I keep hitting the buy button on Broadcom (NASDAQ:AVGO | AVGO Price Prediction) because I do not have to be right about which AI chip architecture wins. I only have to be right that AI itself keeps getting built. That is the trade I want in a retirement account, and the numbers keep confirming it.
Here is the shape of my conviction in plain terms. If hyperscalers keep buying merchant GPUs, they still need the switching fabric to lash those clusters together, and Broadcom’s Tomahawk and Trident silicon control over 70% of high-end data center switching. If they pivot to their own custom accelerators, Broadcom is the co-design partner behind Google’s TPU, Meta’s MTIA, and upcoming OpenAI custom chips. Either road leads through the same tollbooth. That is why I am not agonizing over whether NVIDIA (NASDAQ:NVDA) or somebody else takes the GPU crown next year.
The Data That Keeps Me Adding
AI semiconductor revenue is compounding. Q3 FY2025 AI revenue was $5.20B (up 63% YoY), then $6.20B (up 74%), $8.40B (up 106%), and $10.80B (up 143%) in Q2 FY2026. Management guided Q3 FY2026 AI semiconductor revenue to $16.0 billion, over 200% YoY, with total revenue of approximately $29.4 billion, up 84% YoY. CEO Hock Tan said flatly on the call, “For the full year 2026, we expect to achieve AI semiconductor revenue of $56 billion, up approximately 180% from fiscal 2025,” and reiterated “in excess of $100 billion in 2027.”
The cash economics behind that growth are what let me sleep. Q2 FY2026 free cash flow landed at $10.26 billion, 46% of revenue, with an adjusted EBITDA margin of 69%. Operating income rose 85.07% YoY on 48% revenue growth, which is textbook operating leverage. Cash on hand climbed to $19.63 billion, up 107.22% YoY, while total liabilities fell 3.76%.
Then there is the shareholder return record retirees actually care about. Broadcom has raised its dividend for 15 consecutive years since fiscal 2011, with the latest bump a 10% increase to $0.65 quarterly. On top of that, a $10 billion buyback runs through December 31, 2026, with $7.8 billion already executed in Q1 FY2026.
Why Not Just Buy NVIDIA
NVIDIA is the reflex pick, and I own some. My issue is concentration of outcome. NVIDIA’s business rides on merchant GPUs staying the preferred training substrate. Broadcom’s Q2 bookings tell a different story: AI semiconductor bookings of over $30 billion against the $10.8 billion shipped, with Tan noting “our visibility now extends into 2028.” Add multi-generational commitments with Google, Anthropic, OpenAI, and Meta, plus a VMware subscription base doing $7.18B in Q2 revenue at 79% operating margin, and I get AI upside plus recurring software cash flow in the same holding.
The Risk I Refuse to Wave Away
Customer concentration is real. A handful of hyperscalers drive the AI number. If one pauses capex, the growth curve bends. What steadies me is that the same customers are ordering years ahead because “they need lead time for wafers, memory, power infrastructure, and facility-level planning.” That reflects structural demand tied to multi-year infrastructure planning.
At $416.08 and a forward P/E of 23, I keep adding because Broadcom does not need me to pick the AI winner. It collects either way.
Contact [email protected] for any questions or corrections.
Top Gaining Stocks
Top Losing Stocks