Cisco & Cerebras Orders Up, Stocks Down

Cisco shares fell 7.4% after results despite strong AI-linked demand, as service revenue came in slightly below expectations. Cisco’s AI orders were $9.3B vs guidance around $9B, but management provided no new 2027 AI-order guidance and remaining performance obligations rose only ~7% in the quarter, contributing to a cautious read. Cerebras, a new AI chip IPO, dropped about 13% post-close after GAAP revenue rose 74% but missed estimates, with core revenue beating guidance while margins fell nearly 600 bps and investors grappled with complex reporting; the company still expects core revenue to triple in 2027. Overall, the segment highlights solid AI traction but mixed execution and valuation risk across AI infrastructure names.

Analysis

Cisco’s reaction looks more like skepticism about duration than about execution: the market is discounting that AI bookings are still a small part of a very large, slower-moving annuity base. The near-term fight is not revenue growth, but mix quality—if services and renewals stay the ballast, the multiple stays capped even with strong product orders. The cleaner expression of AI networking remains ANET, which should keep trading at a premium as long as buyers prefer faster-growing pure-play exposure over Cisco’s conglomerate profile.

Cerebras is less a direct Nvidia disruptor than a demand-expander for the whole stack. If custom silicon lowers inference cost, the first-order winner is the foundry/packaging chain, especially TSM, while the second-order winner is still NVDA if cheaper AI simply drives more token consumption. The main risk is not technological rejection; it’s capital markets and execution—opaque accounting, margin volatility, and customer concentration can keep CBRS in the penalty box for months even if the long-term thesis remains intact.

Contrarian takeaway: the market may be overpricing complexity and underpricing compounding. Names with cleaner operating metrics and visible customer adoption—XMTR, MQ, and TBBB—look more investable than the more narrative-heavy AI stories because their growth is easier to underwrite and less dependent on future disclosure. For Cisco, the falsifier is simple: if product-order growth doesn’t translate into better forward revenue and guide-ups over the next 1-2 quarters, the stock remains a value trap rather than a hidden AI winner.

Key Decisions for Investors

• Buy CSCO on further weakness for a 1-2 quarter trade; thesis only works if AI/product orders convert into visible revenue acceleration. Falsify if next print shows order growth below ~20% or management refuses to raise AI expectations again.
• Pair trade: long ANET / short CSCO over 3-6 months to express the market’s preference for pure-play AI networking over legacy annuity exposure. Target is relative multiple divergence if AI capex stays strong.
• Long TSM as the second-order beneficiary of custom-ASIC and AI hardware ramp. Use any post-earnings pullback as entry; thesis breaks if advanced-node or packaging commentary softens materially.
• Avoid chasing CBRS until gross-margin recovery is actually visible and disclosure gets cleaner. If you want exposure, wait for confirmation that margin troughs are behind it; otherwise keep it on watch, not in the book.
• Add MQ or XMTR on pullbacks as cleaner compounding alternatives to AI-adjacent complexity. Both are better fits for capital today because the unit-economics story is easier to verify than in the hardware narratives.