since November 2021, and the US allocation is 27% net overweight, the highest since December 2024. A record 56% expect no landing and 43% expect an outright boom, the most since February 2022. Long global semiconductors remains the most crowded trade at 53%, though that has fallen sharply from 82% a month ago. The biggest tail risk is an AI bubble at 32%, and 38% identify hyperscaler capex as the most likely source of a systemic credit event, yet 71% expect no capex cuts this year. Some 72% expect no Fed hike before the November midterms. Gold screens as the most undervalued asset since March 2023.

Our take: the survey describes a market that has extended itself on a specific macro assumption, which is that growth stays above trend, inflation is tolerable and long rates behave. Two of those three were challenged overnight. The 53% semiconductor crowding, down from 82%, is the only genuinely constructive datapoint here, since it means the most exposed trade has already been partly unwound. The rest is a positioning profile that gives up more on a term premium shock than it gains on a benign surprise. Note also the internal contradiction: 38% naming hyperscaler capex as the likeliest credit event and 71% expecting no cuts is the same investor holding both sides of the same trade.

Asia and Europe

Japan took the worst of it. The Nikkei fell 2.54% to 67,460.73, with Kioxia, Murata, Taiyo Yuden, Sony and Fast Retailing leading the decline, as the 10-year JGB reached 2.95% and the government’s proposed consumption tax cut on food arrived without an identified funding source. That is a domestic fiscal premium stacking on top of the global one, and it is happening while the market prices an imminent Bank of Japan hike. Korea’s Kospi erased a 2% intraday gain to close 1.55% lower at 6,869.83, with the Kosdaq down 3.52%, after five straight sessions of foreign buying. Hong Kong finished roughly flat at 25,471.15 as energy names offset an AI pullback. Shanghai rose 0.19% to 3,990.30 while the Shenzhen Component fell 0.56% and the ChiNext lost 0.92%, with the CSI 300 down 0.8% at the lunch break.

Europe opened lower with Stoxx 600 futures down 0.4% and Euro Stoxx 50 futures down 0.6%, after Monday’s closes at 10,720.30 for the FTSE 100, 8,579.60 for the CAC 40 and 26,338.61 for the DAX. German ZEW expectations improved to 34.2 from 26.3, which is a genuinely better number that nobody traded because the bond move overwhelmed it. Energy shares are the one European sector working. Canada’s July CPI accelerated to 3.0% against 2.9% expected, another small confirmation that the energy pass-through is showing up in headline inflation across developed markets.

Also Worth the Time

Equinor rose 1.6% to NOK 394.20 on a $940 million US power acquisition, which is the European majors continuing to buy into the electricity demand story rather than the barrel. Braskem Idesa filed Chapter 11 to cut more than $920 million of debt. Rapid7 cut 12% of staff last week under a new chief executive, part of a 2026 tally that now stands at 322 layoff events affecting roughly 205,832 workers, with several companies explicitly attributing headcount reductions to AI deployment. Huber+Suhner fell 5.7% in Europe after missing on core profit and communications orders, which is a small but relevant datapoint for the optical connectivity chain given how much of that complex trades on AI datacentre orders. Space names remain firm after last week’s run, with Rocket Lab still carrying the $397 million Space Force Flatellite award and SpaceX digesting a debut quarter that beat on revenue at $7.81 billion and spooked investors with $18.4 billion of capex.

Key Events Today

What We Are Actually Watching

Three threads.

The first is whether the long end is signalling inflation or supply, because the trade is different in each case. If this is an inflation premium, it is oil-led and it unwinds when Hormuz does. If it is a supply and sponsorship problem, with foreign holders reducing Treasuries and every major sovereign issuing heavily into the same window, then it persists through any Middle East resolution and it keeps compressing the multiple on long-duration equity indefinitely. The evidence this morning leans toward the second reading, since the move is simultaneous across the US, Japan, the UK and Europe, and Japan’s fiscal politics have nothing to do with Iranian crude. That is the more uncomfortable interpretation.

The second is that the AI trade’s financing structure keeps getting more sophisticated while the rate environment gets less accommodating, and those two lines are converging. Nvidia has now arranged $500 billion of chip financing capacity and converted its Ohio exposure into exclusivity at $105 billion. That is a well-engineered structure. It is also a structure whose cost of capital is set by a 30-year yield that has risen forty basis points in seven weeks. The buildout does not need demand to disappear to slow down. It only needs credit to get expensive enough that the marginal gigawatt stops clearing.

The third is the gap between what investors say and how they are positioned. Thirty-eight percent of surveyed managers think hyperscaler capex is the most likely trigger of a systemic credit event, and the same survey shows cash at 3.5% and the largest equity overweight since November 2021. That is not hypocrisy, it is a crowded trade with an identified exit and no one willing to leave early. It works while the tape is orderly. The reason to watch the VIX at 15.75 rather than the S&P at 7,735 is that the first sign of that changing will be a bid for protection, and there is not one yet.

Marks above are pre-market and will move by the open. There is no first-tier US data today, which means the tape trades on oil headlines and the bond auction calendar until the minutes tomorrow afternoon.