US "inflation fever" but "consumption strain"! PCE and GDP suggest the Fed is caught in a dilemma.

On Thursday, the U.S. Commerce Department's Bureau of Economic Analysis (BEA) delivered two reports at once: the preliminary estimate of third-quarter real GDP growth at an annualized quarter-on-quarter rate of 3.4%, above economists' forecast of 3.2% and the prior reading of 2.2%; the September PCE price index rose 0.5% month-on-month and 3.6% year-on-year, while core PCE, excluding food and energy, rose 0.3% month-on-month and 3.1% year-on-year, with all four readings landing within market expectations. The economic picture First, the GDP components (all at annualized quarter-on-quarter rates): personal consumption expenditures rose 3.8%, maintaining that pace for a second consecutive quarter; business investment rose 12.7%, the strongest since 2020, with both information-processing equipment and industrial equipment posting sharp gains; the contribution of equipment spending to GDP was the highest since 2011excluding the post-pandemic rebound period; net exports were a 3.3-percentage-point drag, the largest since early last year; inventories contributed 1.6 percentage points; residential investment remained a drag, after mortgage rates had at one point risen to 7.5%. Now the September income and spending data: personal income rose by US$54.5 billion, up 0.2% month-on-month (market expectation 0.3%); disposable personal income rose by US$52.7 billion, up 0.2%, but real disposable income fell 0.3%; personal consumption expenditures rose by US$113.1 billion, up 0.5% in nominal terms (market expectation 0.6%), while real spending rose by only US$2.6 billion, an increase of less than 0.1%, essentially zero; personal saving was US$911.1 billion, and the saving rate fell to 3.8%. On the inflation side: the PCE price index rose 0.5% month-on-month (0.3% in August) and 3.6% year-on-year (3.4% in August); core PCE rose 0.3% month-on-month and 3.1% year-on-year, the highest since May, versus 3.0% in August. AI equipment investment and inventory restocking The growth structure this quarter was highly concentrated. The contribution of equipment spending to GDP was the highest since 2011, and this was heavily concentrated in information-processing equipmentthat is, AI data center-type investment. The 12.7% growth rate in business investment was the strongest since 2020. In other words, if AI-related spending were stripped out, this quarter would be far less hot than 3.4%. The 3.3-percentage-point drag from net exports looks contradictory but stems from the same source: much of the equipment that AI investment relies on must be imported, and the surge in imports of data center equipment itself depressed net exports; combined with companies front-loading shipments to avoid tariffs, the import side was pushed even higher. The 1.6-percentage-point inventory contribution likewise has a one-off charactergoods rushed in will eventually have to be worked off in later quarters. And residential investment, shrinking under 7.5% mortgage rates, is the only unambiguously negative item in this report. This also explains the severe divergence between the data and how it feels: the GDP reading is one of the strongest in three years, yet consumer confidence as measured by the University of Michigan and the Conference Board is at historic lows, and Trump's approval rating is at a low for his termand this report was published less than a week before the November 3 midterm elections. Some economists are blunt: given the growing weight of AI-related spending, 3.4% may not be the true picture of the economy. The real problem is consumption: the divide between nominal and real More alarming than the GDP components is the crack in the September income and spending accounts. Nominal spending rose 0.5%, while real spending was roughly zero; nominal income rose 0.2%, while real disposable income fell 0.3%. Income growth has fallen to its lowest range since April 2022, yet consumption is still growing, and the gap can only be filled by saving: the saving rate was 4.6% in July (revised), 4.1% in August (the lowest since November 2022), and fell further to 3.8% in September. There is also a timing mismatch that must be pointed out: GDP covers July through September, but real consumer spending had already stalled in September. In other words, 3.4% is a rearview mirror, while zero growth is the current readingthe strong momentum of the third quarter had already lost its handoff from the consumption side as it entered the fourth quarter. And the pressure is still mounting. The U.S. average gasoline price remains above US$4 per gallon; in