The market and its bodyguard.

Less urgency for another Federal Reserve increase has not resolved the market's cost-of-capital problem. Softer employment helped equities recover on Friday, but long-term Treasury yields remained demanding. For the coming months, the question is whether improving earnings expectations and AI demand can keep absorbing expensive financing, persistent operating costs and growing commitments to infrastructure.

A bodyguard can become concerned long before the person being protected notices anything wrong. After years of keeping that person safe, the guard begins checking rooftops, changing routes and keeping the car closer. Nothing has happened to the protected person. Indeed, keeping that person unaware of danger may be part of a successful day's work. But the guard's behavior has changed, and that change contains information.

Credit deserves the same attention. Lenders protect capital by asking for more collateral, accepting less leverage, charging more or becoming more selective. These are precautions, not proof of impending failure. Yet if enough lenders take them together, transactions shrink, investment is delayed and borrowers preserve cash. The market can still look healthy while the people protecting its financing are becoming less comfortable. Their response can eventually change the conditions they are trying to survive.

Fed relief, without an all-clear from bonds

September payrolls increased by 29,000, unemployment reached 4.2%, and average hourly earnings rose 3.0% from a year earlier. July and August payroll estimates were revised down by a combined 60,000. The Bureau of Labor Statistics' October 2 report gives policymakers a reason to weigh slowing employment alongside inflation.

Reuters' post-report update, published at 9:04 a.m. Eastern, put the futures-implied chance of an October increase below one in five, with a December increase still close to 90%. Those are dated morning probabilities, not Friday closing prices. They show less immediate tightening pressure, not an assured easing cycle. A probability for one meeting also differs from the cumulative policy change expected across a year. Reuters, October 2.

The ten-year Treasury had reached 5.34% on Thursday, according to Reuters' October 1 report. Friday's initial decline in yields after payrolls subsequently retraced as oil recovered. The S&P 500 rose 0.7% on Friday but fell 0.3% for the week; the Nasdaq Composite gained 1.2% on Friday and 0.5% for the week. AP closing report, October 2.

The interpretation matters more than a particular yield threshold. A long-term yield incorporates expected short rates, compensation for uncertainty and the price required to attract buyers of duration. Treasury supply, inflation risk, global borrowing and private investment demand can keep that price high even when the next Fed increase becomes less likely. This week's response supports giving those channels greater weight; it does not identify a precise term-premium contribution or prove that monetary policy has stopped mattering.

For borrowers, the practical test is the cost and availability of money over the life of a project. A less threatening signal from the Fed helps. It does not automatically make lenders comfortable with the project's leverage, collateral or future cash flows.

Earnings support is real—and increasingly important

FactSet's October 2 preview estimates S&P 500 third-quarter earnings growth at 29.5% and full-year 2026 growth at 32.4%. Third-quarter revenue growth is forecast at 12.3%, with earnings growth expected in all eleven sectors. Analysts raised quarterly EPS estimates 1.4% during the quarter. These are forecasts, not reported results, but upward revisions indicate improving expectations for corporate profits.

FactSet also reports a forward twelve-month P/E of 19.0 using Wednesday's closing price. A multiple near a historical average is not proof of fair value: the return available on competing assets and the reliability of forecast earnings both matter. Strong growth can offset some discount-rate pressure. It cannot make the discount rate irrelevant.

Our assessment therefore depends on delivery rather than the size of the headline forecast. An earnings beat against a lowered estimate is less useful than sustained forward estimates, credible guidance and cash available after investment and financing costs. The next few reports will be evidence, not a verdict on the entire season.

The same earnings multiple also means something different when safe long-term yields are above 5% than when they are near 1% or 2%. There is no universal formula for translating that difference into P/E points. The principle is simpler: investors have more attractive alternatives, and distant or uncertain profits must earn their place. Strong earnings are a real defense, but the amount of valuation and debt resting on that defense matters too.

The scorecard for the coming months

We will compare subsequent evidence with the October 2 expectations above and evaluate five connected tests:

1. Forward estimates: Do the level and growth of expected earnings hold up, or do downward revisions erode the apparent valuation cushion?
2. Breadth: Are revisions improving across sectors and companies, or does a small group carry the aggregate?
3. Operating delivery: Do revenue, margins and forward guidance support the earnings trajectory as energy, labor and financing costs pass through?
4. Cash conversion: Does profit become cash available for debt service and investment, after working capital and contractual commitments?
5. Market confirmation: Do earnings-sensitive shares and corporate credit remain resilient when Treasury yields rise?

Support is holding if estimates, guidance and cash generation broadly sustain or improve. Pressure is gaining the upper hand if weaker margins or cash conversion combine with declining forward estimates, tighter credit and lower valuation multiples. Where results are not available, the score remains unresolved. One week's price move cannot settle a multimonth thesis.

AI demand and AI credit are different underwriting questions

AI adoption can be economically important while some infrastructure investments earn inadequate returns. The investor must follow four questions: Who owes the money? Who pays? Who captures the margin? When does the cash arrive? Aggregate technology earnings do not automatically service a separate data-center borrower's debt.

Oracle illustrates both the opportunity and the necessary distinctions. Its August 31 filing reports $664 billion of remaining performance obligations, with about 13% expected to become revenue within twelve months. It also reports $11.4 billion of quarterly customer prepayments with a significant financing component. Those prepayments are meaningful funding support and counterevidence to a blanket claim that the backlog is unfunded. But backlog, recognized revenue, cash receipts and cash available after delivery costs remain different quantities. Oracle Form 10-Q.

