Soybeans: Harvest Lags as China Demand Weakens

Only 25% of the US soybean crop had been harvested by 4 October, against a 33% five-year average. Yet the more consequential signal may be emerging in China, where state-linked companies have bought US beans while private crushers face negative margins, record inventories and little commercial incentive to follow.

TL;DR

US soybean harvest reached 25% by 4 October, up eight percentage points in a week but still eight points behind USDA's 33% five-year average.
Crop development is much closer to normal: 85% was dropping leaves against an 87% average, separating the harvest delay from crop maturity.
USDA currently forecasts a 52.8 bu/acre yield and a 4.535bn-bushel crop from approximately 85.9m harvested acres, with exports at 1.69bn bushels and crush at 2.78bn.
US soybean stocks on 1 September were 3% below a year earlier, providing a tighter starting point as the new crop enters the system.
Chinese state-linked companies have made substantial US purchases, but private crushers remain reluctant amid tariffs, negative margins and high inventories.
Soybean inventories at 111 Chinese crushing plants reached 7.96m tonnes in late September, while only 37.3% of beans offered in the latest Sinograin auction were sold.
Chinese buyers booked around 50 soybean cargoes in the first three weeks of September, the fewest for that period in four years, with state firms accounting for roughly 30 US cargoes.
November CBOT soybeans settled at $12.80¾ on 5 October, up 2½ cents, after a 30¾-cent fall on 28 September when soybeans were excluded from China's latest tariff relief.
The next supply catalyst is USDA's Crop Production report at 12:00 ET on Friday 9 October, when the 52.8 bu/acre yield estimate meets more harvested evidence.

A Harvest Gap Has Opened

The US soybean harvest accelerated substantially in the week ended 4 October, with national progress increasing from 17% to 25% — an eight-percentage-point move in seven days. That still left harvest well behind its seasonal benchmark, since USDA's published 2021–25 average for the week is 33%.

The distinction between crop development and completed harvest is important. Soybeans dropping leaves advanced from 75% to 85%, only two percentage points behind the 87% five-year average. The crop itself is therefore considerably closer to its normal development schedule than the harvested acreage suggests, which points to field access rather than maturity as the binding constraint.

The national figure also conceals large state differences. Illinois was 31% harvested, Indiana 27%, Minnesota 25%, North Dakota 34% and South Dakota 17%. Iowa, however, was only 5% harvested.

The delay has already mattered in the physical market. Reuters reported on 25 September that persistent rain in the western Midwest had interrupted soybean harvesting and left processors competing for immediate supplies. Some plants reduced operations, while cash bids in parts of the region reached as much as $1 a bushel above November futures as processors sought either early new-crop beans or remaining old-crop stocks.

That is evidence of a real near-term economic effect, not of a smaller national crop. With leaf drop close to normal, favourable conditions could close part of the deficit quickly during October. The more important supply question is whether field results begin to challenge USDA's yield estimate.

USDA Still Has a 4.535bn-Bushel Crop

The official supply picture remains substantial. USDA's September estimates put national soybean yield at 52.8 bushels per acre and harvested area at approximately 85.9m acres, implying production of about 4.535bn bushels.

That matters when interpreting weekly Crop Progress. An eight-point lag in harvested acreage changes when supply reaches elevators, exporters and processors. It does not by itself change the number of bushels ultimately produced.

There is nevertheless less old-crop cushion than a year ago. USDA's 30 September Grain Stocks report showed soybean inventories on 1 September down 3% year on year, which helps explain why local harvest interruptions have produced noticeable nearby basis effects without any evidence yet of a national production shortfall.

The next production estimate is therefore more important than the harvest percentage alone. USDA NASS has scheduled Crop Production for 12:00 ET on Friday 9 October. If harvested evidence moves USDA materially away from 52.8 bu/acre, the supply calculation changes. If yield survives broadly intact, the present harvest deficit looks increasingly like a timing problem.

China Is Buying — But Not All Chinese Demand Is the Same

The more significant development is on the other side of the balance sheet. China has not stopped buying US soybeans; the distinction is between state-directed purchases and the economics facing private commercial crushers.

Reuters reported on 30 September, citing three Asia-based agricultural traders, that Chinese state-run companies had bought around 13.7m tonnes of US soybeans following the trade agreement reached in May, while private Chinese crushers had taken only South American shipments. That 13.7m-tonne figure is an industry estimate reported by Reuters rather than an independently verified USDA cumulative sales total, and given its scale that distinction matters.

The underlying state-versus-private split is nonetheless supported by several independent indicators. US soybeans still face an additional 10% Chinese tariff, while Brazilian cargoes have been quoted at roughly comparable pre-tariff prices with typically higher oil content, adding to their attraction for crushers.

For private processors, the immediate economics remain difficult. November-shipment soybeans from the US Pacific Northwest and Gulf were generating estimated crushing losses of 120–200 yuan a tonne, according to Shanghai JC Intelligence figures reported by Reuters, with Brazilian beans also producing negative margins of around 120 yuan.

