The Early Signal Fired. The Peak Is Still Ahead.

In mid-July we named the state, the week, and the driver — and it fired on schedule. The reading that matters most is still ahead, and still a forecast.

By João Pedro Rodrigues Morciani · Senior Analyst, Helios AI

The Tell We Named in July Fired on Schedule.

In mid-July, the reading we published to our own desk named Nebraska as the early tell. On model runs through Jul 13, it said the state's Helios AI climate-risk score — the share of production under climate stress, 0 to 100 — would fire in the week of Jul 27, roughly three weeks ahead of the Belt-wide peak, with heat as the driver.

It fired on Jul 30. Nebraska corn hit 52.5 and soybeans 54.9 (observed days, pulled Aug 12) — inside the named week, heat-driven as called, harder than the roughly 48 and 47 pencilled in — the panel's highest realized readings for either crop. The US Department of Agriculture (USDA) registered the same hit: Nebraska corn 54% good-to-excellent, its second-lowest early-August rating since 2012 (Reuters, Aug 10).

One caveat we owe you on the soybean number. The spike was real — heat during pod set, the stretch Helios's scoring treats as the soybean season's most yield-sensitive. The slide since, 52.1 on Aug 1 to 13.3 by Aug 12 (pulled Aug 12), reflects two things: weather genuinely moderated, and the crop moved out of that maximum-sensitivity window into grain fill. The reading has not said the stress reversed — pod fill is where August makes or breaks beans, precisely the window the late-August forecast now flags.

The Peak Is Still Ahead. And Still a Forecast.

The same mid-July read called a national risk peak near 41 for late August — and nothing at 41 has been realized. As of our Aug 12 pull, the model shows roughly 40.5 for the week of Aug 17 — within a point of the July call — with the absolute peak now centred on Aug 22–27 (43.7 corn, 41.9 soybeans, both Aug 27). The magnitude call is intact; the clock has slipped about a week. The highest observed national reading is corn's 35.2 on Jul 30 (pulled Aug 12); everything above it is a forecast.

The Leg We Would Have Led With Did Not Verify.

Warm nights at pollination were the July thesis. They did not happen. The mid-July read forecast a +3.2°F overnight-low anomaly across the corn pollination window (Jul 10–Aug 10), with 44% of nights at or above 70°F. Observed, window fully elapsed (pulled Aug 12): +0.58°F, and one night of 32 — 3%. That is a miss, and we print it.

The yield claim was conditional, and it behaved like one. The July read said: if the heat verifies, 8 to 10 bushels an acre off the 183 trend. The heat did not verify — and USDA's first survey-based corn yield, in the Aug 12 World Agricultural Supply and Demand Estimates (WASDE) report, was 180.7 bushels an acre — 2.3 below trend (the pre-report consensus comparison is in the FAQ). A forecast that scales down when its condition fails is honest conditionality.

The Thesis Was Heat. The Model Now Points to Cold.

The August risk is not the July risk wearing a later date. Soybeans went back on Watch on Aug 9 on record Reproductive Development cold — 100th percentile of the 8-season record across Indiana, Ohio and Minnesota — with Helios's own caveat that only about a third of the phase is observed (pulled Aug 10).

The forecast peaks ahead have changed colour too. The late-August peaks in Indiana, Ohio and Illinois are too-cold-dominated, not heat, and Ohio soybeans' 55.5 forecast for the week of Aug 21 is too-wet-led (pulled Aug 12). A reader tracking "heat dome" headlines into late August will be watching the wrong stress.

Cold in August works through the clock, and it splits the two crops. Crops develop on accumulated warmth, so a cool stretch slows grain and pod fill and pushes maturity toward frost. Soybeans are the sensitive crop: nights below about 60°F during pod set and seed fill cut pod set and seed size by impairing overnight starch export (Ohio State University; South Dakota State University Extension). Corn is more forgiving — moderately cool nights have historically helped grain fill by trimming respiration losses — while heat shortens the fill window and drought aborts kernels and pods. The same cold reading is a bigger soybean story than a corn story.

None of this is a crisis call. The national corn rating is Watch, the reduced WASDE yield still leaves the second-largest US corn harvest on record, and Illinois — the worst state rating — is 15% of national production: divergence, not disaster.

Wheat gets one paragraph, because its story is closed. Take Action since Jun 21 (re-confirmed Aug 12); Kansas heat and drought at the 100th percentile of the 8-season vegetative-growth record; Oklahoma's worst drought in 20 years capped by an April freeze; the crop essentially harvested; production confirmed down at 1,531 million bushels in the Aug 12 WASDE. The damage is booked — and wheat has no South American season to hand off to.

The Same Discipline Now Moves South.

Write the read down before the season. Score it after. The same clock now starts for the hemisphere carrying more of the world's soybeans than this one: Brazil alone is projected at roughly 41.5% of world soybeans and 10.7% of world corn for 2026/27 (Helios AI supply-yield baseline, pulled Aug 12) — a trend-based projection for a crop not yet planted.

