Is the AI trade over? Bernstein weighs in

The SOX semiconductor index is down 38% from its June 22 high, while Bernstein says the decline reflects multiple compression rather than weaker estimates; all five covered networking companies reporting since the firm’s coverage launch issued guidance above consensus. Analyst Daniel Zhu remains positive on AI longer term but says recovery could take time, citing financing and land/power risks, and sees greater capacity risk in 2028 than 2027. Bernstein left forecasts unchanged but cut price targets on Lumentum to $635 from $900, Celestica to $340 from $495, and Corning to $120 from $140.

Analysis

The key distinction is between demand durability and investment returns: strong token use or lab revenue does not establish that customers can monetize capacity enough to support hyperscaler returns. If that proof remains absent, multiples can stay compressed even with stable near-term estimates. Installed capacity is therefore a weak near-term floor for networking and optical suppliers; forward order conversion and customer concentration matter more than reported guidance beats.

Over days to weeks, positioning and liquidity can dominate fundamentals, so a technical bounce is possible but not a catalyst-backed recovery. Over 1–3 months, monitor AI-lab revenue and financing, hyperscaler capex commentary, and order/backlog revisions at CIEN, COHR, LITE and CLS. Downward estimate revisions—not another price-target cut alone—would be the more meaningful bearish confirmation. Over 6–18 months, permitting and power access could shift value toward already-powered sites and grid capacity, while making 2028 deployments more vulnerable than near-term capacity. That risk is not resolved simply by willingness to pay: interconnection and approval timelines can constrain supply.

Contrarian point: the selloff may be over-discounting 2027 delivery risk, but the analyst’s return-hurdle math is a scenario, not independent evidence of lab economics. With catalysts described as sparse, avoid treating a large drawdown as a valuation signal by itself. The promotional CSCO chart tool provides no investment evidence or defensible entry/stop level.

Key Decisions for Investors

• Do not chase a broad AI-infrastructure rebound pre-catalyst. Keep risk small until lab monetization/financing and hyperscaler capex evidence stabilize; a further cut to supplier order or backlog outlook falsifies the recovery setup.
• Set an alert—not an unconditional buy—for CIEN and COHR: consider staged long exposure only after reported orders/backlog and guidance hold or improve through the next relevant results. Invalidate on a guidance cut, backlog deterioration, or renewed evidence of customer capex deferral; use position sizing rather than an invented price stop pending current prices and volatility.
• Track LITE and CLS alongside CIEN/COHR for relative estimate revisions, not just share-price performance. If optical/networking suppliers diverge, investigate customer mix and order quality before rotating; the article does not provide enough company-level data to justify a pair trade.
• For GLW, do not infer a short thesis from a lower target alone. Reassess only if estimates weaken or the market prices in a sustained reduction in AI-related demand; verify segment exposure and valuation first. No CSCO trade is indicated by this article.