Two and a half months of lower highs, then a second low that held above the first — and on Friday the oscillator turned.

The claim: a two-and-a-half-month downtrend in the chip designers has produced the three things a downtrend needs to produce before it ends — a higher low, an oscillator turn and a broken line — and all three are now on the chart.

Start by being precise about what this is, because the loose word does real damage here. This is not a consolidation in the sense of a sideways pause. From the 22 June top near 68 the group has made lower highs and lower lows: the mid-August advance to about 60.5 died under the descending line and gave the whole move back. That is a downtrend, and it has run for two and a half months. Every index page in our tracker shows the drawdown; the drawdown is not the information. The information is what has to change for a downtrend to stop being one, and this week two of those things changed at once.

Start with what the consolidation actually is. Over three months this fund is down 17.4%. Over twelve it is up 58.0%, and over six up 45.1%. It sits 19.1% under a 52-week high of 67.86 and 56.9% above a 52-week low of 35.01. It is below its 20-day average near 56.55 and its 50-day near 57.44, and 14.7% above its 200-day near 47.88. That is a specific shape: a group that doubled, then spent a quarter giving back a fifth of it without ever threatening the trend that produced the gain. Our own technical engine labels it without prompting as a pullback inside an uptrend stack.

Now the timing evidence. The slow stochastic reads 20.6 against 21.5 — the bottom of its range. That is not a prediction; it is a statement that the selling has already done most of what selling does. Consolidations end from there, or they deepen from there, and the difference is usually visible in how the group behaves on a day it should not like.

And the chart turned today rather than merely tightened. The downtrend line drawn from the 22 June top near 68 has been broken: the fund trades at 55.93 as we write, up about 1.9% on the day, above a line it spent the whole of August beneath. It broke out of a base rather than off a single low — the correction put in one low in late July near 51.5 and a second in early September near 53.5 that held above it, a double bottom with a rising floor. The oscillator moved with it: Thursday’s slow stochastic of 20.6 against 21.5, pinned at the bottom with no crossover, reads 33.78 against 24.22 on Friday, K above D and thirteen points higher in a session. Triangles resolve in the direction of the trend that built them more often than they do not, and the trend that built this one carried the fund 90% in three months.

None of that is a guarantee and a broken trendline can be given straight back. What it does mean is that the wait is over: this setup gets answered within weeks, not months, which is the only reason to write about it today rather than in October.

The catalyst was Friday’s jobs report, and it is worth being exact about why it helps rather than hurts here. The August employment report landed at 162,000 jobs against a consensus near 53,000, with July revised from a loss of 23,000 to a gain of 21,000 and June revised up as well — 55,000 of upward revisions on top of a number three times expectations. Unemployment held at 4.1% while participation rose, so the labour force grew and the hiring was absorbed. The rate market read it the obvious way: odds of a September rise jumped from about 50% back above 58%, the two-year yield spiked to 4.425%, the dollar turned higher, gold fell 1.4% towards $4,400, gold miners 1.9%, bitcoin was sold and five of the seven largest technology stocks fell.

That is the discount-rate effect, and it hits anything whose value sits in the distant future. But a jobs report is also a statement about demand, and semiconductors are a cyclical business before they are a growth story — chips go into machines, cars, phones, factories and servers, and the order book follows industrial and consumer activity. A labour market running at triple expectations says the economy is stronger than the market had assumed, and that is straightforwardly good for the people who sell chips into it.

The tape sorted itself accordingly. The Philadelphia Semiconductor Index rose more than 3%; Nvidia is breaking out to new all-time highs; Applied Materials rose 5%, Marvell, Micron, ASML and Arm about 4%, AMD and Intel 3%, Astera Labs 11%. Across the funds, SOXX is up about 3%, SMH and the equal-weight XSD about 2.5%, this fund about 2%. The Dow was roughly flat and the software ETFs were down about 2%.

That split is the observation worth paying for. The same number that raised the discount rate on long-duration assets raised the near-term earnings expectation for cyclical ones, and the market put semiconductors on the cyclical side of the line. A downtrend that has run for two and a half months does not usually end without a reason; this one now has one, and it arrived on the same session as the higher low and the oscillator turn.

Why the fabless cut rather than the whole sector. Because this week graded the layer from the inside, and the grades were not even. Nvidia spent $12.93 billion on Hugging Face, was bought for it, and is breaking out to new all-time highs — at 23.5% of the fund, the heaviest weight is also the strongest chart in it. Broadcom delivered a faultless quarter and was sold 2.7% for guiding to merely excellent; at 16.2%, it is the weak spot rather than the second engine. That asymmetry is the honest version of the bull case and its main vulnerability in one sentence: the fund is turning because its largest holding is making new highs, not because the basket has broadened. What it does not do, either way, is send money to the foundries or the equipment makers — which is precisely what this fund excludes.

There is a second-order reason, and it is the argument this page made in today’s Daily Pulse. If the model market splits into a few enormous closed frontier systems and a long, fragmenting tail of small open ones, the tail runs on inference — billions of cheap calls rather than a handful of vast training runs. The companies that design inference silicon, connectivity and power delivery for that shape are the ones in this basket. That is a slower thesis than a jobs report, and it is the reason the twelve-month view is the one that matters here.

And the horizon on this page is twelve months, which is the frame everything above should be read in. A higher low, an oscillator crossing and a broken line are entry evidence, not the thesis; they tell you the two-and-a-half-month downtrend has produced the conditions under which it usually ends, and they can all be given back inside a fortnight. The twelve-month view rests on slower things: a trend that delivered a 58% gain over the past year and still sits 14.7% above its 200-day line, an order book that its largest holding describes as supply-constrained with $279 billion of forward supply commitments booked, and a demand mix that a stronger economy helps rather than hurts. On that horizon we think this fund trades higher than it does today. It is a diary view held at a two-and-a-half-times market beta, and the levels that would change our mind are in the section below.