Weekend Intelligence Report

The Weekend Brief

• Friday’s Iran-related headlines eased oil pressure, but they did not establish that shipping has recovered. A new shipping analysis estimates Hormuz traffic remains about 75% below pre-conflict levels and Gulf exports at roughly 12 million barrels a day, versus 20 million before the conflict. Those are source-reported estimates, not verified post-talk flow data.
• Long-term borrowing costs remain high despite Friday’s retreat from the week’s peak. The 10-year Treasury was quoted at 5.165% late Friday, after reaching 5.23%; it was 4.95% on September 18. Analysts disagree on how much reflects stronger real returns, Treasury and corporate supply, or concern about future demand.
• The week’s activity data and household signals point in different directions. PMIs accelerated, claims remained low and core capital-goods orders rose 1.6%. Final Michigan sentiment fell to 48.1 from 51.7 in August, while one-year inflation expectations rose to 4.6%.
• AI demand is visible, but the return on the investment is not yet established. Akamai reported a seven-year, $11.6 billion Anthropic cloud agreement; separate reports describe costly project financing, large commitments and sharply lower prices for replacement models.
• Friday’s equity gains did not settle the underlying tensions. The S&P 500 closed up 0.51% and the Nasdaq Composite up 0.5%, while the 10-year remained near multiyear highs. The supplied weekly S&P return estimates conflict, and there is no fresh Friday breadth measure to show whether participation improved.

The Big Story

Diplomacy moved oil; it did not yet restore Hormuz capacity

Friday brought a market-relevant change in tone. Reports of possible phased U.S.–Iran terms coincided with lower oil, a halt in the Treasury-yield surge and a 0.5% rise in the S&P 500, according to Trading Economics. That is a change in expectations about a possible path to de-escalation—not evidence that an agreement was signed or that tankers and cargoes are moving normally.

The distinction matters because the shipping evidence describes an impaired route even as some vessels continue to pass. What’s Going on With Shipping? reports 129 vessels transiting Hormuz over the preceding seven days, with traffic roughly 75% below pre-conflict levels. It also cites approximately 30 U.S.-reported facilitated transits a day, compared with about six AIS-derived transits a day. Those counts are not directly interchangeable: they use different methods, and some ships do not broadcast AIS. The discrepancy is itself a reason not to treat a single passage count as a complete measure of reopening.

The same analysis puts Gulf exports at roughly 12 million barrels a day, against about 20 million before the conflict. It reports 106 maritime events, including 87 vessel-damage reports, 33 injuries or fatalities and five confirmed total losses. These are figures reported by the source, not an independently verified current balance sheet of the conflict. But they describe risks that a negotiating headline alone cannot remove.

Shipping costs reinforce the gap between limited passage and normal capacity. The analysis reports Middle East Gulf–China VLCC rates of $1.26 million a day, versus roughly $50,000–$100,000 before the conflict. A quoted route rate does not represent every realized fixture, but the scale of the difference suggests that vessels, insurance and routing remain constrained. Reported pipeline and Red Sea difficulties can also lengthen voyages and absorb tanker capacity even when some ships pass Hormuz.

The diplomatic accounts describe possibilities, not settled terms. Chas Freeman says an Oman-mediated proposal could reopen Hormuz within seven days if the U.S. blockade ends, while also saying the Gulf Cooperation Council had not endorsed it. Trita Parsi expects mutual distrust to obstruct a durable agreement. Friday’s reports are consistent with negotiations being explored; they do not confirm Freeman’s conditions or resolve Parsi’s concern.

The incentives are consequential on both sides. Any arrangement described in the research links reopening to the status of the U.S. blockade and measures against Tehran. For the parties, negotiating terms that produce passage would matter more than an announcement that leaves restrictions and security risks in place. For energy importers and shipping customers, the relevant outcome is delivered cargo at lower cost—not simply a lower crude quote. Tanker operators with available capacity may benefit from unusually high rates, while carriers and cargo owners remain exposed to delays, route risk and expensive freight.

