CITIC Securities: A-Share Correction Largely Over, August Recovery Expected
N.R. Finch
CITIC Securities argues this A-share pullback is a crowded-trade unwind, not a deleveraging shock — margin balances fell just 14% vs a -32% global average, tech ETFs kept buying on the way down, and the odds of a broad August recovery are rising.
What kind of sell-off was this?
CITIC's diagnosis: a crowded-trade collapse, not a Korea-style leverage blowout.
Margin balances fell from a peak of ¥3.01 trillion to ¥2.59 trillion, a cumulative drop of about 14%. This means → global deleveraging episodes average -32%; this round is less than half that, and the leverage cushion remains intact.
The market-wide margin collateral ratio stood at 264.5% as of July 30 — down from 296.4% at end-June, but still well above warning levels.
What were ETF flows doing?
The ChiNext Index fell 25.8% from June 25 to July 30. Over the same period, tech ETFs logged cumulative net inflows of roughly ¥165.5 billion.
Of 15 down-trading days, 13 saw net subscriptions; the single-day peak hit ¥23.6 billion. This means → the market was not "out of buyers" — crowded trades were unwinding while ETF capital stepped in from the other side.
Broad-market ETFs posted ¥478.4 billion in net subscriptions for all of July, with net inflows on 19 of 23 trading days — clawing back about 29% of first-half net redemptions. In plain terms = nearly a third of the money that left in H1 has already come back.
Which tech names got hurt the most?
CITIC screened 37 high-liquidity-pressure pan-AI tech stocks. Their average Q2 share-price gain was 114%, concentrated in April–June, and none appeared in active mutual funds' top-30 holdings.
This means → these names were fueled mainly by margin and retail buying, with little institutional backing — fast on the way up, unprotected on the way down.
As of July 31, the weighted cost basis of May–June buyers showed an average paper loss of ~34%. CITIC tracks the excess-return gap between "fringe tech" and "core tech" as a bottoming signal: it plunged from +31% at end-June to -26% on July 21, then rebounded to around -14% — signs that the liquidity shock inside the tech sector is largely done.
Is external pressure on non-tech easing?
The Fed's July meeting offered no substantive hawkish action. U.S. Q2 GDP and June PCE data both came in below expectations, further weakening the near-term rate-hike narrative. This means → the "strong dollar + rate hikes" combo that had been capping demand expectations for non-AI sectors is loosening.
Domestically, the Politburo meeting shifted its tone from April's "strong start, key indicators better than expected" to "pay high attention to difficulties and challenges in economic operation," emphasizing timely incremental policy and stronger counter-cyclical adjustment.
In plain terms = cooling rate-hike expectations abroad plus a policy-readiness signal at home — the macro backdrop for non-tech sectors is improving at the margin.
Did more stocks actually go up in July?
2,546 A-shares posted monthly gains in July, or 46.0% of the market — far above June's 1,419 (25.7%).
Of those, non-tech gainers totaled 2,291, or 50.6% of the non-tech universe. This reflects a broadening: the rally is no longer confined to a handful of tech trades.
In plain terms = in June, only a quarter of stocks were rising and almost all were tech. In July, nearly half were rising, and the majority were non-tech. Market breadth is healing.
What does CITIC recommend?
CITIC maintains its "three convergences" medium-term call: upstream AI hardware's outperformance over downstream platforms narrows; non-AI industrials' valuation discount to global peers narrows; the extreme tech-vs-non-tech split narrows.
Within tech: use any bounce in fringe names to rotate into core assets — optical-communications leaders, foundry platforms, and semiconductor equipment.
Outside tech: add exposure to energy-chemicals, non-ferrous metals, non-bank financials, and innovative pharma. CITIC cautions that even if the tech sector recovers broadly in August, performance within it will likely diverge — a fundamental fix to the current chip structure still requires a major industry-level breakthrough to draw fresh capital.
Content is for reference only, not financial advice.