September, gasoline and other energy commodity prices rose 9.2% month-on-month (4.4% in August), and energy spending rose in six of the first nine months of this year. Observations from the corporate side point to the same divergence: UPS CEO Carol Tom said on this week's earnings call that as long as employment remains stable, "consumers will spend because they have assets," but she also stressed that those without assets are already struggling; Coca-Cola said households across income levels are prioritizing "value." The previously released September retail sales total beat expectations, which is easily read as "consumption remains strong." Breaking down the structure yields a much more cautious conclusion. First, the 0.8% increase was highly dependent on the gasoline station category alone. Excluding that, it was only 0.2%; excluding autos, building materials and food services as well, only 0.1%. Retail data are seasonally adjusted but not inflation-adjusted; if the 0.6% monthly price increase in September is factored in, real retail sales were "flatter than flat." Second, real purchasing power is declining. Reports noted that inflation-adjusted wages fell again in September; consumers are still spending largely because of high-income households supported by stock gainsthe so-called K-shaped economy. The Fed's Beige Book released on Wednesday confirmed this: from late August to early October, most districts reported consumer spending "changed little or not at all"; even in districts such as Dallas where businesses reported solid retail sales, there were "some reports of more cautious spending by lower-income customers." Inflation: energy is driving the headline, and the core is also picking up The headline PCE year-on-year rate rose from 3.4% to 3.6%, almost entirely attributable to the energy component. But that is not the whole answer. Core PCE rose from 3.0% to 3.1% year-on-year, the highest since May; and in the August data, "supercore" services inflation excluding housing costs had already reaccelerated. This means that even if statistical adjustments have reduced the weight of some goods prices, domestic price pressure on the services side has not disappeared. Middle East conflict and the Russia-Ukraine war are pushing up fuel prices, and the Trump administration's tariffs are raising some goods prices; both channels are feeding through to the core. This report means PCE has now been above the Fed's 2% target for 68 consecutive monthselevated inflation has persisted for nearly six years. The Fed: a December hike is already priced, with disagreement centered on 2027 On October 28, the Fed held rates steady by a 10-2 vote, with the two dissents coming from Governor Lisa Cook (her first dissenting vote in her tenure) and Cleveland Fed President Beth Hammack, both of whom favored a 25-basis-point hike. The Fed had just hiked once in September, the first in three years. Three signals are worth noting: the statement upgraded its description of economic activity from "solid" to "expanding at a strong pace," while reiterating that "inflation remains somewhat elevated"; Chair Kevin Warsh said at the press conference that "economic activity is expanding at a strong pace" and "domestic spending remains resilient," and said the next six weeks would be an "information-dense" period; he also explicitly rejected forward guidance, announcing that starting in 2027, one of the FOMC's eight annual meetings will become a pure strategy discussion, with no rate vote and no policy statement. It should be noted that the 3.4% growth rate is clearly above the roughly 2% "non-inflationary growth rate" that some Fed officials believe in. CME's "FedWatch" showed that after the decision, the probability of a 25-basis-point hike in December briefly rose to 85.0% (78.3% the previous day), and remained around 81% after the GDP and PCE releases. Institutional disagreement is concentrated on 2027: Kay Haigh, global head of fixed income and liquidity solutions at Goldman Sachs Asset Management, expects the Fed to resume hiking in December and possibly tighten further next year if energy prices rise again or data remain strong; former Philadelphia Fed President Patrick Harker expects a 25-basis-point hike in December followed by no change in 2027; Ian Lyngen of BMO Capital Markets believes a third hike after December may not be necessary. The tail risk was raised by Brent Schutte of Northwestern Mutual: if October inflation shows another hot reading similar to October's, the market will begin to question the Fed's commitment to its 2% target, and a 50-basis-point hike in December may be needed. Political constraints also exist. Inflation