智谱
Price 626.50 HKD | 1d -2.3% | 5d -1.9% | 20d -43.5% | RSI 67
SECTION 1 - LIQUIDITY
Turnover rate (real): 0.24%
Turnover value: 7.22 亿HKD
Volume z-score (20d): -1.51
Struct dVol (shares, 10v10): -31%
Struct dValue (money, 10v10): -40%
dValue/dVol ratio: 1.28
Block days (20d): 2
Amihud illiquidity (20d): 0.010
Float / Total cap: 1,648.1 / 3,054.7 亿HKD
LIQUIDITY, EXPLAINED (each question answered for THIS stock)
Liquidity = how easily you trade WITHOUT moving the price.
Q1. HOW DEEP is the liquidity?
ANSWER THIN
WHY only 0.24% traded (about 1 in 423 shares) - very few shares change hands. A small order can gap the price and exiting real size is genuinely hard; liquidity is a risk here, not a convenience.
DRILL HK$7.22亿 actually changed hands today against a HK$3054.7亿
total market cap - i.e. turnover = traded value / market cap
= 0.24%. The thin pool (vs HK$1648.1亿 float) is what sets the slippage.
CAVEAT: HK$3055亿 total cap is a LARGE company - turnover % naturally
looks low for big floats, and that is NOT illiquidity; check the
Tradeability (Amihud) and Pool-size lines below, which are absolute.
(Depth = turnover rate: traded HK$ / the company's OWN market cap. It is
RELATIVE to size, so a mega-cap shows a low % even when it trades hundreds of
millions a day. The ABSOLUTE 'can I trade my size?' test is the price-impact
(Amihud) + HK$/day scorecard below. Big pool = you trade quietly; thin pool
= your own order moves the price. NB: NONE of this tells you direction.)
Q2. HOW FRESH is the liquidity? (new money, or the same crowd?)
ANSWER COOLING - participation fading
WHY average daily volume is 31% below the prior 10 sessions (~0.69x): fewer players engaged than before, so follow-through is less reliable.
DRILL last-10d avg volume 5,868,362 shares/day vs 8,538,983 in the prior 10d (-31%).
money per 10d: HK$40.92亿 now vs HK$68.42亿 before (-40%) -
freshness is measured in both share count AND committed capital.
(Freshness = 10-day-vs-prior-10-day volume shift. Rising = new blood,
which is what makes a move matter. Flat = churn, and churn means little.)
Q3. WHO is winning the change of hands?
(Shares moving sellers->buyers only says HOW MUCH traded; price says WHO won.)
ANSWER Weakly sellers - price down on flat/fading volume = interest dying (exhaustion).
WHY 20d price change is -43.5% (5d -1.9%) while 10v10 volume is -31% - matching price and volume direction is what separates a real win from churn.
DRILL today's volume sits -1.51 sigma vs its 20d mean (no single-day anomaly).
DATA: 1d -2.3% / 5d -1.9% / 20d -43.5% | Amihud 0.01 (low impact).
VERDICT: [LIQUID tradeability / BEARISH tape] deep book, falling tape - easy to exit, but no long edge here.
Q4. Any SYSTEMATIC / STRUCTURAL change - or just a one-day spike?
ANSWER volume 31% over a full 10-day window = structure draining away persistently
2 block days in 20 = occasional large prints only
DRILL the 2 block day(s) (>2x the prior 20d base) fell on 2026-09-16, 2026-09-23.
WHY structural change = sustained over weeks and shows up in the 10v10 and
block-day counts. A one-day z-score spike alone is noise and fades.
Q5. WHOSE money is it - institutions or retail?
ANSWER INSTITUTIONAL-LEANING (confidence moderate - 2 institutional vs 1 retail signals).
WHY big orders must lift the book to fill (money rises faster than share count);
many small orders nibble without moving the price. That ratio - plus who shows
up in the settlement register and the block prints - is the evidence, not one number.
EVIDENCE:
· [1] money vs shares (10v10): -40% vs -31% = ratio 1.28 -> the MONEY side moved more than the SHARE-COUNT side, i.e. bigger average trade tickets (institutional-size lots).
· [2] block prints (20d): 2 day(s) above 2x base -> occasional large orders.
· [3] settlement register (CCASS, ~14d): biggest builder = FUTU SECURITIES INTERNATIO [B01955] +908,367 sh (+0.35pp; RETAIL online-broker seat).
