The numbers say what the marketing won’t. Across 5,550 A-share companies, first-half revenue reached ¥37.74 trillion, up 7.09%, and net profit reached ¥3.58 trillion, up 18.57%. That is the broad stroke. The interesting line is the one below it: the sci-tech board’s net profit rose 437.6%.
Strip the hype away and you get this: an aggregate earnings number tells you the market is healthier; a segment number tells you where the health is. Revenue up seven percent, profits up eighteen — that spread itself is already a statement about margins. But the 437.6% line is the real signal, and it deserves closer reading than the headline.
Reading the 437.6% line with discipline
Look at the yield curve on that line — the sci-tech board’s gain. A 437.6% jump in net profit is not organic steady-state growth; it is a base effect on top of a recovery. The numbers say the denominator mattered: profits were depressed, and they are normalizing while the technology cycle runs hot.
That is not a criticism. It is a specification. When you read the figure as “earnings recovering from a low base inside a strong tech cycle,” the true content is: the electronics sector posted net profit growth of 195.11% year on year. That is the load-bearing number. AI compute demand and resource prices together pulled the whole index’s quality up.
What the sector mix actually shows
The composition is the constraint. The aggregate 18.57% would be easy to misread as “everything is fine.” It is not a uniform rise — it is a concentrated one. Electronics at +195% carries a disproportionate share of the improvement; the sci-tech board’s +437.6% is a subset of the same force.
Here is the honest engineering read: the earnings mix is migrating toward science- and compute-heavy sectors, and that is durable in direction but not guaranteed in magnitude. The numbers say the structure of A-share earnings changed this half; they do not say the slope will hold next half.
Let me correct a term I nearly used. “AI-driven” sounds like a narrative; say it as a mechanism instead — compute demand raises semiconductor prices and volumes, which flows through to equipment and materials. Mechanism first, narrative never.
The constraint to watch
For anyone reading the ledger, the real constraint is not this half’s growth — it is the sustainability of the concentration. If electronics and resources carry the market, then a cooling in AI capex or a dip in commodity prices hits the same concentrated lines hardest.
No hype needed on either side, and the numbers say the same at scale. The half-year books show a market whose profit engine has shifted toward science and materials — a structural change worth noting, not a guarantee worth betting a plant on. Watch the segment trend over two more quarters before calling the migration a trend.
The numbers say the mix changed. That’s the spec. The tolerance on that spec is the next two quarters. And the numbers say it in the mix: read at scale, that is the whole engineering statement, and the tolerance is what the next two quarters will test.
Breaking down the margin spread
Let me put the margin story under the lens, because the revenue-to-profit spread is the line an operator reads first. Revenue up 7.09% while profit rises 18.57% means margins expanded by roughly eleven points across the index. That is not a small number, and it did not come from pricing power alone. It came from a mix shift: the profit weight moving toward sectors where the cost curves are favorable and the demand curves are steep.
The first-order cause is the one everyone names: AI compute demand running hot while commodity prices lift the resource side. Both are real. The second-order cause is what the aggregate hides — the companies growing fastest are the ones with the highest operating leverage, so their marginal revenue converts to profit at a higher rate. When that kind of firm dominates the increment, the index’s margin line looks better than the median company’s actually is.
The engineering discipline here is to ask which of these causes is durable. Demand from compute is structural if the build-out continues; commodity prices are cyclical and can reverse without asking permission. The mix that produced +18.57% is a blend of both, and blends are exactly what you re-test each quarter.
The sector-level inspection
Now the inspection the aggregate cannot do. Electronics at +195.11% is the load-bearing beam; the sci-tech board at +437.6% is the concentrated expression of the same force. Reading them together is the only honest way: the broad index rose because the technology sector rose, and the technology sector rose because the demand for compute, memory, and advanced components is genuinely at a cyclical and structural high.
What the numbers do not show is uniformity. Inside electronics, the spread between the AI-facing names and the consumer-facing names is wide — one group is running at full utilization, the other is digesting inventory. The aggregate is a blend, and blends hide tolerance bands. When you read a +195% sector number, read it as an average of a hot core and a lukewarm periphery, not as a single temperature.
That is the real constraint on the whole report: the quality of the mix is better than its breadth. Profit growth is concentrated in the sectors where the cycle is strongest, and concentration is a risk characteristic, not a comfort. The numbers say the mix changed; the tolerance on that spec is the next two quarters.
