Coffee Futures Market Weekly Update
Two things defined the week: a front-month roll and dry Brazilian weather. With September arabica (KCU26) moving into delivery, the reference contract shifted to December (KCZ26), which rallied to a 6.5-month high mid-week before falling -2.02% (-6.65c) on Friday to settle at 322.65c. Against the prior Friday's December close of 314.30c that is a weekly gain of +2.66% (+8.35c). The Friday slide was triggered by weekly rainfall in Minas Gerais of just 0.6mm — 11% of the historical average, which points to a faster harvest. Pulling the other way, ICE certified arabica stocks fell again to a 2.75-year low of 227,992 bags.
※ A front-month roll occurred this week. September (KCU26) entered the delivery process and the reference contract switched to December (KCZ26), so the prior report's 337.80c September print is not directly comparable with this week's 322.65c. The weekly change is computed December-versus-December (314.30c on 8/14 to 322.65c on 8/21 = +8.35c / +2.66%). Friday's 322.65c settle and robusta November's $3,618 are confirmed Barchart settlement prints; Thursday's 329.30c is derived from Barchart's Friday change (-6.65c) and cross-checked against the Trading Economics benchmark quote of 329.30. The 314.30c baseline is confirmed by StoneX's weekly report ("Dec/26 arabica closed the week at US¢314.3, down 0.5%"). The week's lowest close (Monday) could not be confirmed by two independent sources and is therefore not quoted as a figure — StoneX reported only that December traded above 318c on Monday. Barchart's Friday commentary cites ICE robusta stocks as both 4,732 and 4,722 lots, so we state "about 4,730 lots." USD/BRL prints ranged 5.176–5.194 across sources and is stated as approximately 5.18.
The first event of the week was not a price move but a change in contract structure. September arabica went into delivery and the market's reference shifted to December. StoneX reported that on August 20 alone 4,074 September lots changed hands along with 458 EFP/EFS trades posted off exchange, and the September–December (U/Z) spread went out at 31.80c. In other words, the spot month held a backwardation of more than 30 cents over December right into expiry — a direct read on how tight nearby physical supply remains.
The price path was firm for four days, then sharply lower on Friday. December opened the week above 318c and set a 6.5-month high on Tuesday (per Barchart's Thursday commentary). Safras & Mercado data showing the Brazilian harvest still behind both last year and the five-year average, together with collapsing certified stocks, underpinned the advance. Thursday's 329.30c was the week's highest close.
Friday then erased much of that gain with a -2.02% (-6.65c) drop. The trigger was rainfall data published by Somar Meteorologia on August 17: weekly rain in Minas Gerais through August 16 came in at 0.6mm, just 11% of the historical average, down sharply from the prior week's 5.8mm (92% of average). The market read this as a delayed harvest about to accelerate — and Cooxupé's progress figure duly jumped 7 points in a week to 81.1% as of August 14.
The week's central dynamic: the same "dry Brazilian weather" could be read in two opposite directions. For the harvest, dryness means faster near-term supply (bearish); for the September–October flowering window, dryness means damage to the next crop (bullish). On Friday the market chose the first reading. Yet StoneX's field reporting notes that Matas de Minas has already received enough rain to trigger new flowering, with further precipitation forecast from August 22. Friday's bearish logic therefore faces an immediate re-test next week.
The most information-dense number of the week was not a settlement price but the September–December spread of 31.80c. A nearby contract holding a 30-cent-plus premium over the next month right into expiry says the market's judgement that immediately deliverable, exchange-grade coffee is scarce never wavered. The fact that the spread swung across a wide 29.60–35.40c intraday range shows how violently positions around that premium were unwound.
The practical consequence is a level shift in the headline price. The prior report's 337.80c (September) and this week's 322.65c (December) are not measured with the same ruler. On a December basis the market actually rose 2.66% on the week. Any internal reporting or hedge valuation that spans this period must apply a continuous-contract adjustment.
Implication: saying "coffee fell from 337c to 322c" across this roll is effectively wrong. What happened is that the benchmark moved 30 cents lower; on a like-for-like basis prices rose. Contract renewals, budget builds and margin calculations must strip out this roughly 15-cent optical illusion.
