Broadcom Lines Up $60 Billion to Build Chips for Anthropic

Today’s Focus

Broadcom has begun assembling about $60 billion in debt financing to build custom artificial intelligence chips for Anthropic, the maker of the Claude chatbot, according to Bloomberg News. The arrangement includes a reported $42 billion loan commitment from Broadcom to Anthropic to fund the infrastructure buildout, Yahoo Finance reported in a summary of market coverage.

The deal would be one of the largest single corporate borrowings tied to AI infrastructure to date. It extends a pattern of giant, interlocking agreements between the handful of companies that design AI chips, train frontier models, and sell cloud capacity.

Anthropic, founded in 2021 by former OpenAI researchers, is already a major customer of Nvidia processors and uses Amazon and Google data centers to train and serve Claude. Both Amazon and Google have invested billions in the startup. The Broadcom tie-up would give Anthropic a second, custom silicon pipeline distinct from Nvidia’s general-purpose GPUs.

Broadcom shares rose on the news, with Seeking Alpha and Barron’s noting that the company is increasingly viewed by investors as a direct beneficiary of AI capital spending alongside Nvidia. Micron, another chip supplier, fell on the same trading day despite reporting strong earnings and guidance, according to Yahoo Finance’s midday wrap.

Neither Broadcom nor Anthropic has published a full term sheet. Bloomberg reported that the $60 billion figure reflects the total debt Broadcom is seeking to raise over time to fund chip production for Anthropic, not a single bond issuance. The companies previously disclosed a partnership earlier in 2026 to co-design accelerators, but had not attached a dollar figure of this size.

The announcement lands as regulators in Washington and Brussels are scrutinizing the financial entanglements between leading AI firms, their investors, and their suppliers.

The Debate

Supporters argue

Backers of the deal frame it as the kind of long-horizon industrial investment the United States needs to keep a lead in AI. Broadcom CEO Hock Tan has told investors that custom accelerators designed for a single customer’s workloads can deliver far better performance per dollar than off-the-shelf chips, a pitch Bloomberg has reported repeatedly in coverage of the company’s AI segment.

Anthropic executives have argued that diversifying away from a single chip supplier reduces risk and lowers long-run training costs. In public comments cited by Reuters earlier this year, co-founder Dario Amodei said access to compute is the single biggest constraint on building safer and more capable models.

Industry groups including the Semiconductor Industry Association have said deals of this scale support domestic chip packaging and advanced manufacturing jobs. Wall Street analysts quoted by Barron’s described the transaction as “a validation of Broadcom’s AI roadmap” and a sign that hyperscaler-style spending is spreading to model developers themselves.

Critics argue

Skeptics say the AI sector is building a web of circular financing that could amplify losses if demand softens. Economist Paul Kedrosky, writing on his newsletter, has described the current pattern of chipmakers lending to customers who then buy the chipmakers’ products as “vendor financing on a scale we have not seen since the telecom bubble.”

Senator Elizabeth Warren (D-MA) and other Democrats on the Senate Banking Committee have urged regulators to examine whether cross-investments among Nvidia, OpenAI, Microsoft, Amazon, Google, and Anthropic constitute concentration risk for the broader economy. The Open Markets Institute, a nonprofit focused on monopoly policy, has argued that the structure of these deals forecloses competition from smaller labs that cannot access similar credit.

Consumer advocates note that Anthropic is unprofitable and that lenders are effectively betting on continued hypergrowth in enterprise AI spending. The Financial Times has reported that several bond investors have begun pushing back on AI-linked debt pricing, warning of thin covenants.

What the experts say

The Bank for International Settlements, in its September 2026 quarterly review, flagged rising concentration in AI-related corporate debt and warned that a small number of counterparties now account for a growing share of new issuance in the technology sector. The BIS did not single out specific firms but cited the risk of “correlated exposures” if AI revenue projections disappoint.

Researchers at the Stanford Institute for Human-Centered AI estimated in their 2026 AI Index that private investment in generative AI reached about $56 billion globally in 2024, with the United States accounting for the majority. Capital expenditure by the top four U.S. hyperscalers is on pace to exceed $300 billion in 2025, according to filings compiled by Bloomberg Intelligence.

Historian of technology Margaret O’Mara of the University of Washington has compared the current spending cycle to the 1990s fiber-optic buildout, noting that overcapacity ultimately produced cheap bandwidth but wiped out many of the companies that financed it. Economists at the Federal Reserve Bank of Dallas have cautioned that AI-driven capex is now a measurable contributor to U.S. GDP growth, making a downturn in the sector a macroeconomic concern.

By the Numbers

$60 billion: approximate total debt Broadcom is seeking to raise to fund custom AI chips for Anthropic, per Bloomberg.

$42 billion: reported size of Broadcom’s loan commitment to Anthropic for infrastructure spending, per Yahoo Finance.

$56 billion: global private investment in generative AI in 2024, per the Stanford AI Index 2026.

$300 billion+: projected 2025 capital expenditure by the top four U.S. hyperscalers, per Bloomberg Intelligence.

2021: year Anthropic was founded by former OpenAI researchers, per Reuters.

September 2026: date of the BIS quarterly review warning on concentrated AI-linked debt.