· next: INTERACTIVE BROKERS HONG K [B01590] +181,600 sh (retail-broker).
· biggest cutter: GOLDMAN SACHS (ASIA) SECUR [B01451] -2,607,153 sh (-0.99pp; institutional).
· [4] mainland (southbound, largely institutional/quasi): +10,612,170 sh over ~30d -> committed demand building.
Q6. How much supply is free to trade (float)?
ANSWER SMALL float - only 54% free to trade.
Fewer shares to absorb demand, so the same turnover moves price
MORE: squeezes fly up, air pockets fall fast. Exciting and dangerous.
DRILL free-float HK$1648.1亿 of a HK$3054.7亿 total cap (54% free, 46% locked).
BOTTOM LINE - LIQUIDITY HEALTH
SCORECARD (1 weak / 2 ok / 3 strong) - ABSOLUTE tradeability (can you trade your size?), not direction (see Q3):
Tradeability (impact) 3/3 Amihud 0.01 - low impact: can trade size cheaply
Pool size (HK$/day) 3/3 HK$7.22亿 traded/day
Freshness (10v10) 1/3 10v10 -31% (fading)
TOTAL 7/9
ACTIVITY (turnover vs float - relative to size, not scored): 0.44% of free float changes hands per day - looks low, but this is a LARGE cap: a big float dilutes the % (activity is relative to size, so it is NOT scored).
Activity (turnover/float): thin | Freshness: fading
VERDICT: HEALTHY (7/9) - the weak link is FRESHNESS (10v10 -31% (fading)), so tradeable size is limited. Offsets: mainland 14% & adding -> committed demand soaking float; institutional-tone flow.
TRADE READ: Grinding lower (price down on flat/fading volume) - exhaustion, not panic; and the pool is deep, so exits are cheap and moves are orderly - which cuts BOTH ways (a fall can grind without a squeezy bounce).
Rule to remember: big turnover is bullish ONLY when price rises WITH it.
READING THE FLOW BEHIND LIQUIDITY
short = 12.38% of turnover (fading) | mainland = 14.17% held (+3.1pp/30d)
READ a moderate 12% of turnover is borrowed supply (fading) - real but not dominant; it colours the tape rather than drives it. mainland keeps adding (+3.1pp) and now owns 14% - committed demand soaking up float, so the smaller remaining supply makes moves sharper both ways.
SECTION 2 - MOMENTUM
Price: 626.50
1d / 5d: -2.3% / -1.9%
20d / 50d: -43.5% / -46.5%
MA5 / MA20 / MA50: 627.50 / 756.88 / 969.96
RSI-14: 67.5
ATR-14: 64.29 (10.3% of price)
Q1. How FAST is it moving - speeding up, or rolling over?
ANSWER DECELERATING DOWN (still falling, but slowing).
WHY pace = the per-day speed of the move. 5d -1.9% over 5 days = -0.38%/day; 20d -43.5% over 20 days = -2.17%/day. A faster recent pace means the move is heating up; a slower one means it is fading.
DRILL pace 5d -0.38%/day · 20d -2.17%/day · 50d -0.93%/day | MA20 slope -22.3%/10d · MA50 slope -12.6%/10d
Q2. Is it REAL momentum, or a fake drift? (price must move WITH volume)
ANSWER NO CLEAN MOMENTUM (price and volume not aligned).
WHY real momentum needs price AND participation to rise together. Price drifting up on
flat/falling volume is not momentum - it is a thin tape that can reverse on no news.
DRILL 5d return -1.9% with 10d/10d volume -31% and today's volume -1.5σ vs its 20d normal.
Q3. How EXTENDED is it - room to run, or stretched?
ANSWER BELOW the mean - pullback / weak positioning.
WHY extension = how far price sits from its average, measured in ATR units
(1 ATR = a typical day's swing = 64.29 pts). Fewer ATRs from the mean =
mid-range; more = stretched and prone to snapping back. Read it WITH the trend:
in an uptrend 'above the mean' is strength; in a downtrend 'below the mean' is weakness.