ICE certified arabica stocks declined again every day this week, reaching 227,992 bags. The trajectory runs 251,417 bags on August 7, 236,430 on August 13, 231,445 on August 17 and 227,992 on August 21 — roughly -9.3% over two weeks, and about -70% versus the same point last year (roughly 754,516 bags). StoneX attributed the drawdown not only to withdrawals but to Brazilian lots being rejected during classification alongside an absence of significant new certifications — that is, the shortage reflects supply quality rather than demand strength.
Robusta moved in exactly the opposite direction, with stocks at an 8.75-month high of about 4,730 lots on expectations of Vietnam's incoming crop and continued Asian supply pressure. The two inventory cycles are now running fully counter to each other.
Dry weather has finally begun to unwind the harvest delay. Cooxupé's progress rose from 74.6% on August 7 to 81.1% on August 14, a 6.5-point weekly gain that narrowed the year-on-year gap from 5.8 points to 5.0 points. Nationally, Safras had the harvest at 90% on August 12, seven points behind last year.
At the same time the market's focus is rotating from harvest to flowering. StoneX's field team reports that conilon/robusta areas in Espírito Santo, Bahia and Rondônia are already flowering, with more than half of expected potential blooms emitted. On the arabica side, Matas de Minas received weekend rain sufficient to trigger new flowering, with more rain forecast from August 22. Cerrado is forecast to see moderate rain from August 25, while South Minas currently shows few signs of meaningful precipitation.
Implication: Friday's move priced a single, linear story — dry equals faster harvest equals bearish. But the same dryness can be reinterpreted as flowering-failure risk within three to four weeks. The weak rainfall signal over South Minas, Brazil's single largest arabica region, is the variable most likely to become a bullish catalyst over the next two to three weeks. Rainfall should be tracked region by region, not as a national aggregate.
The CFTC report released August 21 (positions as of August 18) put Coffee C non-commercial net length at 30,365 contracts, up 3,186 from 27,179 the prior week. The composition tells a different story from the headline: longs fell 662 contracts while shorts fell 3,848. The increase in net length is therefore capitulation by existing shorts, not fresh bullish conviction.
More telling still, open interest dropped 10,540 contracts (-6.1%) to 163,299. Expiry roll mechanics explain much of that, but the combination of rising price, rising net length and falling open interest is the textbook signature of a low-quality, short-covering rally. Managed money net length actually slipped to 31,612 contracts (-821), indicating speculative capital stayed largely on the sidelines.
The most recent confirmed trade statistics, Cecafé's July data, again illustrate the "record crop, tight arabica spot" paradox. Total July exports were 3.03 million bags (+10%), but arabica fell to 1.82 million bags (-8.7%) while robusta surged to 851,235 bags (+84.4%). Year-to-date shipments stand at 18.4 million bags (-7.7%), with arabica at 15.0 million (-16.6%), and revenue of USD 903.9 million (-13.2%) — more volume, less money.
The production consensus remains firmly bearish. The USDA projects 2026/27 world output at a record 189.7 million bags (+6.0%), Brazil at 71.9 million bags (+14%) and world ending stocks up 1.9 million to 26.3 million bags. CONAB's second survey puts Brazil at 66.7 million bags. On robusta, Vietnam's January–July exports rose 21.1% to 1.31 MMT, with July alone at 147,600 tonnes (USD 639.5 million) and a seven-month average export price of USD 4,537/t (-19.9%) — the classic oversupply pattern of rising volume against collapsing unit value.
Vietnam's next crop is more ambiguous. StoneX relays local estimates of a roughly 20% decline in parts of the Central Highlands due to dry weather and irrigation difficulties, alongside pest pressure from excessive rain elsewhere. These are single-source local estimates and have not been independently cross-confirmed.
USD/BRL, which had weakened from 5.09 to 5.22 the prior week, retraced to approximately 5.18 this week (source prints ranged 5.176–5.194). A firmer real means Brazilian producers receive fewer reais per exported bag and have less incentive to sell. As StoneX notes, the exchange rate is the key short-term driver of Brazilian commercialization pace, and the currency's continued strength versus the 5.50 level seen early in the year helps explain persistent producer selling resistance.
Brazil's domestic physical indicators (Cepea, week to August 14) diverged: arabica at R$1,794.15/bag (+3.6%) against robusta at R$1,049.17/bag (-3.7%). Producer resistance and limited immediate availability of higher-quality lots lifted the arabica indicator, while robusta tracked London lower.