What the tolerance band allows
Let me state what the report does and does not permit. It permits saying this: the earnings cycle is real, the technology leadership is confirmed, and the margin recovery is broad enough to matter. It does not permit saying: the recovery is uniform, the commodity tailwind is permanent, or the concentration is sustainable at current speed.
The next two quarters are the test. If the electronics core holds its margin while commodity prices normalize, the mix change is structural and the spec is upgraded. If the margin spread narrows as the commodity contribution fades, then part of the +18.57% was cyclical wallpaper, and the honest reading was always that it was a blend.
Watch three numbers quarterly: the electronics sector’s profit growth, the sci-tech board’s profit growth, and the spread between revenue and profit growth for the whole index. Those three lines will tell you, with far more precision than any headline, whether the earnings quality is compounding or merely recovering.
The numbers say what the marketing won’t. The marketing will say the market is strong. The numbers, read as a spec sheet, say the strength is real and concentrated, and concentration is the dimension that gets tested next.
The operating-leverage read
Let me examine the spread through the operating-leverage lens, because it explains the headline better than any sentiment. Operating leverage is the ratio of fixed to variable cost in a business: when revenue rises and fixed costs stay put, the profit line jumps faster than the top line. A 7.09% revenue rise converting into an 18.57% profit rise is the signature of exactly that mechanism running through the index’s heavier sectors.
The machinery, the memory, and the foundry names are fixed-cost-heavy: their capacity is the expensive part, and the marginal wafer or module is cheap to produce once the line is paid for. When demand climbs, their profit accelerates. That is why the concentration matters — the leverage is not distributed evenly, it sits where the fixed capital is, and that is precisely the electronics and semiconductor complex that delivered the +195% line.
The operator’s caution follows automatically. Operating leverage works in both directions. The same fixed costs that magnify profit on the way up absorb margin on the way down, and the unwind is as sharp as the run-up. The +18.57% is partly a one-way valve that only looks robust while demand keeps rising.
What the resource leg contributes
Now the second leg of the blend, because leaving it out would be a tolerance error. Commodity and resource names contributed meaningfully to the aggregate profit rise, and their contribution is a different species of signal. Resource earnings respond to price, price responds to supply and demand tightness, and tightness right now is partly real (disrupted supply, strong industrial demand) and partly cyclical (inventory rebuilding).
The distinction between the two legs is the whole forecast. The compute leg is a build-out story with a visible horizon; the resource leg is a price story that can reverse when supply normalizes. A spec sheet separates the two, and the honest read of this report separates them too: the durable part of the earnings improvement is the technology cycle, and the cyclical part is the commodity tailwind. Treating them as one number is how blend errors get made.
This is also why the next two quarters matter so much. They will separate the legs. If the electronics margin holds while commodity prices soften, the improvement is real quality. If both soften together, the market gets its correction. Either way, the aggregate number was never the spec — the decomposition was.
The re-test, scheduled
Let me close by scheduling the re-test, because a number without a re-test date is not engineering, it is decoration. The quarterly cadence gives us three checkpoints before the year closes. At each one, run the same three lines: the electronics sector’s profit growth, the sci-tech board’s profit growth, and the revenue-to-profit spread of the whole index.
If the spread narrows while the electronics core holds, the mix change is real and the earlier read stands. If the spread narrows because the electronics core itself fades, the concentration argument needs rewriting. And if the spread holds while commodity prices fall, that is the strongest possible signal — it means the margin improvement has detached from the cyclical leg and become structural.
That is the spec, properly stated. The numbers say the mix changed — and that is a real statement. The tolerance on it is the next two quarters, and the instrumentation to measure it is already in the report, waiting to be re-read.
One closing note on the tolerance, because it is the kind engineers leave on a drawing. The report is a good report — the earnings mix genuinely changed, and the technology complex is genuinely leading. The discipline is simply to hold the two legs apart in your head: the compute cycle that compounds, and the commodity tailwind that can reverse. Read them separately, re-test them quarterly, and the +18.57% becomes a spec you can manage instead of a headline you can only quote.
And when the tolerance holds, the report stops being a data point and becomes a baseline. Baselines are what the next quarter is measured against.