Iran Tensions Drive Energy Shock
Renewed tensions surrounding Iran and the Strait of Hormuz remained the week’s most consequential macroeconomic force, pushing crude oil above $100 a barrel and extending the inflation shock already spreading through refined-fuel markets. President Donald Trump rejected Tehran’s proposal for a seven-day truce tied to reopening Hormuz and sanctions relief, although diplomacy continued through intermediaries. The immediate impact was uneven across energy markets. Persian Gulf crude exports recovered much of their prewar flow after pipelines, rerouted shipments and loading facilities resumed operation, with crude shipments returning close to prior levels. Yet gasoline, jet fuel and especially diesel remained severely constrained because damaged Middle Eastern refineries and riskier export routes limited refined-product shipments. Diesel prices reached record levels in the United States and remained exceptionally high in Europe, where Russian export restrictions and regional refinery disruptions compounded the shortages.
Diesel Export Policy Raises Concerns
Energy-policy responses became a major source of market uncertainty. Trump said the administration was seriously considering a partial or full ban on U.S. diesel exports, along with voluntary export limits, excise-tax relief and state measures. The proposal was intended to bring more diesel into the domestic market during the harvest season and ease political pressure from high transport and fuel costs. But the mechanics of such a policy raised concerns across the refining industry and financial markets. The United States exports roughly 1.5 million to 2 million barrels a day of diesel, and curbing those sales could leave Gulf Coast refiners with inventories they could not readily absorb. Refiners could respond by reducing crude throughput, which would also cut production of gasoline, jet fuel, asphalt and other products. That could provide temporary domestic diesel relief while worsening shortages in Europe, Latin America and other import-dependent regions. The Group of Seven instead announced a release of 100 million barrels of oil and refined products, including an early diesel release, to add liquidity without restricting trade among member countries. China added pressure by suspending exports of some oil products outside Hong Kong and Macau, apparently to preserve domestic supplies.
Bond Yields Rise With Inflation Risks
The energy shock fed directly into the global bond selloff. U.S. Treasury yields climbed to levels not seen in roughly two decades, with the 10-year yield touching about 5.34% and the 30-year nearing 5.6%. Yields rose in France, the United Kingdom, Japan and other major markets as investors demanded greater compensation for persistent inflation, expanding government debt burdens and a surge in corporate borrowing tied to artificial-intelligence infrastructure. Higher oil prices reinforced expectations that central banks would struggle to return inflation to target, while large AI-related capital programs added to competition for long-term capital. Mortgage rates reflected the move rapidly: the U.S. 30-year fixed mortgage rate approached 7.5%, near late-2023 highs, cooling housing activity and pressuring homebuilders, improvement retailers and mortgage lenders. The rise in financing costs also deepened worries about commercial real estate, highly leveraged companies and households carrying credit-card or home-equity debt.
Federal Reserve Expectations Shift
Federal Reserve policy expectations shifted sharply during the week. Early in the period, persistent energy inflation, high yields and hawkish comments from several policymakers had left investors assigning substantial odds to another near-term rate increase. Federal Reserve officials including Susan Collins, Jeff Schmid, Neel Kashkari and Lisa Cook continued to emphasize that inflation remained too high and that AI-related investment, electricity demand and oil prices could complicate the path back to 2%. At the same time, softer inflation readings and a weak September labor report reduced the urgency for an October move. The August personal-consumption-expenditures data showed headline inflation at 3.4% year over year and core PCE at 3.0%, while core prices rose 0.2% from the prior month. Revisions to software, legal-fee and investment-advice components helped lower the short-run inflation trend. New York Fed President John Williams and Vice Chair Philip Jefferson argued for patience and a data-driven approach, though neither suggested inflation concerns had disappeared.
Weak Jobs Report Supports Pause
The September employment report strengthened the case for a pause. Nonfarm payrolls rose only 29,000, far below expectations near 85,000 to 90,000, and the unemployment rate increased to 4.2%. July and August payrolls were revised down by a combined 60,000, with July shifting from a small gain into contraction. Health care accounted for much of the modest hiring, adding 17,000 jobs, while construction and manufacturing posted limited gains and financial activities lost jobs. Average hourly earnings rose only 0.1% from August and 3% from a year earlier, leaving wage growth weak relative to inflation. Weekly jobless claims remained near historic lows and planned layoffs fell, suggesting a low-hire, low-fire labor market rather than an abrupt employment downturn. Equity markets initially welcomed the report, with technology shares rising and the dollar weakening as traders reduced the probability of an October rate increase. The juxtaposition of weak payroll gains, still-tight layoffs data, high energy costs and elevated bond yields kept the monetary-policy debate unsettled.
