AI Is Supplying Blowout Earnings Again

Cisco reported 12% revenue growth and 35% overall orders growth, with hyperscaler AI infrastructure orders up more than 3x to $1.9B in the quarter and fiscal-year hyperscaler orders guided to $9B versus $4B of expected recognized AI revenue. Lumentum delivered even stronger results, with revenue up 90% year over year, adjusted operating margin expanding 21 points to 32%, and guidance calling for 22% sequential growth; shares have risen sharply on the AI infrastructure demand outlook despite rich valuations. The episode also highlighted several non-AI diversification ideas, including Deckers, Casella Waste Systems, Trex, Berkshire Hathaway, and Disney.

Analysis

The market is starting to price a real re-rating in network infrastructure, but the second-order effect is that spend is concentrating into a narrower set of winners with the cleanest exposure to hyperscaler capex. That favors the names with the strongest order book visibility and the most differentiated power-efficiency or interconnect products, while commoditized networking vendors risk being left with lower-margin spillover demand once the front-end buildout normalizes. The key tell is that order growth is outrunning recognized revenue, which means the next 2-4 quarters still have positive revision momentum even if headline shipment growth eventually slows.

The more interesting read-through is that AI infrastructure is becoming a portfolio of bottlenecks, not a single trade. Cisco is benefiting from the “plumbing” layer, but Lumentum sits closer to a capacity-constrained, efficiency-sensitive segment where every incremental watt saved matters; that tends to support higher pricing power and a longer runway than generic optics or legacy telecom gear. The flip side is valuation dispersion: once the market believes supply remains meaningfully behind demand, any small improvement in throughput can trigger multiple expansion faster than fundamentals catch up.

The risk is not demand disappearance but timing mismatch. If hyperscaler capex pauses for even one or two quarters, the most expensive beneficiaries—especially those trading at triple-digit earnings or sales multiples—will de-rate first because expectations are already extended into 2027-2028. By contrast, more mature names with visible buybacks and secular AI attach rates can absorb a slowdown better, which is why the setup is more favorable for barbell exposure than outright chase-the-momentum exposure.

The non-AI diversification segment is notable because investors are implicitly looking for businesses with low narrative beta, not just low correlation to semis. Waste, consumer staples, and capital-light home-improvement adjacencies should hold up better if AI sentiment reverses, but the best risk-adjusted ideas are the ones with idiosyncratic operational catalysts and balance-sheet support. That makes the cleaner compounding stories more attractive than generic defensives if you want diversification without surrendering upside.

Key Decisions for Investors

• Long CSCO over the next 3-6 months on pullbacks, targeting a continuation of order-driven estimate revisions; risk/reward is favorable while hyperscaler order momentum remains intact, but size modestly because this is still a cyclical re-rating story.
• Initiate a paired long LITE / short a weaker legacy networking or optics peer with lower AI leverage for 6-12 months; the thesis is that efficiency-sensitive interconnect demand will compound faster than the market can service it, but the short leg should be a structurally lower-growth comparator.
• Avoid chasing LITE outright at current levels unless using staged entries; if you want exposure, consider long-dated call spreads 9-15 months out to cap multiple-contraction risk while preserving upside from continued supply shortfall.
• For portfolio diversification, add a starter position in CWST or DECK over 1-2 quarters as non-AI compounders with lower narrative beta; these names offer more durable cash generation if AI multiples compress, though upside will be slower than the current AI complex.
• Use BRK.B as a hedge against an AI-led drawdown in growth multiples; it is the cleanest liquid ballast among the mentioned names and should outperform on a relative basis if risk appetite rolls over.