1. Strip out the roll illusion: the headline appears to have fallen 15 cents from 337.80c to 322.65c, but that is a benchmark shift, not a price move — on a like-for-like basis the market rose 2.66%. Any August–September budget, hedge valuation or contract negotiation citing this period must explicitly state the continuous-contract adjustment.
2. The quality of the rally was poor: the +3,186 increase in net length was really -3,848 in shorts, with longs actually declining, and open interest fell 10,540. A rally with no new money behind it is vulnerable to reversal — Friday's -2.02% proved the point. Range trading beats trend-following in this configuration.
3. Arabica–robusta decoupling has reached an extreme: arabica stocks at a 2.75-year low against robusta stocks at an 8.75-month high is an unusual combination. For roasters with blend flexibility, the economics of substituting robusta are the most favourable in roughly a year; conversely, washed-arabica specialty procurement should budget for wider premiums.
4. The next three weeks hinge on South Minas rainfall: Matas de Minas has already met flowering conditions and Cerrado rain is forecast from August 25 — but South Minas, Brazil's single largest arabica region, shows a weak precipitation signal. Layered on top is the September–October rain-delay risk from one of the strongest El Niño events in 75 years. Disaggregated, region-level rainfall tracking is the highest-value risk management action this quarter.
5. The cause of the inventory drawdown has changed: StoneX attributes the decline to Brazilian lots failing classification and an absence of new certifications. That makes the shortage a structural supply-quality signal rather than a bullish demand signal. Even as a record crop arrives, the volume that clears exchange grade may stay limited — in which case the coexistence of a record crop and low certified stocks persists longer than the consensus assumes.
On a December basis the market is in a wide range with a mild upward bias. Support runs 321.8c (10-day moving average), 318c (early-week lows) and 314c (prior Friday); resistance runs 329.3c (week's high close), 335c and 345c. Friday's 322.65c settle sits just above the 10-day moving average, making the defence of that line next week's first test.
Technical view: the weekly closing range was roughly 318–329c. Friday's -2.02% surrendered about half the week's advance, but the weekly change remained positive (+2.66%) and the close held above the 10-day moving average (321.83). StoneX observes that arabica has found meaningful support at the 20-period moving average over the past two weeks. A break of 321c opens a gap toward 314c; a recovery of 329c puts 335c into play.
The medium-term anchor remains record supply (189.7 million bags globally, 71.9 million in Brazil), and this week's harvest acceleration strengthened that downward pressure. Three buffers remain: (1) certified stocks at a 2.75-year low of 227,992 bags, with a structural cause in classification failures; (2) Brazilian arabica exports down 8.7% year-on-year in July, showing the record crop has not yet converted into shipments; and (3) the September–October flowering risk from one of the strongest El Niño events in 75 years. We retain the base case of structural downward pressure punctuated by event-driven spikes, while noting that near-term downside risk widened this week on the harvest acceleration.
This was a week in which how you read the numbers became the story. On the headline the market appears to have collapsed from 337.80c to 322.65c, but that is simply the consequence of September's expiry shifting the reference contract. Measured December-to-December, the market rose from 314.30c to 322.65c, up 2.66%. And the 31.80c backwardation held into expiry testifies that nearby tightness eased not at all.
The quality of that advance, however, was poor. Non-commercial net length rose 3,186 contracts, but the mechanism was short-covering (-3,848) while longs actually declined, and open interest fell 10,540 (-6.1%). A rally without new capital is fragile — and Friday's -2.02% collapse on a single rainfall datapoint (0.6mm, 11% of average) demonstrated exactly that.
Trading view in brief: December in a 314–335c range, with 321.8c (the 10-day average) the first line to defend. Key monitors: (1) flowering rainfall in South Minas and Cerrado; (2) the 220,000-bag line for ICE certified stocks; (3) Cooxupé harvest pace; (4) whether open interest rebounds and new longs appear; (5) whether robusta stocks keep setting highs; (6) the completion date for Buenaventura's normalization.
For roasters and importers there are two practical conclusions. First, strip the roughly 15-cent roll illusion out of your procurement benchmarks and re-base them on a continuous contract. Second, with the arabica–robusta inventory divergence at an extreme (2.75-year low against an 8.75-month high), this is the most economically compelling window in roughly a year to revisit blend ratios. On direction, a clear bearish driver (harvest acceleration) and a potential bullish catalyst (absent South Minas flowering rain) will collide head-on within three weeks. On a risk-adjusted basis, scaled hedging into the top of the range is preferable to a directional bet.