CITIC Securities: AI Price-Hike Chain's Risk-Reward Ratio Drops Sharply, Investors Should Return to Reasonable Expectations
Nashnova编辑部
CITIC Securities calculates that China's AI re-pricing chain has seen its risk-reward ratio compress from 6× to just 1.2× in two weeks of rallying; the firm urges investors to stop chasing a one-off valuation reset and rotate into core assets.
Odds dropped from 6× to 1.2× — what does that mean?
On July 29, the median forward P/E of CITIC's AI re-pricing sample sat at 1.18× its cycle starting point. Upside far exceeded downside — the implied reward-to-risk ratio was roughly 6×.
By August 14, after a sharp rally, that median had climbed to 1.5× the starting point. The same calculation now yields a ratio of just 1.2×. This means → every dollar of potential gain is matched by almost a dollar of potential loss — the asymmetry is gone.
CITIC's bottom line: the industry trend hasn't changed, but trapped longs above and short-term profit-takers below coexist. Betting on another one-shot valuation reset is extremely difficult.
So many new narratives — why not jump in early?
Markets have recently built stories around AI for Science — using AI to accelerate scientific research — and RSI (Robotic Super Intelligence), but CITIC's review shows a clear pattern: a significant time lag between a technology breakout and the equity rally it eventually triggers.
The o1 reasoning model launched in late 2024; its equity trade didn't fire until June 2025. Coding Agent exploded in February 2026; North American compute names didn't move until late March, and China's AI chain waited until May for a broad rally.
In plain terms = a new concept doesn't equal an immediate trade. Investors can wait for real product traction and verified adoption rates before deploying capital.
Grok 4.6 looks strong — can it lift the whole sector?
Grok 4.6 has impressed, and some investors read it as proof that "bigger models win" — reviving the "brute-force scaling" narrative on the training side.
CITIC argues that unless model capabilities show a non-linear jump — not gradual improvement but a sudden leap — this marginal gain is only enough to support a bounce in core names, not to re-rate the entire sector.
This reflects a market that now prices AI narratives with more granularity — one strong product no longer automatically lifts all boats.
Exports and domestic demand — where is each constrained?
Export-facing names face two headwinds: ① Sino-EU trade friction remains tense, with September–October as the key window for interim results; the EU's carbon border adjustment mechanism — an extra levy on high-carbon imports — and the forced-labor product regulation will steadily raise compliance costs. ② RMB appreciation is already biting: among 384 non-financial companies that have reported H1 results, Q2 financial-expense ratios rose 0.47 percentage points year-on-year; for firms with over 30% overseas revenue, the increase reached 1.04 to 1.74 points, driven mainly by FX translation losses.
Pure domestic plays do enjoy a clear lead: since June, China semiconductor materials & equipment is up +22.8% and STAR Market chips up +1.3%, outperforming the Philadelphia Semiconductor Index (−3.2%) and South Korea's KOSPI 150 (−19.1%).
But CITIC estimates that since August, some capital exiting North American compute names has crowded into China's compute chain, keeping positioning density elevated and capping further valuation upside.
A record number of stock doublers — why is that a warning sign?
In Q2 2026, 275 companies listed in Shanghai and Shenzhen saw their share prices double — the highest count since Q1 2007.
CITIC puts it bluntly: "This kind of routine doubling is a once-in-a-decade event, not market normal." This means → future opportunities will look more like phased valuation repairs in names with earnings support, not a broad-based surge.
The market is likely entering a "withdrawal period" — a transition from elevated volatility back to normal ranges — with overall volatility grinding lower.
How should investors reposition now?
Within tech: use the bounce in AI re-pricing names as a window to rotate toward gas turbines, wafer-fabrication platforms, and semiconductor equipment — core assets where "volume certainty" matters more than "price explosiveness."
Outside tech: add exposure to energy chemicals, non-ferrous metals, innovative pharma, and leading brokerages with export potential.
In plain terms = a declining-volatility phase is also when positioning structures clean up — momentum chasers exit, quality assets get re-priced, and the foundation for a longer-term rally is laid.
Content is for reference only, not financial advice.