Weekly market perspective
Week ending September 4, 2026
Last week, our conclusion was that the market’s central argument remained earnings growth versus the cost of capital, with earnings still winning—but by a narrowing margin.
This week changed that assessment.
Earnings did not deteriorate. In fact, the evidence supporting the AI investment cycle arguably strengthened. What changed was the other side of the equation. Oil moved sharply higher again, sovereign yields returned toward their recent extremes, the U.S. economy produced materially stronger growth data than expected, and by Friday the market was once again contemplating a Federal Reserve rate increase in September.
SPY and QQQ consequently went almost nowhere for the week despite some unusually violent movement underneath them. The S&P 500 finished September 4 at 7,718.13 versus 7,711.76 the previous Friday, while the Nasdaq Composite ended at 26,505.44 versus 26,402.42. A nearly unchanged index therefore disguised a substantial repricing of several of the variables that matter most to us.
The important conclusion is different from last week:
Earnings are no longer clearly winning the argument with rates. For the moment, they are holding the line.
What Actually Moved the Market
At first glance, this looked like another geopolitical week.
The United States and Iran exchanged strikes again, shipping through the Strait of Hormuz remained severely impaired, crude moved sharply higher and global risk assets initially reacted accordingly.
But geopolitics itself was not the primary equity transmission mechanism.
Oil was. And oil mattered because rates cared about it.
Brent gained more than 6% during the week and WTI more than 8% by Friday morning; later settlement data showed still larger weekly advances. U.S. diesel prices reached record levels. Tanker traffic through Hormuz remained far below normal, although analysts noted that the renewed fighting had not yet produced a corresponding collapse in actual Middle Eastern oil exports. In other words, a meaningful portion of the move remained risk premium rather than realized physical shortage.
That distinction is important because it tells us what to watch next.
If hostilities cool and crude gives back the risk premium, the inflationary impulse can disappear surprisingly quickly. If actual barrels begin disappearing, however, the character of the problem changes.
This week the market increasingly treated the geopolitical story as:
Iran → oil → inflation → yields → equity valuation.
That is considerably more useful than simply observing that stocks fell when missiles were flying.
The Most Important Change Was in Bonds
Last week we concluded that the stress in the long end of the Treasury market had stopped worsening.
That conclusion has to be revised.
By Wednesday, the U.S. 10-year yield was roughly 4.80%, within about a basis point of its 2025 high, while the 30-year had returned near a 19-year extreme despite Treasury's earlier effort to calm long-duration markets. The move was global: Japanese, German and British yields simultaneously reached levels not seen in decades.
This matters because the breadth of the move makes a purely U.S.-specific fiscal explanation less convincing.
Government deficits remain part of the problem. So does energy inflation. But something larger may also be occurring: the world's equilibrium cost of capital may simply be migrating upward.
One reason is investment.
The AI boom is beginning to connect the equity and Treasury stories in a way that deserves much more attention. Governments are issuing enormous amounts of debt at the same time hyperscalers are competing for the same capital to finance data centers, chips, power generation and networking infrastructure. Reuters reported this week that large technology companies' borrowing for AI investment is adding pressure to sovereign markets because these companies are willing to pay sufficiently attractive rates to pull capital away from government bonds.
That creates a fascinating circularity.
AI investment supports growth.
Growth supports earnings.
AI investment requires capital.
Capital demand raises yields.
Higher yields reduce the present value of those earnings.
We have been treating AI capex and Treasury yields as interconnected variables. The market is increasingly providing evidence that they actually are.
The Economy Is Not Providing the Fed an Escape Hatch
The economic data made the rates problem harder, not easier.
August services activity accelerated sharply. The ISM services index rose to 55.4 from 54.1, while new orders jumped to 60.9, their strongest reading in three and a half years. More uncomfortable for the Fed, the prices-paid component increased to 72.6. ISM specifically identified strong domestic demand—including AI-related spending—as one contributor to the strength.
Then Friday's employment report eliminated much of the remaining argument that the labor market was deteriorating rapidly.
The economy added 162,000 jobs in August, almost three times the Reuters consensus of 56,000. June and July were revised upward by a combined 55,000. Unemployment remained 4.1%, labor-force participation improved, the household survey showed substantial employment growth, and the workweek lengthened.
There is an important nuance, however.
Wage growth actually moderated to 3.1% year over year. This does not look like a classic wage-price spiral. The inflation problem increasingly appears to involve energy, supply constraints, tariffs, capital demand and persistent service-sector pricing rather than runaway employee compensation.
That distinction matters economically.
It matters somewhat less to the Fed in the immediate term.
A central bank confronting inflation above target does not need to see wage acceleration if demand remains strong enough to permit businesses to pass higher input costs through to customers.
By Friday afternoon, markets were assigning roughly a 58% probability of a 25-basis-point September rate increase, up from about 49% the day before. The 2-year Treasury moved back toward 4.38% and the 10-year toward 4.78%.
So the macro problem has evolved again.
It is no longer primarily:
Can the economy avoid recession?
The more relevant question is becoming:
Can inflation decline sufficiently while the economy refuses to slow down?
AI Demand Passed Another Test
If the bear case depended upon AI spending suddenly rolling over, Broadcom did not cooperate.
Broadcom increased its fiscal-2027 AI-chip revenue expectation to approximately $115 billion from $100 billion and expects that figure to roughly double to $230 billion in fiscal 2028. Third-quarter AI-chip sales exceeded $16 billion, more than tripling from the prior year, while quarterly AI bookings exceeded $30 billion.