For any heavily committed infrastructure borrower, the review should examine customer concentration, funded versus unfunded commitments, enforceability, utilization, delivery costs and the timing of receipts. It should also test pricing competition, equipment replacement and refinancing. A long contract can provide visibility while still leaving construction, customer or technology risk. The decisive protection is cash reaching the legal entity that owes the debt, in sufficient amounts and on the required timetable.

Bain's September 29 analysis frames the scale conditionally: $1.5 trillion of potential annual infrastructure investment by 2031, at capital spending equal to 25% of revenue, would require a $6 trillion annual revenue pool. Its estimate for existing consumer and enterprise applications is $1.2–1.8 trillion, leaving a large requirement for additional applications. The $6 trillion is a model-implied requirement, not demonstrated demand; the application estimate is a future forecast, not current sales.

That is an argument for examining monetization and the distribution of returns, not for declaring AI a failure. Valuable technology can generate enormous customer benefits while competition limits supplier margins. Equity gains at leading platforms can coexist with losses at leveraged developers, equipment owners or lenders. Those differences require examining each balance sheet and business model separately.

Here, a lender need not conclude that AI will fail. It may simply require more equity, firmer customer commitments or a higher return before funding the next project. That can be rational underwriting and still slow the buildout. The size of the investment commitments makes modest changes in funding assumptions consequential.

Credit is an early test of the remaining cushion

The ICE BofA U.S. High Yield option-adjusted spread series published through FRED widened over the observations from September 25 through October 1, the latest available observation in the October 2 update. That is a developing warning rather than evidence of a funding crisis. The next question is whether widening persists and reaches refinancing terms, lender behavior and company guidance.

Credit does not have to collapse. It can simply become more rational. A borrower may still obtain financing, but at a price or on terms that make a smaller acquisition, slower expansion or delayed investment the sensible choice. Across many borrowers, those individual decisions can weaken the growth assumptions supporting today's equity prices.

That is why we are watching a pattern rather than waiting for one spectacular default. Wider spreads alone do not establish that lenders are closing their doors. Confirmation would come from tighter terms, smaller commitments, more collateral demands and management decisions to conserve cash. Several related signals may reflect a common pressure; they should not be counted as independent proof that failure is inevitable.

When protection changes the environment

The distinction between vulnerable and broken matters. A system can continue functioning after much of its reserve has disappeared. What happens next depends both on the weakness and on the load above it: leverage, demanding valuations, refinancing needs and investment commitments. A small disappointment under a heavy load can matter more than a dramatic headline in a well-capitalized system.

Consider the feedback loop. Lower asset prices reduce collateral values. Lenders respond with less credit or more protection. Borrowers trim investment, hiring or purchases, weakening cash flows and growth. That can put further pressure on asset prices. Each decision can be reasonable for the institution making it, while their combined effect makes the environment harder for everyone.

This is the bodyguard's dilemma. Changing the route is a precaution. Restricting every movement eventually changes the protected person's life. In markets, financing can become restrictive well before defaults make the danger obvious. An initial repricing need not begin as a crisis to create conditions in which further stress becomes more likely.

None of this establishes that such a spiral is underway or unavoidable. Profitable companies, improving earnings, committed customer cash and stable financing can interrupt it. The question is how much ordinary disappointment the structure can absorb. A delayed project, a refinancing at a higher rate, expensive energy or a cautious consumer may be enough to expose thin assumptions. We do not need to invent an extraordinary disaster; we need to ask what happens if ordinary life fails to match a demanding plan.

Evidence to watch

The questions are whether earnings revisions broaden, cash generation holds and financing remains available on workable terms. Deterioration in forward estimates alongside rising refinancing costs would indicate increasing pressure. Improving cash flows and stable lending terms would point to greater capacity to absorb it.

For technology investment, the relevant distinctions are profitability, capital intensity, customer commitments and reliance on refinancing. These are questions about economic resilience and financing structures, rather than a directional view on an index, fund or security.

HER4 currently places more weight on rates, inflation, financing and earnings delivery than on historical midterm seasonality; election-related policy uncertainty can be a catalyst, but seasonality is not an independent guarantee of a rally.

The evidence remains mixed. Equities continue to absorb demanding conditions, and the earnings evidence deserves respect. Yet relief about the next Fed meeting does not by itself repair the cost of financing long-lived assets. Durable earnings, cash conversion and the behavior of credit will decide how much of that pressure the market can continue to bear.

The protected person remains fine. Watch whether the people protecting that person are behaving differently. A change in their behavior is not a prediction of catastrophe. It is a reason to reassess the remaining reserve, the commitments resting on it and the consequences of a more cautious financing system. We will give improving evidence the same weight as deterioration. For now, resilience deserves respect, and the price of maintaining it deserves closer attention.

Previous issues

September 24, 2026 — Higher costs are becoming the environment.

September 18, 2026 — The lines were crossed. The market held.

September 11, 2026 — Oil raises the stakes for earnings and valuations

HER4 Market Interactions is general market commentary, not investment advice or a recommendation to buy or sell any security. Views are as of the issue date and may change as conditions evolve.