$AVGO | Broadcom offers a very broad product portfolio, but I am focusing exclusively on the artificial intelligence sector today. The company is a leader in the two most critical areas of AI infrastructure: custom AI chip design and high-speed networking. Its closest competitor, Marvell, is currently trailing far behind; while I don’t foresee that changing, I’ll avoid making any definitive predictions. Broadcom not only develops and manufactures custom chips (ASICs/XPUs) tailored to the specific needs of hyperscalers, but it also provides the infrastructure required to interconnect thousands of these chips. Through these ASICs, the company also drives sales of its other proprietary components.
When it comes to chips, Nvidia is the largest and most powerful player—that is beyond debate. The second-largest player is not AMD, but Broadcom. Keep in mind that this applies specifically to high-end inference chips—I am comparing Nvidia and AMD GPUs against Broadcom’s ASICs. As shown in the visual, Nvidia is the clear leader, followed by ASICs (Broadcom), with AMD in third place. You might point out that Broadcom doesn’t produce *all* AI ASICs—and you would be correct—but it is projected to hold more than 50% of the market by 2030; its current market share is already approaching 70%.
You might then ask: if the company is so exceptional, why has its stock fallen more than 20% from its peak, effectively entering its own bear market? There are several reasons, but in my view, the most significant factor is its attempt to emulate Nvidia. Because Nvidia possesses such robust cash flow, it has been able to offer financing to entities looking to acquire its chips—investing in or providing capital to companies like Anthropic, OpenAI, Coreweave, and Nebius—thereby further anchoring these players within its own ecosystem. Broadcom, noting its own strong cash flow and growth, decided to facilitate financing for the ASIC chips it is developing for Anthropic and OpenAI. For Anthropic, it has initiated a $60 billion funding round, with $42 billion provided by Broadcom itself. It is also seeking $50 billion in financing for OpenAI; however, the arrangement for OpenAI differs from that of Anthropic—in this case, Broadcom acts more as an intermediary between lenders and OpenAI. Other clients for whom Broadcom is developing chips—without participating in their financing—include Google, Meta, ByteDance (TikTok), and two additional, undisclosed companies.
As one might expect, this surge in the client base is driving significant growth. Broadcom is projected to grow by 24% in 2025, 66% in 2026, 64% in 2027, and 56% in 2028. A key question is whether profit margins are keeping pace with this expansion; indeed, they are. Between 2026 and 2028, the EBITDA margin is expected to remain between 67% and 68%, while the free cash flow (FCF) margin is projected to hold between 46% and 50%. Furthermore, earnings per share (EPS) growth—forecast at 94% for 2026, 78% for 2027, and 68% for 2028—is outpacing revenue growth.
Regarding valuation, the P/E ratios stand at 37x, 21x, and 12x, while the PEG ratios are 0.4x, 0.3x, and 0.2x—a PEG multiple even lower than that of Nvidia for the year 2028. Are there risks? Certainly, there are. Should Anthropic and OpenAI fail, it would deal a major blow to Broadcom—that much is clear. This aspect represents a significant risk. Furthermore, while the TPU agreement with Google is set to run until 2031, the question of whether future generations of this technology might shift to competitors like Qualcomm or Marvell is critical; if that were to happen, current workloads might remain stable, but investors could become uneasy regarding future growth. Finally, looking at the technical side, I have identified two potential scenarios: one where the indices enter a bear market, and another where they undergo a correction. This post is for informational purposes only and reflects my own subjective deductions; please conduct your own research and take full responsibility for the decisions you make.