FACTS (distance from each average, in % and ATR):
· vs MA5 627.50: -0.2% (-0.0 ATR) below
· vs MA20 756.88: -17.2% (-2.0 ATR) below
· vs MA50 969.96: -35.4% (-5.3 ATR) below
· RSI-14 67.5: firm, 60-70 (50 = neutral midline)
· 52w range: 22% up (131.50-2,410.00): lower third
· MA20 slope -22.3%/10d · MA50 slope -12.6%/10d (mean falling)
READ Below the mean inside a falling structure: not 'cheap', but meaningfully stretched to the downside - a bounce can come, yet only a reclaim of the MA20 repairs the trend.
SO WHAT: weak, not cheap - treat a bounce as tradeable only once price RECLAIMS the MA20; until then the trend is still down.
Q4. Is there a DIVERGENCE? (the early exhaustion tell)
ANSWER NONE - price and momentum agree; no warning sign.
WHY divergence = price prints a new extreme while RSI fails to follow. It is a quiet
warning the move is running on fumes BEFORE the price itself turns. We split the last
20 sessions into two 10-day halves and compare each half's swing high and low.
· highs: recent 794.00 (RSI 39) vs prior 1,075.00 (RSI 46) -> no higher high this half (price below the prior peak)
· lows: recent 614.50 (RSI 31) vs prior 680.00 (RSI 31) -> RSI confirmed the low (lower low, RSI not higher)
DRILL neither a higher-high-with-lower-RSI nor a lower-low-with-higher-RSI appeared -
price and momentum are moving in step, so there is no exhaustion tell yet.
SO WHAT: no exhaustion tell - the trend can be trusted for now; a divergence would be your FIRST warning that it is fading, so it is the thing to watch.
Q5. Is the move the STOCK, or just the MARKET?
ANSWER MOSTLY THE MARKET (beta).
WHY beta 0.96 = the stock normally moves about 0.96% per 1% HSI move. Over the last 5d it did -1.4% = market -0.9% (beta x HSI) + stock-specific -0.5%.
DRILL R2 0.01 over 59d - 1% of this stock's swing is explained by the index; the rest is its own story.
BOTTOM LINE - MOMENTUM HEALTH
SCORECARD (1 weak / 2 ok / 3 strong) - STRENGTH & quality of the move ("clean, confirmed trend with room to run?"), not direction (see the VERDICT):
Trend structure 3/3 MA5<MA20<MA50 with the mean falling - clean downtrend structure
Participation 1/3 10v10 -31% (drying up)
Room to run 2/3 -2.0 ATR from MA20 - moderately extended, RSI 67
TOTAL 6/9
DIVERGENCE (the exhaustion tell - a warning flag, not scored): none - price and momentum are moving in step (no exhaustion tell).
VERDICT: HEALTHY (6/9) - the weak link is PARTICIPATION (10v10 -31% (drying up)), so the move is capped there. Offsets: move is mostly HSI beta - it can turn with the index.
TRADE READ: Price and volume are not aligned - no clear direction; so there is no momentum edge regardless of the score.
Rule to remember: momentum is real ONLY when price and volume move together - and even a strong score cuts both ways (strong DOWN momentum is still down).
MOMENTUM VERDICT: [NO CLEAN MOMENTUM]
NOTE: Price and volume are not aligned - no momentum edge.
SECTION 3 - NARRATIVE (trading story)
Net news tone: BEARISH (score -25/100). Active themes: SUPPLY, CATALYST.
NARRATIVE SCORE: -25/100 [MILDLY BEARISH]
DRILL 1 bullish vs 2 bearish headline(s) of 5 (last 30d) | bull theme: CATALYST | bear theme: SUPPLY.
READ News flow leans negative (67% bearish) - the narrative is working against a long, but not decisively.