This is why the Friday oil move should be treated as relief in the diplomatic risk premium, not proof of physical normalization. The report’s early Saturday Brent indication of $104.32 is unchanged and does not establish Saturday trading; Friday’s supplied Brent summaries also conflict. The more useful tests are comparable vessel and cargo counts, Gulf export volumes, confirmed pipeline throughput, and whether realized tanker rates fall.

Who benefits if the route normalizes: energy importers, shippers and customers paying high freight or insurance costs would have a path to relief; lower delivered fuel costs could also ease pressure on refined products and transport. Who remains exposed: users dependent on Gulf flows, tanker operators whose routes or capacity remain disrupted, and refiners and fuel consumers if crude relief fails to reach product availability.

The unresolved question is whether the diplomatic process can change the physical economics of the route. Until comparable traffic, export and freight data improve, Friday’s market response is evidence that the prospect of a deal matters—not that the supply shock is over.

What the Market May Be Missing

Strong production signals and weak household confidence can coexist

The week’s data do not support a simple story of either broad economic strength or an already completed collapse. September manufacturing and services PMIs rose to 57.0 and 58.7. Initial claims were 197,000, below the 202,000 consensus, and core capital-goods orders excluding aircraft rose 1.6% in August. Those readings challenge claims that employment and business investment had already fallen off a cliff.

But they do not establish that household finances are healthy. Final Michigan sentiment fell to 48.1 from 51.7 in August. Respondents’ one-year inflation expectations rose to 4.6% from 4.0%, and five-year expectations reached 3.4%, up from 3.3% in each of the prior three months. Views of current and year-ahead personal finances also worsened, according to Trading Economics.

These indicators measure different things and periods: surveys of activity, realized orders, jobless claims and reported household attitudes are not interchangeable. The policy implication is still important. Strong activity and higher inflation expectations can support a restrictive Fed stance even as confidence deteriorates. Conversely, sentiment is not realized spending, and it cannot by itself establish that demand has already weakened. The evidence leaves open how long current activity can coexist with weaker confidence and elevated financing costs.

Deep Dives

Treasury yields are high; the cause is not settled

The 10-year Treasury was quoted at 5.165% late Friday, after reaching a 5.23% high. That is about 21 basis points above the 4.95% level reported on September 18, even after Friday’s retreat. The two-year was quoted at 4.86% a few minutes later; the resulting spread is only an indicative calculation from separately timestamped records, not a published closing curve measure.

There are credible but competing explanations for the move. Art Laffer emphasizes stronger expected real returns. David Busch argues that Treasury supply competes with corporate AI borrowing and other uses of capital, including private-credit allocations by insurers. Henry Peabody also points to real rates and capital competition, while warning that persistent inflation could make long-duration bonds a poor hedge. Earlier Eurodollar University analysis read curve flattening as a possible warning of later demand weakness.

The week’s data give both sides evidence, without resolving the decomposition. Stronger PMIs and capital-goods orders are consistent with resilient activity; Michigan’s higher inflation expectations may also matter to rate pricing. But neither release tells us how much of the 10-year yield reflects expected Fed rates, real yields, inflation compensation or term premium. Nor do the cited auction results show a uniform failure of Treasury demand: the two-year auction was reported as robust and the five-year as weak.

The implications reach beyond bond portfolios. The 10-year is a financing reference for mortgages and long-lived corporate projects; another move toward 5.23% would make the recent pressure more acute. Yet the evidence does not establish that high yields have already caused a broad investment stop. The next test is whether yields remain elevated as forthcoming employment and inflation data arrive—and whether credit and project financing show broader deterioration.

AI demand is real evidence; profitability remains the harder test

Akamai’s reported seven-year, $11.6 billion Anthropic cloud agreement is meaningful evidence of customer demand. Finviz says the deal includes warrants for up to 5% equity and could expand to about $20 billion. Microsoft rose approximately 3.7% Friday after Copilot announcements; Meta also ended the week higher, though it fell on Friday. These are company-specific developments, not proof that AI investment across the sector earns adequate returns.

