Broadcom Is a Verified AI Powerhouse, but Is Risk Rising, Too?

Broadcom (NASDAQ:AVGO) has transformed from a smartphone chip stalwart to a formidable AI powerhouse rivaling giants like Nvidia (NASDAQ:NVDA). Its pivot to AI infrastructure — custom accelerators (XPUs), Ethernet networking, and data center solutions — has fueled explosive growth. 

A landmark deal with OpenAI this week underscores this shift, with Broadcom supplying 10 gigawatts of custom AI chips for GPT and o1 models starting in 2026. This follows partnerships with hyperscalers like Google, Meta Platforms (NASDAQ:META), and Amazon (NASDAQ:AMZN). 

However, as Broadcom’s AI business surges, it’s narrowing its focus to an elite cadre of top-tier large language model (LLM) developers and hyperscalers who are driving 40% to 50% of its revenue. This concentration amplifies its AI credentials but raises questions about heightened risk if the AI boom falters. By limiting its customer base, is AVGO stock a riskier investment?

The OpenAI Deal Is a Game-Changer

Broadcom’s $10 billion-plus agreement with OpenAI is a cornerstone of its AI strategy. Unlike the vendor-financing deals with Nvidia, Advanced Micro Devices (NASDAQ:AMD), or Oracle (NYSE:ORCL) where equity stakes or investments grease the wheels, this is a cash-based supply contract. Broadcom will deliver 10 gigawatts of custom-designed AI accelerators — tailored for OpenAI’s training and inference workloads — alongside Ethernet networking and full rack systems, starting in the second half of 2026 through 2029.

The deal builds on 18 months of co-development, positioning Broadcom as a critical enabler of OpenAI’s frontier models. This diversifies OpenAI’s supply chain away from Nvidia, while Broadcom gains a high-profile client to showcase its custom silicon prowess. The deal’s scale is massive, contributing to Broadcom’s $110 billion order backlog, with analysts estimating it could add $2 billion to $3 billion annually to revenue by 2027. 

Heightened Risk Component of Interconnected Partners

Yet the deal exposes AVGO to some “contagion risk.” OpenAI has been locking up deals to accelerate its growth, but while its revenue is growing, it doesn’t have the means of paying for its purchases yet. 

In fact, OpenAI agreements with Nvidia, AMD, Oracle, CoreWeave (NASDAQ:CRWV), and now Broadcom mean its total commitments exceed $1 trillion across vendors. Should AI demand become saturated due to a lack of return on investment, there could be a very fast unwinding of these stocks. Shadows of dot-com era implosions and ghosts from similar failed circular deals of the past hang over the AI boom.

However, Broadcom is better protected than OpenAI’s other partners because it doesn’t have a financing entanglement or equity component to its deal. Still, if deployment timelines slip or AI demand cools, there could be a hit to future revenue expectations.

Broadcom’s Focus on the AI Elite

At the Goldman Sachs Communacopia + Technology Conference last month, CEO Hock Tan emphasized Broadcom’s “very narrow” focus on four elite LLM developers and hyperscalers. Although AVGO doesn’t identify its customers, they are building “superintelligence” platforms with million-XPU clusters, which limits the names to only a handful of potential companies. Google, Meta, Amazon, and TikTok owner ByteDance are often mentioned as likely candidates.

They drive 75% to 90% of Broadcom’s AI revenue — projected at $20 billion this fiscal year, and $32 billion to $40 billion in FY2026. This concentration, up from 30% of revenue in AVGO’s last fiscal year, heightens customer concentration risk. Losing one client, like OpenAI shifting back to Nvidia, could dent growth by 10% to 15%. Revenue projections show these four contributing $50 billion to $80 billion by FY2027, half of Broadcom’s total addressable AI market. 

To mitigate the risks, Broadcom is expanding to two additional hyperscalers and LLM prospects like Anthropic, potentially adding $20 billion to $30 billion in revenue by 2027. Sticky co-development contracts and Ethernet’s open-standard dominance reduce switching risks, while VMware’s $7 billion to $8 billion software revenue provides a non-hardware buffer.

Key Takeaway

Broadcom’s AI pivot has propelled its stock up 51% year-to-date and 93% over the past year, reflecting its status as an AI leader. However, its reliance on four hyperscalers — driving nearly half its revenue — creates real concentration risk. Hyperscalers are facing a massive debt load to finance their AI ambitions, and even a 10% capex cut by any of them could shave $2 billion to $3 billion off earnings, and a broader AI slowdown would hit hard, given flat non-AI segments. 

Yet, Broadcom’s runway remains compelling: a $110 billion backlog, multi-year contracts, and expansion to new clients diversify its base. Its valuation assumes AI’s durability, but sticky partnerships, a lack of entangling partnerships, and software revenue mitigate downside. 

For investors, the stock isn’t a screaming buy but offers balanced risk-reward for those bullish on AI’s long-term growth. If hyperscalers keep spending as they are, Broadcom’s upside outweighs near-term risks, but caution is warranted if the AI “music” slows.