WHAT THE MARKET IS TRADING ON:
DRIVERS: 禁售承諾一過即再配股兼發CB，事隔兩個月再「抽水」，合共集資淨額近393億元用於下一代GLM基礎模型研發及訓練；上半年收入按年增4倍、經調整淨虧損19.6億元，年底ARR指引據報由24億上調至30億美元。
WHAT BULLS BELIEVE: 收入高增長加上ARR指引上調、ZCode將正式開源，大摩績後睇1,800元、多間券商升目標價，重金投入GLM反映對模型進化的信心。
WHAT BEARS BELIEVE: 密集配股發CB攤薄股東並構成持續供應壓力，富瑞削目標價9%至1,183.79元，憂ARR路徑存風險、下半年雲端毛利率或受壓，市場亦關注集資後「默默」打包用戶代碼庫等爭議。
DEMAND READ: 基本面對ARR上調有支持，但「抽水」供應壓力與券商分歧令股價兩頭受制，短線偏兩向震盪。
[SUPPLY / DILUTION]
- 【2513】智譜禁售承諾一過即再配股兼發CB、集資淨額近393億元資金主用於下一代GLM基礎模型研發及訓練 (13 Sep 2026 · 香港經濟日報HKET)
- 【2513】智譜事隔兩個月再抽水 擬配股兼發CB合共集資逾391億元 (11 Sep 2026 · 香港經濟日報HKET)
[CATALYST / FUNDAMENTAL]
- 【2513】智譜股價曾大 據報年底ARR指引由24億上調至30億美元 (16 Sep 2026 · 香港經濟日報HKET)
[OTHER]
- 【2513】智譜回應ZCode代碼爭議 宣布已刪除雲端數據並推出開源新版 (28 Sep 2026 · 香港經濟日報HKET)
- 【2513】智譜唐傑清華課堂塞滿學生 唐傑：Chat這一仗基本打完了 2026年讓模型自己進化 (17 Sep 2026 · 香港經濟日報HKET)
KEY DRIVERS (filtered):
* [SUPPLY] 【2513】智譜禁售承諾一過即再配股兼發CB、集資淨額近393億元資金主用於下一代GLM基礎模型研發及訓練
* [SUPPLY] 【2513】智譜事隔兩個月再抽水 擬配股兼發CB合共集資逾391億元
* [CATALYST] 【2513】智譜股價曾大 據報年底ARR指引由24億上調至30億美元
* [OTHER] 【2513】智譜回應ZCode代碼爭議 宣布已刪除雲端數據並推出開源新版
* [OTHER] 【2513】智譜唐傑清華課堂塞滿學生 唐傑：Chat這一仗基本打完了 2026年讓模型自己進化
LATEST HEADLINES (last 30 days):
· 【2513】智譜回應ZCode代碼爭議 宣布已刪除雲端數據並推出開源新版 (28 Sep 2026 · 香港經濟日報HKET)
· 【2513】智譜禁售承諾一過即再配股兼發CB、集資淨額近393億元資金主用於下一代GLM基礎模型研發及訓練 (13 Sep 2026 · 香港經濟日報HKET)
· 【2513】智譜股價曾大 據報年底ARR指引由24億上調至30億美元 (16 Sep 2026 · 香港經濟日報HKET)
· 【2513】智譜事隔兩個月再抽水 擬配股兼發CB合共集資逾391億元 (11 Sep 2026 · 香港經濟日報HKET)
· 【2513】智譜唐傑清華課堂塞滿學生 唐傑：Chat這一仗基本打完了 2026年讓模型自己進化 (17 Sep 2026 · 香港經濟日報HKET)
THEME TAGS: CATALYST, SUPPLY
SECTION 4 - FLOW (who is buying, and is it real?)
Regime: step-down
Verdict: FADING (exhaustion)
Ticket size: large-ticket / higher-priced (institutional tilt)
This week vs month: 0.41
IN PLAIN WORDS: LOW ENERGY. Little is happening; no strong buyer or seller. Nothing to trade on.
WHOSE HANDS - THE SETTLEMENT REGISTER (CCASS)
Q. Whose stake actually moved over the last ~14 days?
ANSWER biggest build = FUTU SECURITIES INTERNATIONAL [B01955] +908,367 sh (+0.35pp; now 1.37%).
over 2026-09-18 -> 2026-10-02 (~14d): 142 seat(s) built, 32 trimmed, net ADDING overall (+72,154 sh); custodian/omnibus banks excluded.
TOP BUILDERS:
+908,367 sh (+0.35pp) -> FUTU SECURITIES INTERNATIONAL [B01955]
+181,600 sh (+0.07pp) -> INTERACTIVE BROKERS HONG KONG [B01590]
+122,250 sh (+0.05pp) -> HANG SENG SECURITIES LTD [B01284]
TOP CUTTERS:
-2,607,153 sh (-0.99pp) -> GOLDMAN SACHS (ASIA) SECURITIE [B01451]
-1,156,562 sh (-0.43pp) -> MERRILL LYNCH FAR EAST LTD [B01224]
-821,827 sh (-0.32pp) -> CHINA INTERNATIONAL CAPITAL CO [B01654]
WHY CCASS is HK's central settlement register - every share sits in a
broker/custodian seat. The huge custodian/omnibus seats just hold
everyone's shares, so we skip them and follow the specific seats whose
stake actually changed. A seat that keeps building = its clients are
accumulating. (pp = change in % of the company held.)
