- Scale of the raise. Broadcom is raising more than $60 billion in debt — and potentially up to $100 billion including junior financing — for AI infrastructure projects, primarily involving Anthropic, according to reporting published August 21. The figure would represent one of the largest single-company debt transactions in corporate history.
- Structural shift. The move signals that chipmakers and infrastructure providers are no longer financing AI build-out from corporate balance sheets alone; instead, capital markets — institutional debt buyers, credit funds, bond markets — are absorbing the risk and scale of compute expansion.
- Parallel deal. On the same day, Nvidia structured an unusual $6 billion licensing agreement with AI coding startup Poolside, combined with a $1 billion equity investment at a $12 billion pre-money valuation, reflecting a different model for acquiring AI capability without a formal M&A transaction.
The AI infrastructure financing story took on new dimensions this week as two separate transactions illustrated how the industry is evolving its approach to scale-out capital. The larger of the two involves Broadcom, the semiconductor and infrastructure software company, which is in discussions to raise more than $60 billion in debt — with the figure potentially expanding to $100 billion when junior financing tranches are included — according to reporting published by Tech Startups on August 21, 2026.
Broadcom’s Debt Raise: Scale and Purpose
The debt raise is described as primarily supporting AI infrastructure projects that involve Anthropic, the AI safety company whose models are increasingly central to enterprise AI deployments. Broadcom has developed a significant custom AI chip business, designing application-specific integrated circuits (ASICs) for hyperscalers and AI labs, making it a natural counterparty for infrastructure financing connected to AI compute demand.
The scale of the proposed raise — $60 billion at minimum — represents a meaningful shift in how AI infrastructure is being financed. Earlier in 2026, Nvidia committed to guarantee $105 billion for OpenAI’s 8-gigawatt Ohio data centre through a consortium of Wall Street firms, establishing a precedent for off-balance-sheet AI infrastructure financing. Broadcom’s reported debt raise extends that model: rather than each company funding its AI obligations from retained earnings or equity raises, the industry is drawing on the depth of credit markets to fund infrastructure at a speed and scale that internal capital generation cannot match.
Nvidia’s Poolside Licensing Model
The same week produced a separate deal that illustrates a different form of AI capital deployment. Nvidia structured a $6 billion licensing agreement with Poolside, an AI coding startup, alongside a $1 billion equity investment at a $12 billion pre-money valuation. Approximately 109 Poolside employees received Nvidia job offers as part of the arrangement.
The structure is notable because it bypasses formal acquisition — which would trigger antitrust review — while achieving many of the same outcomes: access to Poolside’s models, talent, and intellectual property, combined with a financial stake in the company’s upside. Legal and technology analysts are likely to scrutinise whether this model, which combines licensing, investment, and talent acquisition without a change of control, will attract the same regulatory attention as an outright merger. The arrangement echoes Microsoft’s relationship with OpenAI in its early stages, which was structured as a licensing and investment partnership before evolving into a deeper integration.
The Broader Financing Trend
Together, the Broadcom debt raise and the Nvidia-Poolside structure point to an AI industry that has outgrown its early self-financing model. The compute requirements for frontier AI training runs, and the infrastructure needed to serve inference at scale, require capital commitments that no individual company — even one valued in the trillions — can absorb without drawing on external markets. Intel’s $15 billion public share sale earlier this month signalled that even legacy chipmakers are tapping equity markets for AI-era capital. The question now is whether the pace of capital commitment across the sector is running ahead of the revenue base that will eventually service it.