Broadcom starts amassing $60 bln to fund chips for Anthropic- Bloomberg

Broadcom-led banks are preparing to syndicate up to $60 billion in financing tied to Anthropic and other AI companies, including a $42 billion senior-secured tranche. Anthropic's IPO filing disclosed that Broadcom agreed to lend it up to $42 billion for infrastructure spending, with potential conversion of debt into Anthropic equity. The financing supports Anthropic's planned access to multiple gigawatts of TPU capacity from 2027, but has raised investor concerns about circular financing and conflicts of interest between Broadcom and its major future chip-design customer.

Analysis

The relevant AVGO risk is not incremental accelerator demand but the migration of customer concentration into credit and mark-to-model exposure. If Broadcom retains a meaningful slice of the facility, reported AI revenue could be discounted by investors as vendor-financed demand rather than arm’s-length backlog, raising both receivables/credit-loss risk and the probability of multiple compression. The key underwriting variables are final hold levels, collateral coverage, pricing, covenants, payment-in-kind features, and whether any equity conversion embeds a valuation below implied public-market expectations.

In the next several days, syndication appetite and loan pricing will be the cleanest real-time referendum on whether banks view this as infrastructure finance or speculative AI-credit risk. A wide discount, elevated spread, or weak distribution would force the market to reassess AVGO’s balance-sheet optionality and could spill into AI names with concentrated, highly capital-intensive customers, including ORCL and GPU/ASIC supply-chain proxies. Conversely, broad third-party placement with limited AVGO retention would preserve the strategic demand signal while sharply reducing the circular-financing overhang.

GOOG has asymmetric second-order exposure: higher utilization of its TPU ecosystem improves infrastructure economics, but financing terms that subsidize a major model developer may intensify competition for cloud workloads and model-layer economics. Consensus is likely treating this primarily as a chip-demand catalyst; the underappreciated issue is that credit markets, rather than semiconductor order commentary, now determine the quality and durability of that demand over the next 6-18 months.

Key Decisions for Investors

• Do not add outright AVGO exposure before financing terms are public. Establish an alert on syndication spread, discount-to-par and AVGO’s disclosed retained commitment; a weak placement or material retained exposure is a 1-3 month catalyst to reduce AVGO versus the SOXX semiconductor ETF.
• If the facility is broadly syndicated near par with robust collateral and limited AVGO recourse, buy AVGO on the initial financing-related pullback with a 6-12 month horizon; the upside case is that investors reclassify the arrangement as externally funded capacity demand rather than balance-sheet risk. Exit if subsequent filings show growing financing receivables, weaker collateral, or customer payment terms extending.
• For a hedged expression while terms remain uncertain, favor long GOOG / short AVGO in equal dollar risk over 1-3 months. GOOG retains compute-platform utilization upside with less direct vendor-financing exposure; close the spread if AVGO demonstrates de minimis retained credit risk or if GOOG cloud-margin guidance deteriorates.
• Avoid treating the announced capacity as incremental sector demand for NVDA or AMD until procurement and funding sources are independently disclosed. Monitor their earnings commentary for customer-funded versus vendor-funded AI infrastructure; confirmation of broader vendor financing would be a sector multiple-risk signal rather than a reason to chase AI hardware.