High inventories further reduce the urgency to buy. Soybean stocks at 111 Chinese crushing plants reached 7.96m tonnes in the week of 25 September, according to Mysteel data reported by Reuters — the highest level in at least 15 years. Only 37.3% of the 514,000 tonnes offered in Sinograin's latest imported-soybean auction was sold.

Cargo bookings tell the same story from a different angle. Chinese buyers booked around 50 soybean cargoes during the first three weeks of September, the fewest in four years, according to Marex data reported by Reuters. State-run COFCO and Sinograin accounted for roughly 30 US cargoes; private buyers sourced the remainder from Brazil and Argentina.

That is the central distinction in the current soybean market. China is buying US soybeans. Its private crushing industry largely is not.

A 4.535bn-Bushel Crop Still Needs a Home

The distinction matters because the US balance sheet assumes substantial demand on both sides. USDA forecasts 1.69bn bushels of exports and 2.78bn bushels of domestic crush for 2026/27, with ending stocks at 310m bushels.

Domestic processing provides an important base. US crush capacity has expanded and soybean oil demand gives processors a structural source of consumption independent of Chinese import behaviour.

Exports are the more sensitive variable. China is too important to the international soybean trade for the state-versus-private distinction to be dismissed as an institutional detail. If state-directed buying continues at sufficient scale, it can support the 1.69bn-bushel export assumption even while private crushing economics remain poor. If that buying slows, the underlying commercial picture becomes considerably more exposed, because high Chinese inventories, negative crush margins, weak auction participation and low cargo bookings all indicate that private processors are not currently signalling scarcity.

That matters more if Friday's estimate leaves US production close to 4.535bn bushels. A smaller crop would reduce the volume requiring demand. A crop near the current estimate places greater pressure on the export side of the balance sheet to perform.

Price Has Weakened, But the Sequence Matters

November CBOT soybeans settled at $12.80¾ a bushel on Monday 5 October, up 2½ cents on the session and snapping a three-session slide, with firmer soyoil and a stronger Brazilian real offsetting harvest pressure.

The more important price move came a week earlier. On Monday 28 September, November soybeans fell 30¾ cents to $12.88¼ as traders reacted to soybeans being excluded from China's latest agricultural tariff reductions following the US-China talks. The contract settled at $12.84 on 1 October and $12.78¼ on 2 October before Monday's modest recovery.

That sequence is more informative than any single weekly change. Soybeans repriced sharply on the China tariff news and have since traded in a narrow range around $12.80 rather than continuing to fall.

The positioning data provide additional context. In CFTC data for 29 September, managed money held 280,875 soybean longs against 34,317 shorts, leaving the sector net long 246,558 contracts — down 18,601 from +265,159 the previous week. The composition matters: gross longs fell by 19,867 contracts while shorts declined by 1,266, so the reduction was overwhelmingly long liquidation rather than aggressive new short selling.

That fits the broader fundamental picture. The market has become less willing to maintain an exceptionally large bullish position as uncertainty over Chinese commercial demand has increased, but investors were not yet building a correspondingly large bearish one. US supply has not provided a sufficiently clear bearish signal for that: harvest is behind schedule, nearby processors have experienced temporary tightness, and Friday's national yield estimate remains unresolved.

Friday Tests the Supply Side

The temptation is to frame the soybean story around 25% harvested against 33% normally. That is useful information, but it is not yet the central conclusion. A delayed harvest affects timing; yield determines supply.

USDA currently has production at approximately 4.535bn bushels on a 52.8 bu/acre yield, and Friday's Crop Production report provides the next major test of that estimate. If USDA cuts yield materially, the eight-point harvest deficit acquires greater significance because delayed fieldwork will have coincided with evidence of a smaller crop. If USDA leaves yield broadly intact, the focus shifts towards demand — a delayed but ultimately large crop arriving into a market where Chinese private crushers have poor margins, unusually high inventories and little apparent urgency to secure US supply.

The state-versus-private distinction therefore becomes critical. China can continue buying substantial quantities through state-directed channels even when commercial processors would not choose those cargoes independently. But a market supported by state purchasing behaves differently from one in which crushers themselves are competing aggressively for supply.

Outlook

The immediate supply variable is yield. The 52.8 bu/acre estimate is the benchmark, and Friday's Crop Production report will incorporate more information from harvested fields.

The second is harvest catch-up. With leaf drop only two points behind normal, favourable weather could close part of the eight-point deficit quickly. Persistent delays would keep nearby physical supply tighter for longer.

The third, and potentially most durable, is Chinese commercial demand. State buying has placed meaningful US tonnage on the books, yet inventories of 7.96m tonnes at surveyed crushing plants, negative margins and weak auction participation indicate private processors do not face the scarcity that normally generates aggressive discretionary imports.

