Market Overview September 29, 2026
Summary
Shanghao Jin believes the pageantry of the U.S.-China talks far exceeded the substantive results, but at least reduced the tail risk of a serious conflict breaking out over the next 2 years. The main deliverables were a 2-month tariff extension and 2 postponements, in 2 tranches, of rare-earth export restrictions, though the exact duration was not specified; core disputes over Taiwan, China’s international standing, and chip exports still require painful, drawn-out negotiations. China nevertheless needs a détente with the U.S. economically, and the world needs China to export deflation, so “a thaw is certain.”
The real inflation risk right now is not crude oil but refined products such as diesel and a crack spread as high as about 100. Four forces are squeezing supply at once: disruptions to Middle East supply chains, damage to Russian refineries, China’s export quotas, and the possibility that the U.S. will restrict exports; even if crude rises another 10–20, consumers might actually feel better as long as refined-product prices fall 30–40.
Whether China increases refined-product exports by 1M–2M barrels per day is the key variable for global inflation and Europe’s winter risk. China’s refinery utilization is about 66%, inventories are about 1.1B barrels, and the domestic crack spread is about 20 versus more than 70 in Singapore; citing Goldman Sachs estimates, he says that if the NDRC relaxes export approvals, global crack spreads could fall from 80–90 to 40, while European diesel prices could rise more than 50% if it does not, with the extreme tail being “Europe will run dry.”
He still considers U.S. inflation transitory because the forces pushing it higher are relatively short-lived, while falling rents and the downward pressure AI is putting on employment and wages will last longer. He says about 0.8% of core PCE is still attributable to tariffs, with tariffs contributing about 0.6% month over month, and that effect will continue to fade over coming months; AI-sensitive jobs are falling by about 10,000–15,000 per month, while data-center construction is adding back only about 7,000, leading him to expect the labor required for current jobs to fall at least 25% over the next 5–10 years.
Rising expectations for a rate hike in Q4 do not amount to a long-term reversal in the logic of U.S. risk assets; the real danger is volatility in 10-year yields. Kevin Warsh’s hawkish remarks, combined with the synchronized hawkish turn by global central banks, have markets pricing 1 more hike in Q4; but with U.S. GDP growth around 6%–7% and a deficit around 6%, the country is not yet at debt contraction or austerity, and “rate hikes will pull a lot of money into the U.S.”
A near-term correction is almost unavoidable, but the medium-term bull case for AI and semiconductors is unchanged. A weak 10-year auction, September Treasury and credit-bond supply, widening IG/HY spreads, and about $32B of month-end rebalancing selling before Wednesday have jointly created a missing-buyer problem; however, valuations are not high, hedge-fund net positions and gross leverage are not crowded, and AI supply bottlenecks have not improved, so there is still a “floor” below, with the greatest volatility likely before Thursday.
Deep dive
1. The U.S.-China talks secured a thaw but left structural differences unresolved
Shanghao Jin’s read is that “the optics were executed to perfection”: Trump personally met his counterpart at the airport, China gifted giant pandas, and both sides used friendly language while playing up G20 and APEC; the main tangible deliverables were a 2-month tariff extension and 2 postponements, in 2 tranches, of rare-earth export restrictions, with the exact duration unspecified.
The talks could not resolve in 1 sitting the questions of Taiwan, whether China is a G2 power or operates in a U.S.-led world, or how many chips can be exported. He therefore expects follow-on negotiations to be “quite painful”; the leaders’ meeting may prove to be the high point for atmospherics.
China did not bring a delegation of entrepreneurs, which some on Wall Street saw as a sign of weaker willingness to cooperate. His alternative explanation is that, amid internal changes, China’s leadership is “less willing to casually put its prestige behind entrepreneurs.” Commercial discussions may not actually be on hold, since Intel’s CEO was in Shanghai at the time.
His bottom-line assessment remains constructive: China is suffering severe deflation, while the U.S. is seeing AI investment drive demand for industrial equipment, power-generation equipment, and data centers. The two economies are highly complementary, so at least over the 2 years of Trump’s term under discussion, the tail risk of a serious conflict has declined.
2. Refined products, not crude, are the core transmission channel for this inflation shock
Iran has not fully blocked the Strait of Hormuz, but it has disrupted Middle East refineries and the global oil supply chain. The chart he cited shows related exports down about 7M–8M barrels per day from prewar levels, while Chinese demand is down about 3M–5M barrels per day and the U.S. and Europe are also cutting consumption.
Crude itself is not the biggest problem because demand is falling, other regions are increasing output, and inventories are being drawn down; the real pinch is the crack spread. Singapore is already above 70, with some regions at 90–100: “right after the war started and now, the crack spread is basically 100.”
Refined products feed into PCE through jet fuel, airfares, and logistics, making their impact more visible to households than the crude price. Shanghao Jin’s summary is that if crude rises 10–20 while refined products fall 30–40, perceived inflation could still improve.