The financing side raises a different question. Eurodollar University reports SoftBank marketed debt yields near 8.9%–10% and roughly $18 billion of Oracle-linked Project Jupiter loans quoted at 89–91 cents on the dollar. Patrick Boyle describes a broader web of commitments around planned data-center construction, including reported SB Energy funding needs, a long-term Ohio lease to OpenAI and a reported Nvidia guarantee of up to $105 billion. Those terms are source-reported and do not establish that projects will fail; they do make execution, counterparties and cash generation central to the investment case.

The commercial economics could move in either direction. Boyle reports replacement Anthropic and OpenAI models priced 40% and 50% lower, respectively, and attributes to Epoch AI an estimate that the cost of achieving a given performance level has fallen roughly 13-fold annually since 2023. Falling prices can make AI cheaper for customers and broaden usage. They can also pressure provider margins if revenue per unit of capability falls faster than customer volumes and efficiency improve. The dossier does not establish which effect will dominate.

Valuation claims require similar caution. Boyle reports a proposed Anthropic valuation of $2 trillion against unofficial annualized revenue near $65 billion—about 31 times that reported revenue. Neither a filing nor audited revenue is supplied. The multiple is therefore a useful marker of the expectations embedded in the reported proposal, not a verified valuation or a conclusion about fair value.

Analysts also disagree about the pace of investment. Danielle DiMartino Booth cites a 34.6% decline in Q2 AI-investment growth, but the series’ definition and year are unclear. Henry Peabody argues that large borrowers can still obtain funding. The 1.6% rise in core capital-goods orders and Akamai’s contract challenge a blanket claim that business investment or customer demand has stopped; they do not verify AI-specific returns or settle Booth’s measurement.

The beneficiaries are not necessarily the same as the companies taking the greatest project risk. Infrastructure providers can win contracts while project owners and lenders carry construction, utilization and financing exposure. Users may benefit from lower model prices, while model providers may have to prove that falling costs translate into durable margins. Contract size is evidence of demand; it is not yet evidence of cash returns.

Private-credit stress is visible, but the measures do not establish a systemwide break

Apollo Debt Solutions reportedly received redemption requests for 14.7% of shares and repurchased 5%, its contractual cap. The source estimates about $200 million of inflows and $700 million of repurchases, or roughly $500 million net outflow. This points to a liquidity mismatch between investors seeking withdrawals and a fund holding less-liquid assets. A redemption cap is not proof of insolvency.

Reported default measures also describe different populations. Fitch’s 6.3% measure covers roughly 1,300 U.S. borrowers; Houlihan Lokey’s reported rate is below 1%; PIMCO’s 19% “shadow default” estimate concerns a different, retail-oriented BDC population. They should not be averaged or treated as rival readings of the same borrower set.

The mechanism to watch is whether higher rates and weaker refinancing capacity translate into missed cash interest, more nonaccruals or lower recoveries. Peabody argues that some large borrowers still have access to funding even as highly leveraged borrowers face declining interest coverage. Busch emphasizes that floating-rate loans reduce duration exposure but leave investors with credit, valuation and redemption risk. Neither the supplied Friday research nor the differing default estimates demonstrate a generalized funding seizure.

For investors, the key distinction is between illiquidity and ultimate loss. Requests above a repurchase cap show pressure on fund liquidity; loan-level marks, cash collections, new lending and recoveries would be needed to assess whether that pressure is spreading into borrower solvency. No fresh, comprehensive Friday measure of public-credit spreads or aggregate nonaccruals is supplied.