South America enters that window off a record year — its own lesson in reading a risk score. CONAB's June survey put Brazil's 2025/26 soybeans at a record 180.3 million tonnes; Rosario's exchange put Argentine corn at a record roughly 62 million tonnes, soybeans 51.5 million, on exceptional in-season rainfall. Helios's ratings flagged regional stress episodes along the way — southern Brazil's corn states, Argentina's December–January dry spell — that cost yield locally without denting the records: a climate-stress rating is an early-warning instrument, not a production forecast. Our own early reads on 2026/27 are, so far, unremarkable — at or slightly below each pair's historical average (pulled Aug 13) — and they start filling in as the series flip in-season between Sep 1 and about Nov 1. The reason to watch those dates is the incoming El Niño: the historical pattern in outside climatology leans wet over the Pampas and southern Brazil, drier over Mato Grosso, and the National Oceanic and Atmospheric Administration's Climate Prediction Center (NOAA CPC) gives 81% odds of a very strong October–December peak (Jul 9 discussion), on the planting window.

The Bottom Line.

A month ago we wrote down a specific, dated, checkable read — and August is grading it in public. The tell fired; the miss printed; the peak is still ahead, still a forecast, now cold-and-wet-driven. The same discipline starts this month for South America.

The market, so far, is trading the WASDE, not the weather peak. November soybeans settled Aug 11 at their lowest since early July (CBOT Nov '26, Barchart), and DTN read the Aug 12 report as bullish for corn, neutral for soybeans. [INSERT: post-report close, Aug 12 session] The early signal fired; the decisive week has not.

The forecast peak lands days after our webinar

On August 18th, before the model's Aug 22–27 window, we walk the live state-level read for corn and soybeans, take questions, and preview the South American 2026/27 setup as planting approaches.

Frequently Asked Questions

What is the Helios AI climate-risk score used throughout this piece?

It measures the share of a crop's production at climate risk on a 0-to-100 scale, weighted by production and calibrated to the crop's growth phase, and it decomposes every reading into four stress types: too hot, too cold, too wet, too dry. Each daily point is flagged as observed or forecast, which is why we distinguish realized readings (Nebraska's 52.5 on Jul 30) from forecast ones (the Aug 22–27 peak). Every figure in this piece carries its pull date, because recent days are revised as new data lands.

Has the late-August risk peak already happened?

No. The highest realized national reading so far is corn's 35.2 on Jul 30 (pulled Aug 12). The 43.7 corn and 41.9 soybean readings centred on Aug 22–27 are the model's current forecast (pulled Aug 12) — a scenario that has held through a month of updates, but a forecast all the same.

Why did Nebraska's soybean score fall from 54.9 to 13.3 in under two weeks?

Two things at once. Observed weather genuinely moderated into early August — and on Aug 1 Nebraska soybeans moved out of pod set, the scoring window Helios AI treats as the soybean season's most yield-sensitive, into grain fill, taking the reading from 52.1 (Aug 1) to 13.3 (Aug 12), both pulled Aug 12. What the score has not said is that the stress reversed: pod fill is where August weather makes or breaks the soybean crop, and it is the window the model's late-August forecast now flags.

Why did soybeans go back on Watch on Aug 9 if the summer's story was heat?

Because the driver changed. The Aug 9 downgrade was driven by record cold in the Reproductive Development phase — at the 100th percentile of the 8-season record for that phase across Indiana, Ohio and Minnesota — not by heat. Helios's own assessment carries the caveat that only about a third of the phase has been observed so far (pulled Aug 10), so the reading can still firm or fade as more days land.

How does the August WASDE compare with the mid-July read?

The July read was explicitly conditional: if the pollination heat verified, 8 to 10 bushels an acre off the 183 trend. The heat did not verify (+0.58°F observed vs. +3.2°F forecast), and USDA's first survey-based yield came in at 180.7 bushels an acre — 2.3 below trend, and below the pre-report analyst consensus of 182.4 (Reuters). Direction right, magnitude scaled down with the failed condition, which is how a conditional forecast is supposed to behave.

How does this reading connect to last week's basis finding?

Last week we published the study: Helios AI climate risk leads elevator basis at 4-to-12-week horizons — wheat +1.2¢/bu per standard deviation of risk at 2 weeks (p=0.007) rising to +2.6¢ at 12 weeks (p<0.001), corn +2.38¢ at 12 weeks (p=0.001), soybeans +1.62¢ (p=0.024) — magnitudes modest against the 22–39¢ standard deviation of such moves, with mean reversion dominant: a qualifier, not a standalone signal. Its standing verdict — "Helios AI climate risk helps forecast where basis converges to; it does not beat the elevator in the week of the shock, because the elevator prices local weather almost immediately" — is the reason this week's piece exists. This is the live reading that relationship consumes: where the risk sits now, crop by crop, ahead of the forecast peak.