3. China’s export quotas will determine whether Europe bears the wars’ main costs
China’s teapot refiners are still losing money because domestic demand is weak. Overall refinery utilization is about 66%, but exports are constrained by NDRC approvals; the domestic crack spread is about 20 versus more than 70 in Singapore. Overseas margins are therefore very attractive and China has substantial idle capacity, but strategic considerations are keeping exports in check.
He estimates China could release an additional 1M–2M barrels per day of refined products. With inventories at about 1.1B barrels, that would be enough to sustain exports for a considerable period. Comparing this year’s exports with last year’s alone “doesn’t really make sense,” because it ignores inventories, spreads, and surplus capacity.
He cites a Goldman Sachs scenario in which a larger Chinese quota would take the crack spread from 80–90 down to 40. If China continues to restrict exports while the U.S. also stops exporting, European diesel prices could rise more than 50%.
Europe is simultaneously facing physical cuts at Russian refineries, supply disruptions in the Middle East, and tighter exports from both the U.S. and China. The 2 wars “hit Europe hardest,” affecting not only energy and the winter economy but also data-center construction and AI development; the extreme tail is “Europe will run dry.”
4. U.S. inflation is being pushed up by imported shocks, while disinflationary forces will last longer
He breaks the upside pressure into 3 components: oil and refined products, tariffs, and software and hardware costs driven higher by AI data centers. The total lift is about 1 percentage point, with tariffs contributing about 0.6 percentage points and oil about 0.2–0.4 percentage points. He separately says about 0.8% of core PCE is still attributable to tariffs, with the month-over-month effect of tariffs around 0.6% and continuing to decline. Rate hikes cannot produce diesel; they can only contain inflation expectations and second-round spillovers.
The month-over-month tariff shock was concentrated at the start of the year and in Q1 and Q2. If the current release is for August, then under his path the effect will continue to fade through September–November. Unless a low-intensity war with Iran drags on for a long time, oil could become the main remaining upside driver.
The first downside force is lagged shelter: U.S. rents have been falling for some time, and the official data will follow. The second is wages. AI-sensitive jobs are falling by about 10,000–15,000 per month, while data-center construction is adding back only about 7,000; once cumulative job displacement reaches a tipping point, it will materially suppress the wage spiral.
His long-term estimate is that over the next 5–10 years, the labor required for current jobs will fall at least 25%, and possibly more. His conclusion remains that “the U.S. actually has no core inflation”; current inflation is closer to transitory than to sustained excess demand.
5. Global central banks turn hawkish together, but rate volatility hurts markets more than hikes themselves
Europe is taking a deeper hit from diesel, the Russian energy shortfall, and the Middle East conflict. Unlike the Fed, the ECB does not have an employment mandate and is more focused on inflation. Rate-hike expectations for the BOJ, ECB, and U.S. are all higher than 3 months ago, so this repricing is not Kevin Warsh’s doing alone.
Kevin Warsh sent a forceful signal: inflation was not truly brought under control over the past 5 years, while employment remains close to full employment. The market has therefore priced 1 Q4 hike quite fully, but Shanghao Jin emphasizes that “rate hikes will pull a lot of money into the U.S.” and do not necessarily directly hurt U.S. assets.
He says the key variable to watch is long-end rate vol. The U.S. central bank owns fewer bonds relative to Europe and Japan; at the same time, U.S. GDP growth is around 6%–7% and the deficit around 6%. The fiscal process is difficult, but the country is nowhere near a point where austerity or cuts to debt-related spending are necessary.
6. September supply-demand imbalance is driving a correction; the AI bull case still has positioning and valuation support
Last week’s 10-year Treasury auction was weak, while the market was simultaneously pricing higher for longer and high diesel prices. Even if pension funds recognize that long bonds are beginning to offer value, they are inclined to wait for better entry levels, leaving the market with “missing buyers.”
September Treasury and credit-bond supply is heavy, with IG and high-yield spreads widening in tandem and the latter moving more sharply. This does not necessarily signal panic; it is more a timing mismatch between issuance supply and demand. Oracle project delays, related debt tied to Apollo, and possible redemptions have added to localized pressure, while October supply is expected to be lighter.
Liquidity as reflected by GSFCI has fallen to one of its weakest levels in recent months, close to July. Adding about $32B of month-end rebalancing selling before Wednesday, he concludes that “a short-term correction is unavoidable,” with volatility particularly prone to intensify before Thursday.
The medium-term supports remain in place. Manus looks “like a large Doubao”; it may not yet have proved its profitability or capabilities, but it could broaden AI penetration. Industry supply bottlenecks have not improved, valuations are not particularly high, and hedge-fund gross and net leverage are not crowded. The market can keep pushing lower, but “there is a floor below”; the medium- and long-term view on AI, semiconductors, and the broader market is unchanged.