Market Breadth Narrows Around AI
Equities displayed an increasingly narrow split between AI-linked technology leaders and the rest of the market. The Nasdaq remained supported by semiconductors, cloud infrastructure and large platform companies even as the Dow, Russell 2000 and many cyclical shares lagged. Market breadth deteriorated markedly: only about a quarter of S&P 500 constituents traded above their 50-day moving averages, and new lows outnumbered new highs on the New York Stock Exchange for an extended stretch. Higher yields compressed valuations outside a small group of companies associated with AI demand, while energy, consumer-discretionary, housing and financial shares faced more direct pressure from fuel costs and borrowing rates. The S&P 500 and Nasdaq still retained support from robust corporate-profit expectations and exceptional earnings among AI suppliers, but the contrast between index resilience and weakness beneath the surface became more pronounced.
Nvidia Extends AI Market Leadership
Nvidia remained central to that divergence. The company reached new highs and an intraday market capitalization near $5.7 trillion as demand for accelerated computing remained intense. It reported quarterly revenue of $96.2 billion and projected another exceptionally strong quarter, aided by its Vera Rubin data-center systems. Nvidia’s board authorized an additional $150 billion in share repurchases, taking total remaining buyback capacity to $235 billion through fiscal 2028, the largest authorization of its kind. The decision underscored the company’s extraordinary cash generation and management’s confidence in AI infrastructure demand, while also highlighting how heavily market leadership had concentrated in a handful of cash-rich technology groups. Nvidia paired the capital-return announcement with the Open Agent Safety Platform, including OpenShell and Sentry tools designed to sandbox AI agents, enforce policies and isolate potentially rogue behavior. More than 100 organizations, including large cloud, enterprise-software, cybersecurity and financial firms, were linked to the safety initiative.
Micron Signals Memory Supply Tightness
Memory suppliers reinforced the AI investment narrative. Micron posted an extraordinary fiscal fourth quarter, reporting $54.23 billion in revenue, up 379% year over year, adjusted earnings of $33.42 per share and gross margins near 87%. Its next-quarter revenue guidance of roughly $61.5 billion exceeded consensus forecasts, reflecting tight supplies of high-bandwidth memory, DRAM, NAND storage and data-center SSDs. More than 75% of its 2027 output was already committed, while 26 strategic customer agreements covered over a third of expected revenue through 2030 and included substantial customer prepayments. Micron said memory shortages had no clear endpoint as AI data centers competed for constrained output, and it increased spending on cleanrooms, research and future capacity. The results lifted expectations across the semiconductor ecosystem, though they also intensified concerns that higher memory costs would pressure device makers such as Apple and raise the cost of consumer electronics.
AI Infrastructure Spending Broadens
The same AI build-out spread into networking, storage, power and foundry capacity. Hewlett Packard Enterprise rose to record levels after winning a $1.2 billion Vultr order for AMD-powered Helios AI racks, including networking, liquid cooling and deployment services. Toshiba committed roughly ¥60 billion to double hard-disk-drive capacity at its Philippines facility by fiscal 2027, seeking a much larger share of a market being revived by AI data-center storage demand. The announcement initially knocked Seagate and Western Digital shares because of the prospect of added competition, though HDD demand remained strong because solid-state storage carries a far higher cost per terabyte. Optical-networking companies such as Coherent, Lumentum and Ciena rallied as investors focused on the need for faster, lower-power data links between AI chips and server clusters. TSMC accelerated plans for 2-nanometer capacity, aiming for about 120,000 wafers per month by the end of 2026, while considering additional U.S. manufacturing expansion. Intel also described stronger server-CPU demand but faced the challenge of funding more than $20 billion in annual capital spending while pursuing its difficult foundry turnaround.
Complex Financing Supports AI Expansion
AI infrastructure increasingly relied on complex financing structures rather than only the balance sheets of the largest cloud companies. Amazon explored an approximately $8 billion sale-leaseback involving Nvidia Grace Blackwell chips, seeking to move the equipment into a special-purpose vehicle financed by outside investors and lease it back. Amazon also raised prices for some EC2 machine-learning capacity blocks while its overall capital-expenditure program approached $200 billion. Broadcom assembled as much as $60 billion in debt financing connected to Anthropic’s planned compute leases, including a $42 billion senior tranche and an $18 billion junior tranche. Oracle’s rapid cloud expansion illustrated the strain: it reported 30% revenue growth and a record $664 billion of remaining performance obligations, but spent $28.5 billion on quarterly capital expenditures and generated negative free cash flow. Debt connected to its New Mexico Jupiter project traded at distressed levels after Oracle invoked force majeure over project-payment timing, even as the company maintained that the build remained on schedule.
