China Weekly 22.6.2026

brief · 2026-06-22 · China tech, rare earths, state banks · 1-2 weeks

Aavistus briefs. Independent market intelligence — satellites, flows, filings. Register free to follow new briefs.

§01 · Macro snapshot

Indicator Latest Prior Change As of
US 10Y-2Y Treasury spread (pp) 0.27 0.40 −0.13 2026-06-18
US 10Y-3M Treasury spread (pp) 0.63 0.67 −0.04 2026-06-18
US High Yield OAS (pp) 2.63 2.80 −0.17 2026-06-17
US Initial jobless claims 226,000 230,000 −4,000 2026-06-13
US Sahm Rule indicator 0.10 0.13 −0.03 2026-05-01

§02 · Themes of the week

Rare-earth retaliation adds two more US names

China added two US rare-earth producers to its export-control list this week, extending the pattern of using rare-earth licensing as a discretionary lever. The mechanism is a named-firm blacklist, not a tonnage cap. Two names is not yet a supply shock for downstream US users. It is a process shock for the targeted firms' Chinese counterparts, who now need either licenses or workarounds.

The escalation reads cleanly against the recent US designation of BYD, Alibaba, and Baidu as "Chinese military companies." Beijing's response is symmetric in shape if not in scope: name specific US firms, tie them up procedurally, keep optionality on broader controls. The metals counter-data point reinforces the read: this week Bloomberg flagged that Chinese aluminum supply is blunting what would otherwise be a war-shock premium on London aluminum. China is simultaneously restricting selected rare-earth flows and keeping generic metals abundant. The selectivity is the point.

Counter-thesis: this may be tactical, not strategic. Adding two names the Chinese state can later quietly relicense is the kind of move designed to preserve negotiating room ahead of trade talks. Reading two-firm escalation as a fundamental shift to weaponized export policy may over-read the announcement.

This would be wrong if Beijing keeps adding US firms at a one-or-two-per-week cadence into July, or if Western magnet and EV producers report license delays inside 30 days.

Sources: Bloomberg, Investing.com.

Hang Seng tech complex prints fresh 52-week lows

The week's most visible pattern is broad and ugly. Tencent (0700.HK) is down 7.25%. Alibaba HK (9988.HK) is down 7.44% with RSI at 25.6 and a price essentially at its 52-week low. Xiaomi (1810.HK) is down 9.92% with RSI at 17 and a price 0.4% above its 52-week low. Meituan (3690.HK) is down 9.31%. BYD (1211.HK) is down 9.76%. Four of these names sit within two percent of their 52-week lows. The move spans five mega-cap platform names with no profit warnings, transactions, or guidance changes attached to any individual ticker — either the market is pricing through the rare-earth-plus-Pentagon-list combination above, or the move is taking place inside a broader China-decoupling repricing.

When five mega-cap platform names sell off seven-to-ten percent in a single week without name-level news, the pattern reads as sentiment, not fundamentals. RSI readings in the high teens to low thirties on names this size are technically extended on the oversold side, and that is itself worth flagging.

Counter-thesis: oversold readings on Hang Seng tech have been a poor mean-reversion signal through the recent cycle. "RSI 17 on Xiaomi" is not a bounce thesis by itself — extended-oversold conditions in this complex have historically persisted rather than mean-reverted on a one-week horizon.

This would be wrong if either (a) a name-specific driver lands for any of the five within ten days — earnings cut, regulatory action, accounting question — which would tell us the move was specific not generic; or (b) the index rallies more than five percent into next Friday on no news, confirming the oversold-bounce read.

Sources: market data.

China's physical-AI push gets a public-attention pulse

Several outlets this week converged on a story about China shifting AI capital toward humanoid-robotics training. Alibaba released new AI models targeted at robot tasks. A separate piece flagged Alibaba's scientific-AI model beating Microsoft's "using one grammar" on a science benchmark. A Clarín dispatch framed Beijing's bet as humanoid robots trained by constant physical interaction in real environments, contrasted against the chatbot-writing-and-coding lane that dominates headline AI coverage.

This matters because it sets up a different industrial logic than the OpenAI / Nvidia / hyperscaler narrative. If China commits real factory floor space and provincial subsidies to humanoid-training infrastructure — and the BABA, BIDU, and Xiaomi capex prints over the next two quarters confirm it — the read-through is to industrial-automation suppliers, not to the language-model layer.