and the cost of living are core issues in this election campaign, and the November 3 midterm elections will determine control of Congress, including oversight of the Fed. Holding steady before the election both avoids being interpreted as political interference and leaves time to observe data; but once the election is over, that constraint disappears. Impact on U.S. stocks: earnings are the support, and AI cash flow is being repriced Recently, a chip chain spanning equipment, foundries, memory and finished devices fell in unison, showing that the market is cutting the same variablenot demand, but the money that supports demand. What the market is hearing now is not "demand is good," but "good demand must first turn into bills." This is exactly the narrative shift in this round of AI trading: over the past two years, whoever confirmed orders rose; now, whoever confirms spending falls. At present, what is weighing on U.S. equity macro sentiment is not a demand problem but a financing problem. Alphabet's third-quarter revenue was US$127.58 billion, up about 25% year-on-year, but below analysts' estimate of US$128.2 billion; more glaring was free cash flow of negative US$9.07 billion, negative for a second consecutive quarter, the first time since its listing, with the stock down more than 11% at one point premarket. Microsoft's quarterly revenue was US$92.28 billion, up 19%, with Azure growing 48%, but full-year capital expenditure was about US$170 billion. The AI capex narrative is shifting from "story" to "cash flow"this is the micro foundation for Nasdaq futures falling far more than the Dow. Premarket, the AI hardware chain fell broadly: Marvell Technology dropped more than 12% (on reports it is in talks to acquire data center optics maker Lumentum), CoreWeave fell more than 6%, and Intel, Seagate and ARM fell more than 3%. Support does indeed exist. LSEG data show that 200 S&P 500 constituents have reported third-quarter results, of which 84% beat market expectations, and the expected overall year-on-year earnings growth rate has been raised to 31.8% from 30.6% on October 1. In a high-rate environment, earnings are the main reliance for U.S. stocks to digest valuation pressure. The summary by Parth Talsania, CEO of Equisights Research, is the most precise: this selloff is "a rates event dressed in credit clothing," CoreWeave is the "most leveraged" sample of this round of rate pressure, not "the first evidence of a crack in AI demand"; "companies that collect in cash are protected." U.S. Treasury yields remain elevated The 10-year Treasury yield closed at 5.021% on October 28, having briefly breached 5.3% earlier this month, a 24-year high; the 2-year yield fell below 4.7% during Warsh's press conference. The combination of a lower short end and a higher long end has a clear meaning: the market is cutting near-term hike bets, while the long end above 5% reflects not "how many more hikes" but term premium, Treasury supply and inflation stickiness. In other words, "the Fed not hiking" has not truly eased valuation pressure on stocksthe pressure comes from the long end. Mike Medeiros of Wellington Investment Management warned that Warsh is dovish overall but his communication has been contradictory, meaning that at some point after the midterm elections, term premium could reaccelerate upward. Three potential judgments First, the 3.4% is of poor quality, but sufficient to prevent the Fed from pivoting. Growth was propped up by AI equipment investment (contributing the highest since 2011) and one-off inventory restocking (1.6 percentage points), while net exports were a 3.3-percentage-point drag and residential investment continued to shrink; excluding AI, the economy is far less hot than the reading. But the picture the Fed sees is "demand not cooling + supply hit," precisely the combination least supportive of a policy pivot. Second, inflation cannot be accounted for by energy alone. In the headline 3.6%, the 9.2% month-on-month rise in energy commodities was the main driver; but the core rising from 3.0% to 3.1%, and supercore services having already reaccelerated earlier, show that pressure is feeding through to domestic prices. Being above target for 68 consecutive months is itself a policy credibility problem. Third, for assets, direction is determined by long-end rates and AI cash flow, not by this hike. The 83% probability of a December hike is fully priced, with limited marginal impact; the real variables are twowhether the 10-year Treasury can hold below 5%, and whether tech giants can turn AI capex into free cash flow. Google's negative US$9.07 billion in free cash flow is a signal.