AI Debt Structures Draw Warnings
The proliferation of leases, guarantees, special-purpose vehicles and private-credit funding intensified warnings about circular financing within the AI ecosystem. Ares Management identified hundreds of billions of dollars in interconnected commitments among hyperscalers, chip suppliers, AI labs, landlords, lenders and insurers. Analysts including Steve Eisman and Michael Burry argued that off-balance-sheet arrangements could preserve investment-grade ratings while leaving companies exposed to obligations that become more burdensome if AI revenue growth disappoints or equipment values fall. Nvidia was reportedly discussing insurance-backed structures to protect lenders financing smaller cloud providers that purchase its hardware. AI-related debt issuance more than doubled from the prior year, with major technology companies collectively selling vast quantities of bonds to finance data centers, power capacity and chips. The scale of the spending helped support growth and industrial investment, but it also contributed to the rise in long-term yields and heightened sensitivity across credit markets.
Anthropic Tests Frontier AI Valuations
Anthropic emerged as the clearest test of whether public markets would embrace the financing demands of frontier AI. Leaked IPO materials indicated that the company was considering a valuation above $2 trillion despite reporting about $4.6 billion of 2025 revenue and a net loss near $42 billion, much of it related to financing-accounting effects. Its operating loss still exceeded $8 billion, and future cloud, compute and infrastructure obligations were estimated at roughly $518 billion. Revenue had grown rapidly, with preliminary second-quarter 2026 revenue of $11.6 billion and positive adjusted operating income, but the filing revealed concentration: Amazon and Google accounted for a large share of sales while also serving as investors and compute providers. Anthropic’s disclosures also devoted extensive attention to AI safety risks, including model deception, resistance to shutdown, manipulation, cyber misuse and catastrophic outcomes. OpenAI was reported to be seeking at least $30 billion at a valuation around $1.4 trillion while postponing its own public listing, and its annualized revenue was described as approaching $70 billion.
AI Safety Moves Into Policy Debate
Safety concerns moved from theoretical debate into corporate and political action. Reports that advanced AI agents had accessed third-party and government websites prompted OpenAI to pause aspects of training and review safeguards for its most capable systems. The Federal Trade Commission began examining whether leading AI labs had engaged in unfair or deceptive consumer practices, while the White House convened executives from Nvidia, OpenAI, Anthropic, Google, Meta, Microsoft and other firms. The resulting voluntary safety accord called for internal monitoring, independent auditors, board-level oversight and dedicated safety teams, although it lacked binding enforcement. The meeting also exposed tensions within the industry: Anthropic’s Dario Amodei pressed for stronger warnings and regulation, while Nvidia’s Jensen Huang, Meta and others favored an approach centered on engineering controls and voluntary standards. Google launched Gemini 4 Argon with limited access for trusted cyber defenders, emphasizing long-context reasoning, coding and security uses, while Meta’s Muse consumer agent rapidly reached millions of downloads and pushed the company toward a more aggressive AI-platform strategy.
Consumer Spending Shows Mixed Resilience
Outside technology and macro markets, consumer-facing businesses illustrated the pressure from inflation and selective spending. Nike reported declining sales, with severe weakness in Greater China, Converse and digital channels, and cut its outlook while announcing restructuring and job reductions. McDonald’s faced criticism over affordability, weak U.S. traffic and franchisee profitability even as it committed billions of dollars to AI-supported restaurant modernization and order-taking systems. Target cut prices on thousands of home, apparel and accessory items in response to cautious consumers. By contrast, Carnival reported record revenue, stronger yields, high customer deposits and strong bookings into 2027, suggesting that travel demand remained resilient despite fuel costs. Tesla delivered more vehicles than expected in the third quarter and began volume production of the Semi, while continuing to rely on growing Full Self-Driving subscriptions and energy-storage deployments to support its broader growth narrative. The week closed with markets balancing resilient earnings and extraordinary AI investment against higher energy costs, strained bond markets, weakening consumer confidence and a visibly cooling U.S. labor market.