What would change the view: A material cut to yield would reduce the volume requiring demand. A yield that holds while harvest catches up would make the question whether state-linked Chinese purchases can continue compensating for weak private-sector economics.

Key Risks

Harvest progress is not production. The eight-point deficit measures the proportion of acreage harvested, not eventual crop volume or yield.

The five-year comparison is unusual this year. USDA cancelled the corresponding Crop Progress reports during the October–November 2025 federal funding lapse, so its published 2021–25 averages use observed estimates for 2021–24 and imputed estimates for 2025.

The 13.7m-tonne state-purchase estimate is not an official statistic. Reuters attributed it to three Asia-based agricultural traders, and it should be treated as a reported industry estimate rather than a confirmed government total.

Chinese state and private demand should not be conflated. State-linked purchases can coexist with unattractive commercial crushing economics for private processors.

Chinese inventory and margin figures are third-party estimates. The 7.96m-tonne inventory figure, the Sinograin auction result, the cargo-booking count and the reported crushing margins originate from commercial industry sources cited by Reuters.

CFTC positioning predates the latest Crop Progress report. Positions are dated 29 September while Crop Progress covers the week through 4 October. The datasets provide complementary context but should not be presented as causally linked.

Friday can materially change the supply analysis. Production and yield estimates remain subject to revision as harvested evidence increases.

Intelligence Monitoring Points

USDA Crop Production, 9 October: soybean yield against 52.8 bu/acre and production against approximately 4.535bn bushels.
US harvest progress: whether the 25% against 33% five-year-average deficit begins closing.
Chinese purchasing: whether state-linked US buying continues and whether private crushers begin participating.
Chinese crush margins: whether US and South American import economics move back towards profitability.
Chinese soybean inventories: whether the 7.96m-tonne late-September stock begins drawing down.
CBOT November soybeans: direction from the $12.80¾ settlement of 5 October.
CFTC managed money: whether further reductions in the +246,558 net long remain driven by long liquidation or develop into meaningful new short formation.

FAQ

Why is the US soybean harvest behind normal? USDA reported 25% harvested by 4 October against a 33% five-year average. Weather has delayed fieldwork in parts of the Midwest, while crop development is much closer to normal at 85% dropping leaves against an 87% average.

Does a slow harvest mean the US crop will be smaller? Not necessarily. Harvest progress measures acreage harvested, not yield. USDA estimates national yield at 52.8 bu/acre and production at approximately 4.535bn bushels, and Friday's report provides the next direct test.

Is China still buying US soybeans? Yes, but the identity of the buyer matters. Reuters reported that state companies had bought around 13.7m tonnes following the May agreement, citing three traders, while private crushers had taken only South American shipments.

Why are Chinese soybean inventories important? High inventories reduce processors' urgency to secure additional imports. Stocks at 111 Chinese crushing plants reached 7.96m tonnes in late September, the highest in at least 15 years according to Mysteel data reported by Reuters.

Where are soybean prices? November CBOT soybeans settled at $12.80¾ on 5 October, having fallen 30¾ cents on 28 September when soybeans were excluded from China's latest tariff relief.

What is the next major catalyst? USDA NASS publishes Crop Production at 12:00 ET on Friday 9 October. The principal soybean number is national yield relative to the current 52.8 bu/acre estimate.

Data and source note: US crop-development and harvest figures are from USDA NASS Crop Progress, released 5 October 2026 and covering the week ended 4 October. USDA's published 2021–25 averages require an additional caveat this season, because the corresponding 2025 weekly reports were cancelled during the federal funding lapse and USDA uses imputed 2025 estimates within the five-year benchmark. Production, yield, export and crush figures refer to the latest USDA estimates available before publication; the 4.535bn-bushel production figure is the result of the current yield and harvested-area estimates. The 1 September stocks figure is from USDA's Grain Stocks report of 30 September 2026. CBOT settlement prices are from USDA Agricultural Marketing Service daily grain reports: November soybeans settled at 1280.75 cents on 5 October, 1278.25 on 2 October and 1284.00 on 1 October. Chinese state purchasing, crusher inventories, crushing margins, Sinograin auction participation and cargo-booking figures are industry estimates reported by Reuters on 30 September rather than official Chinese statistics; the 13.7m-tonne state-purchase figure is attributed by Reuters to three Asia-based agricultural traders. The western-Midwest basis discussion is from Reuters reporting of 25 September 2026. Managed-money positioning refers to CFTC positions dated 29 September 2026 and predates the Crop Progress report covering the week ended 4 October.

Sources

This document is published by Bloodstone Research for informational and institutional research purposes only. It does not constitute investment advice, an investment recommendation, an offer or solicitation to buy or sell any financial instrument, commodity or security, or a forecast of future performance. Market conditions and data may change without notice. Readers should conduct their own analysis and, where appropriate, seek independent professional advice before making investment decisions. For institutional enquiries contact research@bloodstonecapital.co.uk.

Soybeans analysis — 2026-10-06