Asia chip stocks climb after Alphabet raises capex forecast; China underperforms

Alphabet/Google nearly doubled AI infrastructure spending outlook, raising its 2026 capex forecast to $195B–$205B from $180B–$190B (+$15B to +$20B), reinforcing that AI computing demand will stay strong despite cost concerns. Asian chip stocks rallied broadly (e.g., SK Hynix +~5%, Samsung +~3%), while some China-linked AI/semis fell (SMIC -3.4%, Hua Hong -6% and Zhipu AI/Z.AI -3%) on U.S. allegations of technology theft and potential trade restrictions/sanctions, alongside investigations into whether advanced U.S. AI chips were improperly accessed.

Analysis

The best read-through is not "AI is hot" but that hyperscaler spend is still flowing to the tightest parts of the stack: HBM, advanced packaging, and back-end test. That makes SKHYV and SSNLF better expressions than generic AI software, because incremental demand there is still capacity-led and can sustain pricing longer than the initial capex headline cycle. ATEYY should also benefit because test intensity rises nonlinearly as accelerator complexity increases; that usually shows up in orders before it shows up in revenue.

GOOGL is a more mixed asset here. Higher capex supports the thesis that AI monetization is strategic, but it also raises the hurdle rate for free-cash-flow conversion and creates a future multiple risk if payback periods slip. Near term, the market may reward the willingness to spend; over 6-18 months, investors will care more about whether incremental AI revenue outruns the depreciation burden and incremental opex.

The losers are the China-facing AI ecosystem and any domestic foundry exposure that depends on foreign model access or advanced U.S. silicon. HHUSF is the cleanest downside proxy because sanctions risk can hit procurement, customer confidence, and financing before it hits reported earnings. Contrarianly, the market may be underestimating how much of this spend accrues to a few bottleneck suppliers rather than the headline spender; if capex is concentrated in a small number of customer platforms, the supplier side can outperform even if AI enthusiasm later normalizes.

Key Decisions for Investors

• Long SKHYV / SSNLF vs short HHUSF for 1-3 months: best risk/reward on the view that capex intensity and sanctions both widen the gap between non-China memory suppliers and China-facing AI names; target 10-15% relative outperformance, stop if U.S. controls broaden into Korean supply chains.
• Buy ATEYY on pullbacks or via 1-3 month call spreads: test equipment has operating leverage to AI accelerator complexity and usually gets re-rated before the next revenue inflection; risk is an order-book miss or a sudden capex pause from hyperscalers.
• Do not chase GOOGL after the print; if long, keep it as a quality core but hedge with a covered-call structure over the next 4-8 weeks because 2026 capex intensity can cap multiple expansion unless Cloud/AI revenue accelerates again.
• Set an alert on SKHYV/SSNLF if HBM pricing or lead times roll over: that would falsify the "capacity-constrained winner" thesis and is the cleanest signal to take profits on the supply-chain longs.
• Use HHUSF only as a tactical short if sanctions headlines escalate; cover quickly if there is any evidence of policy de-escalation or if China AI shares stop underperforming on the next policy headline.