Broadcom to lend Anthropic up to $42 billion to lease its chips, filing says

Anthropic's IPO filing disclosed that Broadcom has agreed to provide up to $42 billion in financing for its AI infrastructure buildout, potentially through convertible debt that could convert into Anthropic equity. The facility could fund roughly one-third of Anthropic's $125.2 billion, five-year TPU-compute lease commitment, while Anthropic is expected to become Broadcom's largest compute customer in 2027. Broadcom forecasts AI semiconductor revenue of about $115 billion in fiscal 2027 and $230 billion in fiscal 2028, but the reciprocal customer-financing structure raises concentration, conflict-of-interest and default-liquidity risks.

Analysis

AVGO’s AI multiple now embeds two linked assumptions rather than a conventional component-sales cycle: Anthropic must continue scaling workloads, and its financing structure must remain fundable. The supplier-financier role can pull forward recognized semiconductor demand and improve visibility through 2027, but it also turns a portion of AVGO’s AI backlog into customer-credit exposure. A slowdown in Anthropic monetization would therefore threaten both order volume and the economics of any financing support, producing a sharper de-rating than a normal hyperscaler capex pause.

GOOG is the underappreciated strategic beneficiary: sustained third-party TPU adoption validates its custom-silicon stack and raises utilization of a platform historically viewed as internal infrastructure. That creates longer-term leverage against NVDA pricing, especially if TPU performance and availability prove adequate at multi-gigawatt scale. Conversely, AMZN’s cloud/distribution relationship risks becoming less economically central if Anthropic’s incremental training capacity migrates toward TPU-led deployments; the relevant watch point is not Claude distribution announcements but AWS workload and cloud-revenue contribution disclosures over the next 1-3 quarters.

The near-term IPO is a liquidity catalyst, not proof of end-demand: public equity can strengthen counterparties’ willingness to finance capacity, while a weak aftermarket would expose how dependent the buildout is on continued external capital. Over 6-18 months, the key falsifier for the AVGO thesis is a gap between AI semiconductor revenue guidance and cash conversion, alongside increased restricted cash, receivables, guarantees, or financing commitments. Consensus may be too focused on AVGO’s headline AI revenue trajectory and insufficiently discounting concentration and correlated credit risk across the AI infrastructure chain.

Key Decisions for Investors

• Maintain an AVGO overweight only with a defined financing-risk monitor: reduce if quarterly disclosures show incremental vendor financing, restricted-cash requirements, or a material deterioration in Anthropic-related payment terms. Upside is continued AI revenue-guide expansion; downside is simultaneous backlog and multiple compression if customer funding tightens.
• Initiate a 6-12 month pair trade long GOOG / short AMZN in equal beta-adjusted notional, sized modestly. The thesis is TPU ecosystem validation and compute-share gains versus AWS concentration risk; exit if AWS growth reaccelerates relative to Google Cloud for two consecutive quarters or if TPU deployment timing slips.
• Use NVDA as the cleaner hedge against AVGO-specific customer-credit risk rather than shorting AI semis outright: long NVDA / short AVGO can work if AI capex remains intact but custom-silicon financing becomes scrutinized. Cover the short leg if AVGO demonstrates AI revenue growth translating into operating cash flow without expanded financing exposure.
• Treat Anthropic IPO pricing and the first post-IPO financing disclosures as event triggers, not a reason to chase AVGO beforehand. A strong IPO with durable trading support lowers near-term counterparty risk; a weak deal or rapid aftermarket break should widen the perceived AVGO credit/concentration discount within days.