Counter-thesis: a press release and a single benchmark do not equal a hardware-commitment level of capex, and the Western press tends to over-read state-direction signals from Beijing. The Polymarket line on "best AI model at end of June" reads as narrative trading, not as ground truth on capability.

This would be wrong if Q2 capex disclosures from BABA, BIDU, or Xiaomi show no meaningful increase in robotics-specific capital outlays.

Sources: Clarín, TechTimes, Polymarket.

The catch ▸ Sahm Rule rolled back to 0.10 while jobless claims dropped to 226k.

Two of the US labour-stress signals in §01 moved the right direction at the same time. The Sahm Rule indicator dropped from 0.13 to 0.10, and initial jobless claims fell from 230k to 226k as of 2026-06-13. The Sahm reading at 0.10 is well below the 0.50 recession-trigger threshold and the directional move is away from it, not toward it. The catch sits in the data because it cuts against the China-decoupling-driven Hang Seng sell-off above — if the US labour print into the July 2 NFP confirms cooling-without-cracking, the global risk backdrop that drives Hang Seng beta becomes less of a one-way bid for safety. The China-specific repricing would have to stand on its own without US-recession tailwind.

§03 · Companies of interest

Research surface — not investment advice.

Ticker Exchange Sector Theme link Technical snapshot
0700.HK Tencent Hong Kong Communication Services Hang Seng platform sell-off Price 430.00 (52w 420.4–683.0); −7.25% 1w; RSI 48.0; below 200d and 30w MAs.
9988.HK Alibaba Hong Kong Consumer Cyclical Hang Seng sell-off; physical-AI tape Price 102.00 (52w 100.8–186.2, at low); −7.44% 1w; RSI 25.6 (oversold); 13w RS vs SPX −25.0%; below both MAs.
1810.HK Xiaomi Hong Kong Consumer Cyclical Hang Seng sell-off; humanoid-robotics read Price 23.60 (52w 23.5–61.5, at low); −9.92% 1w, −56.4% 1y; RSI 17.0 (deeply oversold); below both MAs.
3690.HK Meituan Hong Kong Consumer Cyclical Hang Seng sell-off Price 70.65 (52w 69.5–144.8); −9.31% 1w, −50.9% 1y; RSI 37.7; below both MAs.
1211.HK BYD Hong Kong Consumer Cyclical China EV cycle softness Price 78.10 (52w 76.7–139.9); −9.76% 1w, −44.5% 1y; RSI 28.2; below both MAs.
1398.HK ICBC Hong Kong Financials State-bank divergence Price 6.90 (52w 5.42–7.32); +18.0% 1y; RSI 60.0; above both 200d and 30w MAs. The only Chinese name in this table above its moving averages.
BABA US (NYSE) Consumer Cyclical Physical-AI tape; "Chinese military companies" list overhang Price 107.10 (52w 103.71–192.67); −4.96% 1w, −28.3% 6m; RSI 22.0; 13w RS vs SPX −24.2%; below both MAs. Industry trend NEUTRAL. Carries adjacent regulatory overhang from US "Chinese military companies" designation.
PDD US (Nasdaq) Consumer Cyclical Temu margin question Price 79.56 (52w 78.51–139.41); −2.14% 1w, −27.0% 6m; RSI 36.3; 13w RS vs SPX −28.4%; below both MAs. Industry trend NEUTRAL. Recent commentary frames Temu expansion as growth-vs-profitability question.
NIO US (NYSE) Consumer Cyclical China auto cycle; CEO tone shift Price 5.02 (52w 3.34–8.02); −4.02% 1w; +45.9% 1y; RSI 33.5; below both MAs. Industry trend DEGRADING. CEO publicly stated this week that China auto is unlikely to return to a "golden era."
BIDU US (Nasdaq) Communication Services "Chinese military companies" listing — primary frame Price 111.76 (52w 83.30–165.30); −3.75% 1w, −18.8% 1m; RSI 22.4; below both MAs. Industry trend NEUTRAL. The recent US "Chinese military companies" designation dominates the framing — technical setup not analyzable as an industry-cycle signal in isolation.

Technicals: market data, computed 2026-06-22.

§04 · Calendar ahead

§05 · Methodology + disclaimer

Compiled from public macroeconomic data and financial press.

This is research material, not investment advice.

Brief generated 2026-06-22. Version: weekly geographic · v1.