What Changed This Week

• The rate narrative moved from a possible retreat to renewed pressure. The 10-year stood at 4.95% on September 18 and was quoted at 5.165% late Friday, after touching 5.23%. Friday’s easing did not restore the earlier yield level, and the cause of the rise remains contested.
• Activity data strengthened the case against an already completed broad contraction. September PMIs accelerated, claims remained low and core capital-goods orders rose 1.6%. These readings do not resolve the weakening household-confidence signal or predict how higher financing costs will feed through.
• The Iran story shifted from negotiation possibility to market-moving reports—but not to confirmed implementation. Friday’s oil and equity response came without supplied evidence of signed terms or improved Hormuz cargo flows.
• The AI debate became more specific. A large cloud agreement and product announcements provide evidence of demand, while reported financing terms, project commitments and lower model prices sharpen questions about cash returns and margins. Neither side establishes a sector-wide investment stop or a guaranteed payoff.
• The U.S.–China truce now has a January 10, 2027 end date. The announced extension changed the timetable, but the supplied research still lacks delivered rare-earth quantities, a full tariff schedule and definitive AI commitments.

What Matters Next

• Monday, September 28, 10:30 a.m. EDT — Dallas Fed Manufacturing Index. The dossier supplies no consensus or prior. It will add a regional manufacturing reading after the stronger national flash PMIs.
• Monday, September 28, after the close — Jefferies, Vail Resorts and IDT earnings. No estimates are supplied; listen for company-specific demand and financing commentary rather than treating one report as a macro proxy.
• Following week — consumer confidence, JOLTS, GDP revisions and PCE inflation. Dates and consensus estimates are not supplied. The inflation release matters alongside Michigan’s higher expectations; JOLTS and the labor data will test activity claims against household confidence.
• Friday of the following week — September jobs report. The dossier flags the release but provides no consensus. Payrolls, hours and revisions will help distinguish low claims from broader labor-market momentum.
• U.S.–Iran diplomacy and shipping flows — ongoing, with no confirmed deadline in the dossier. Watch for published terms, Oman or GCC acceptance, and comparable AIS, facilitated-transit and cargo/barrel measures. A negotiating headline without physical-flow evidence would not establish reopening.
• Rates and financing — next trading sessions. Watch whether the 10-year holds above 5%, retests Friday’s 5.23% high or retreats with oil; also watch project-debt pricing and disclosed construction, grid or financing delays. The evidence supplied does not identify a single yield level as a mechanical trigger.

Questions Worth Carrying Into Monday

• Do any U.S.–Iran terms emerge that change actual tanker passage and Gulf exports, or do negotiations remain ahead of the physical evidence?
• Can strong orders, PMIs and low claims persist alongside falling sentiment and higher inflation expectations—and which signal will be reflected in realized spending and hiring?
• How much of the 10-year’s rise reflects real returns and capital supply, and how much reflects inflation compensation or concern about future demand?
• Will AI customers’ gains from lower model prices produce enough additional usage to offset pressure on provider margins?
• Are private-credit redemption limits containing a liquidity problem, or will loan-level cash performance and marks show broader borrower stress?

Sources

• Briefing weekly economic calendar: September PMIs, claims, durable-goods orders and upcoming events.
• Trading Economics, September 25: Treasury yields, oil and equity reporting; durable-goods orders; Michigan sentiment and inflation expectations.
• CNBC quote records: Friday 10-year and two-year Treasury observations; early Saturday Brent indication, used only as an unchanged reference.
• Finviz, September 25 closing digests: equity closes, company developments and upcoming calendar.
• What’s Going on With Shipping?, Inside the 2026 Tanker War: source-reported Hormuz traffic, Gulf exports, maritime events and tanker rates.
• Chas Freeman and Trita Parsi interviews: differing assessments of the possible Oman-mediated terms and durability of an agreement.
• Art Laffer, David Busch and Henry Peabody: competing explanations of Treasury yields and credit conditions.
• Patrick Boyle: reported AI valuation, project-financing and model-pricing evidence.
• Eurodollar University: reported SoftBank and Project Jupiter financing, Apollo redemptions and private-credit measures.
• Danielle DiMartino Booth: attributed AI-investment-growth estimate.