Broadcom: The Market Is Likely Looking At OpenAI/Anthropic All Wrong

The author maintains a buy view on Broadcom despite near-term weakness, citing its strategic position in custom silicon and data-center networking for AI hyperscalers. Broadcom is securing multi-year chip supply deals with Anthropic and OpenAI using debt-financing structures, while its forward non-GAAP PEG of 0.59x is described as below sector and historical averages.

Analysis

The key underwriting question is whether these arrangements create durable, cash-generative economics—or chiefly improve reported backlog while shifting financing, utilization, or counterparty risk onto Broadcom. Verify who ultimately bears the debt, whether customer commitments are take-or-pay, and how much of the expected value converts to free cash flow. Until those terms are clear, treat “multi-year visibility” as a claim to diligence, not an earnings floor.

Over 1–3 months, the stock’s reaction is likely to hinge more on AI infrastructure spending signals and guidance than on the headline PEG: that ratio is sensitive to non-GAAP definitions and forecast growth, and can look cheap just before estimates reset. A deterioration in hyperscaler capex or weaker networking/custom-silicon orders would pressure both earnings expectations and the growth multiple.

Over 6–18 months, custom silicon can broaden AI compute options and sustain demand for networking, but it also gives hyperscalers leverage to renegotiate economics or develop more of the design internally. Marvell and other silicon suppliers may compete for programs; capacity constraints across the supply chain could also delay revenue conversion. The constructive thesis is falsified by weaker order conversion, rising financing exposure without corresponding cash generation, or guidance cuts tied to customer spending. There may be no attractive trade if the current price already discounts the expected growth.

Key Decisions for Investors

• Do not add solely on the stated PEG. Before initiating or increasing a position, verify the growth assumptions, deal economics, cash-flow contribution, and whether financing obligations sit with Broadcom or counterparties.
• For existing holders, retain exposure only while custom-silicon and networking demand converts into reported revenue and cash flow; reassess on guidance cuts, deteriorating order conversion, or evidence that Broadcom is absorbing material funding or utilization risk.
• If seeking exposure, scale in on broad semiconductor weakness rather than chasing a positive narrative; use a defined-risk position and size it for the possibility that AI capex expectations reset before long-term demand does.
• Track hyperscaler capex commentary, customer concentration and contract terms, and competitive program wins by Marvell. A sustained drop in orders or cash conversion would invalidate the constructive thesis; stronger conversion with limited balance-sheet risk would support adding.