Perhaps more interesting were the physical-capacity numbers.
Broadcom described visibility into more than 10 gigawatts of AI infrastructure for Anthropic, more than 5 GW for OpenAI and approximately 3 GW for Meta. Those are not abstract software forecasts. They imply enormous requirements for electricity, data centers, financing, cooling, networking hardware and semiconductors.
The market reaction was instructive. Broadcom's stock initially fell despite those extraordinary numbers because near-term revenue guidance was slightly below expectations.
That tells us something important about QQQ.
The AI thesis can remain fundamentally right while AI equities become increasingly difficult to please.
Expectations are already enormous.
We therefore continue to believe the AI buildout itself is real, but the relevant investment question is migrating from “Will the spending occur?” toward “Who earns an adequate return on the capital required to fund it?”
There is another implication that may ultimately prove more important.
The AI boom is now large enough that it may be affecting the risk-free discount rate against which AI equities themselves are valued.
That is not a contradiction. It is an equilibrium mechanism.
Credit Still Does Not Confirm a Crisis
For all of the drama in sovereign bonds, public corporate credit remains remarkably calm.
ICE BofA U.S. high-yield option-adjusted spreads widened only modestly, from approximately 2.60% on August 28 to 2.66% on September 2. BB spreads were essentially unchanged.
That continues to be one of the strongest arguments against declaring the beginning of a generalized financial accident.
Treasuries are under pressure.
Oil is elevated.
The Fed may hike.
Yet investors are still not demanding dramatically greater compensation to own risky corporate debt.
We would become much more defensive if that changed.
Private credit remains a separate concern because much of it is floating-rate debt. Every additional month of elevated short-term rates keeps pressure on borrowers that were underwritten under materially easier financing assumptions. But this week provided no evidence that private-credit stress has crossed into public-credit contagion.
For now:
Rates are a valuation problem and an economic headwind. They have not yet become a credit event.
The distinction is crucial.
Volatility Is Sending a Strangely Calm Message
Perhaps the most curious market signal remains volatility.
The VIX ended the week around 14, even with Brent near the mid-$90s, a potential Fed hike two weeks away, Treasury yields approaching major historical levels and an active military conflict affecting the world's most important energy corridor.
Meanwhile investors continued withdrawing money from equities.
U.S. equity funds lost another $11.12 billion during the week through September 2 following roughly $22.7 billion of withdrawals the previous week. Technology, financials and industrials all experienced selling. Yet almost $49 billion flowed into money-market funds.
That combination looks less like panic than optionality preference.
Investors are raising cash.
They are not indiscriminately dumping risk.
And the distinction shows up directly in the VIX.
These developments coexist with another observation. Equity volatility is extraordinarily subdued relative to the volatility occurring in oil, sovereign bonds and geopolitics.
Markets can tolerate that divergence for quite some time.
They rarely tolerate it forever.
Earnings and financing conditions
Last week, our broad assessment was that earnings growth still outweighed the pressure from rates. This week, that margin narrowed.
Looking across the coming months, our outlook is:
Constructive on the earnings foundation, but more cautious about the valuations those earnings can support.
That is why QQQ can feel heavy even when the AI data itself is excellent.
Friday provided an encouraging detail: semiconductors materially outperformed despite the hawkish employment surprise. That suggests investors still want AI exposure when the price is attractive. But software and services simultaneously lagged, reinforcing the increasingly important distinction between AI beneficiaries and companies threatened by AI substitution.
For investors assessing SPY and QQQ over the coming months, the central question is whether earnings can grow fast enough to offset a persistently higher cost of capital. SPY offers broader exposure to the economy; QQQ is more concentrated in the growth companies at the center of both the AI opportunity and the financing challenge.
Our base case remains one of resilient earnings facing a more demanding valuation environment. Strong AI demand and orderly credit support that view. Persistent energy inflation and elevated yields would narrow the room for further valuation expansion, even if corporate profits continue to grow.
The strategic distinction is between confidence in business growth and confidence in the price paid for that growth. This week strengthened the evidence for the former while making the latter more demanding.
What We Are Watching Next
The next decisive input is the August CPI report on September 11, followed by the Federal Reserve meeting ending September 16.
A benign inflation report accompanied by a meaningful decline in crude could unwind a substantial portion of the rate pressure very quickly. In that scenario, QQQ in particular would have both the earnings numerator and the discount-rate denominator moving in its favor.
The opposite scenario is more dangerous.
If inflation remains sticky, crude remains around $90–$100, and the 10-year begins challenging 5%, the market will have to decide whether current equity multiples adequately compensate investors for an increasingly attractive risk-free alternative.
That is the line we are watching.
The biggest change from last week is therefore not that the equity fundamentals became worse.
They didn't.
The change is that higher bond yields have raised the return equities must offer to remain attractive.
And when the risk-free rate keeps rising, eventually every asset has to renegotiate its price.
Selected source links
September 4, 2026 employment report and market reaction: Reuters — U.S. payrolls surge in August
September 4, 2026 market close: Reuters — Wall Street falls as Fed-hike bets increase
September 2, 2026 AI/capex update: Reuters — Broadcom raises AI-chip forecasts
September 2–3, 2026 bond-market update: Reuters — Global bond selloff deepens
September 4, 2026 energy update: Reuters — Oil surges for the week as U.S.-Iran tensions return
Previous issue
August 28, 2026 — Rates vs. earnings
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.