Q. Is mainland money involved? (Stock Connect / southbound)
ANSWER southbound holds 37,290,134 sh = 14.17% of the company (+10,612,170 sh / +3.12pp since 2026-09-07).
WHY southbound = mainland investors buying via Stock Connect. Rising holdings
= mainland demand building; falling = they are leaving. It is a real,
named-holder feed, not an inferred proxy.
Q. Is anyone betting AGAINST it? (short selling)
ANSWER short selling = 12.38% of today's turnover (meaningful), and it is falling vs the 5-day average.
WHY short selling = shares borrowed and sold first, hoping to buy back cheaper -
a bet on (or hedge against) a fall. SHORTS COVERING - heavy but fading; covering can itself support the price.
DRILL 02/10/2026 - short HK$89.4M vs 5d avg HK$1.49億; 0.224% of ALL market short-selling.
SIGNALS
[LIQUIDITY/MED] Liquidity fading: last 10d avg volume -31% vs prior 10d — participation drying up.
[LIQUIDITY/HIGH] 2 block-volume days (>2x base) in last 20d: 2026-09-16, 2026-09-23 — institutional-size prints.
[LIQUIDITY/LOW] Amihud illiquidity (20d) = 0.01 -> liquid.
[MOMENTUM/HIGH] Price < MA20 < MA50 — bearish trend alignment.
[RISK/MED] ATR14 = 10.3% of price — very high volatility.
ANALYST NOTES
* Overhang: a SUPPLY/dilution thread is live (placement or CB). This is a structural cap on upside — new float has to be absorbed before rallies stick.
* Structure: only 54% of cap is free-floating — thin float amplifies moves both ways; squeezes up, air pockets down.
SYNTHESIS - HOW TO PIECE THIS TOGETHER
FLOW + MOMENTUM: STABLE: neither flow nor momentum is making a decisive move — no edge from the tape right now.
WHO / WHERE: Money character: large-ticket / higher-priced (institutional tilt). Price is 21% up its 52w range.
NARRATIVE: Headline tone: NEUTRAL. SUPPLY overhang is live (caps upside). A fundamental catalyst is in play.
RISK: volatility extreme (ATR 10%)
PUTTING IT TOGETHER:
Trading at 626.5 HKD. Recent tape: 20d -43.5%, 5d -1.9%. Both volume and price are weakening (interest drying up). Large-ticket / higher-priced (institutional tilt); price sits 21% up the 52w range. Headline tone is neutral with a live supply/dilution overhang that structurally caps upside and a fundamental catalyst in play. Bottom line: NET BEARISH. Flow, trend and/or supply all lean the same way down, so the burden of proof is on the bulls. Don't bottom-fish on cheapness alone — wait for the tape to stop leaking before re-engaging. Key caveat: placement/dilution can cap or reverse rallies; outsized swings demand reduced position size.
ENTRY POINTS - WHERE TO ACT (trade plan)
SETUP: 20d -43.5% / 5d -1.9%, RSI 67, downtrend (MA5<MA20<MA50); price 21% up its 52w range.
PULLBACK ENTRY: buy 587.93-626.50 on weakness | stop 523.64 | T1 755.07 (R:R ~1.2:1)
BREAKOUT ENTRY: distant - the nearest 10-day high is 836.66 (+33.5%), so a breakout is not a near-term trigger; price must first reclaim that level.
TARGETS: T1 755.07 (+2xATR, short-term) | T2 1146.00 (nearest resistance / swing high).
STOP / RISK: ATR 10.3% of price; a stop-out is ~16.4% from spot. Size so a stop costs ~1-2% of the book.
FLAGS: high volatility (ATR 10.3%) -> half size, wider stops | tight float 54% -> squeezes fast both ways | CCASS non-custodian seats net trimming over ~14d (-3,996,684 sh) -> holders reducing | mainland (southbound) adding +10,612,170 sh -> committed dip demand
ENTRY VERDICT: [NO LONG ENTRY] NO LONG ENTRY: tape leans bearish. Only tradable back above a reclaim of 756.88 (stop 692.59); otherwise stay flat and let the tape stabilise.