Cisco & Cerebras Orders Up, Stocks Down

Cisco shares were down 7.4% after reporting strong AI hardware demand (AI orders of $9.3B vs ~$9.0B expectation, and product orders up 35% YoY) but service revenue came in below expectations, and investors wanted more explicit AI order guidance for 2027. Cerebras, a new AI-chip IPO, dropped ~13% following a confusing earnings release: GAAP revenue +74% missed estimates, while “core revenue” more than doubled and gross margins fell ~6 percentage points, with management still projecting revenue to triple YoY in 2027. The podcast also highlighted Marqeta’s improving profitability (GAAP net income positive, $120B payment volume, +32% YoY) alongside strong growth in BBB Foods (+38.7% sales YoY, ~20% same-store sales growth).

Analysis

Cisco is being punished less for fundamentals than for not looking like a clean AI compounder. Near term, that usually means multiple compression until the market sees another order-book inflection; over 1-3 months the key is whether AI-driven mix can offset services normalization and keep EPS outgrowing revenue. The better read-through is for network peers: ANET remains the cleaner beneficiary of enterprise AI refresh cycles, while DELL and even TSM benefit downstream from continued data-center capex rather than from any one vendor winning share.

Cerebras is not a near-term Nvidia killer; it is a proof point that AI buyers will pay for specialization only if it reduces total system cost enough to justify integration risk. The second-order winner is still TSM, because any real scaling path requires foundry capacity, while the loser could be hyperscale cloud margins if customers keep fragmenting workloads across custom silicon. The risk is simple: if margin recovery slips one more quarter, the market will re-rate this as a capital-intensive science project rather than a platform.

The better risk/reward is in the quieter compounders. XMTR has a tangible workflow advantage from instant pricing plus Siemens distribution, MQ is finally reducing single-customer dependence while turning profitable, and TBBB has a durable value-trade tailwind that can compound into store expansion if Mexico consumer demand stays resilient. The contrarian miss is that the market may be over-focused on flashy AI names when the cleaner monetization path is in infrastructure-enabling software, payments, and hard-discount retail.

Key Decisions for Investors

• Pair trade for 1-3 months: long ANET / short CSCO into any post-earnings bounce in CSCO. Thesis: the market pays up for pure-play AI networking over diversified legacy mix. Falsify if CSCO re-accelerates software/services and raises AI order guidance again.
• Stay cautious on CBRS; if expressing a view, use small call spreads only after evidence that gross margin has troughed. This is a 6-18 month optionality trade, not a core long. Falsify if margins stay under pressure for two consecutive quarters or cash burn accelerates.
• Long XMTR on a 6-12 month horizon. The Siemens integration is a distribution catalyst, and the model benefits from more digital procurement in fragmented industrial supply. Best entry is on any broad industrial/AI selloff; stop if active buyer growth decelerates materially.
• Add MQ on weakness as a self-help profitability story with upside if Block concentration keeps falling. This is a 1-2 quarter catalyst with 12-month upside if margins and buybacks persist; invalidate if top-line growth re-accelerates downward with no new customer mix.
• Small long TBBB versus consumer staples/grocery proxies for a 6-18 month structural value-trade. The risk is Mexico consumer slowdown or margin pressure from aggressive store rollout; trim if same-store sales normalize sharply below high-teens.