NVIDIA (NASDAQ:NVDA | NVDA Price Prediction) and Broadcom (NASDAQ:AVGO) just walked into the same AI capex storm and walked out with very different stories. NVIDIA sells the merchant GPU platform hyperscalers default to. Broadcom sells the custom silicon those same hyperscalers use when they want to design around NVIDIA. This week the question is which model captures more of the buildout.
Blackwell Runs Hot. Custom ASICs Run Hotter Percentage-Wise.
NVIDIA’s Q1 FY2027 delivered $81.615 billion in revenue, up 85.23% year over year, with non-GAAP EPS of $1.87. Data Center hit $75.246 billion, and networking alone jumped 199%. Jensen Huang framed it as “the largest infrastructure expansion in human history.” That is a heavy claim, but supply commitments of $119.0 billion suggest management is putting real money behind it.
Broadcom’s Q2 FY2026 came in at $22.19 billion, up 47.9%, with AI semiconductor revenue of $10.80 billion growing 143%. Hock Tan guided Q3 AI revenue to $16.0 billion, implying growth over 200%. VMware still contributed $7.18 billion in recurring-flavored software revenue, a cushion NVIDIA does not have.
Horizontal Platform Versus Vertical Design Partner
| Lens | NVIDIA | Broadcom |
| Core Bet | Merchant GPUs plus CUDA | Custom XPUs plus Ethernet AI switches |
| Customer Shape | Broad: cloud, sovereign, auto, enterprise | Concentrated hyperscaler ASIC contracts |
| Margin Signal | Non-GAAP gross margin 75.0% | Adjusted EBITDA ~68% |
| Key Vulnerability | Zero China H20 revenue this quarter | Customer concentration and VMware debt |
NVIDIA’s next-gen Vera Rubin platform and partnerships with Google Cloud, T-Mobile, and Hyundai suggest a widening net. Broadcom, by contrast, tightens focus on a handful of hyperscalers building their own accelerators, aiming to exceed $100 billion in AI sales by 2027. Different playbooks, same tailwind.
The Next Test Is Guidance Credibility
NVIDIA guided Q2 to $91.0 billion, excluding any China Data Center compute. That is a huge base to accelerate off. I will be watching whether the ACIE and sovereign AI cohort can offset the China gap Huang has essentially written off. On Broadcom, the $16 billion AI revenue figure becomes the number that matters. Retail sentiment has already grabbed the story: a viral Reddit thread noted “34% of the S&P is 10 stocks making the same bet,” and any wobble in hyperscaler capex would sting both names fast.
Why I Would Split the Question by Investor Type
If you want the margin engine and the cleanest AI proxy, I lean NVIDIA. A 11.04% year-to-date move against a 75.0% gross margin and an $80 billion new buyback authorization is a hard combination to ignore. If you want ASIC exposure with a software cushion and a 20x forward P/E, Broadcom fits better, especially for income-tilted investors watching that $0.65 quarterly dividend. Personally, this is not a week that forces a choice. The cleaner setup is to see NVIDIA’s next quarterly earnings report and whether the capex thesis holds before drawing further conclusions on either name.
Contact [email protected] for any questions or corrections.