The artificial intelligence (AI) era continues to gain momentum. Total capital expenditures among top AI companies could hit $700 billion this year. All these investments have driven a chip supercycle that has propelled Nvidia (NVDA +0.10%), the runaway leader in AI GPU chips, to mind-blowing growth over the past several years.
But the AI industry is already changing, and quickly. Demand for AI inference is surging as companies begin applying AI models to real-world applications, which emphasizes chip efficiency over raw performance. Broadcom (AVGO 0.12%) has emerged as a power player in this space, designing custom AI accelerator chips for multiple top AI companies.
Both stocks have been absolute home runs since early 2023, but have pulled back from their highs amid recent market volatility. Which AI chip titan is most likely to add to your wealth in this ongoing supercycle?
Image source: The Motley Fool.
Nvidia is gearing up for inference with Vera Rubin
One major advantage Nvidia carries is a massive existing installation base. The top AI companies have already committed hundreds of billions of dollars to building on Nvidia's hardware and its proprietary CUDA parallel computing platform, which helps lock developers into its ecosystem. Nvidia's current flagship AI chip, Blackwell, is still going strong.
NASDAQ: NVDA
Key Data Points
Nvidia will soon go after more of the server rack with Vera Rubin. It's not just a GPU; it's a full-rack server platform with a CPU and networking hardware. Additionally, Nvidia designed Vera Rubin for inference applications, such as AI agents, which could be the largest immediate AI opportunity today, given how enterprises can deploy what are essentially digital employees.
CEO Jensen Huang has said that Nvidia, currently at $215.9 billion in revenue over the past four quarters, has a line of sight to more than $1 trillion in sales from Blackwell and Vera Rubin through 2027, noting that agentic AI is accelerating demand.
Broadcom's massive ASIC ramp-up
Obviously, AI companies would prefer not to give Nvidia so much supply chain leverage. Custom accelerator chips have become a rapidly growing AI chip subcategory that Broadcom has increasingly embraced. An ASIC is a chip designed to perform one specific application extremely well. AI companies are seeking efficiency from ASICs purpose-built for their specific needs.
NASDAQ: AVGO
Key Data Points
Broadcom has inked partnerships with six major AI companies to design chip road maps for their specific AI needs. Alphabet worked with Broadcom to co-develop its Tensor Processing Unit, used to train its Gemini 3 AI model. That success has been a boon for Broadcom, which has since scored partnerships with Anthropic, OpenAI, Meta Platforms, and others.
CEO Hock Tan forecasts Broadcom's AI chip sales will surge significantly over $100 billion in fiscal year 2027. The company's total semiconductor solutions sales were $36.8 billion in fiscal year 2025. That's companywide chip sales, not just AI.
Which stock do you want to own?
As stocks, Nvidia and Broadcom are quite similar. They trade at similar valuations, ranging from 19 to 21 times trailing-12-month sales.
Nvidia's business is almost entirely focused on AI data center chips at this point, which gives it an edge in growth, but also makes it riskier. If the AI supercycle ends, Nvidia may actually have more downside than Broadcom, given the latter's non-AI revenue streams.
NVDA PS Ratio data by YCharts
Broadcom also pays a meaningful dividend. Sure, it yields just 0.86%, but the company is committed to growing it. Broadcom has increased the dividend for 14 consecutive years. For Nvidia, the stock's 0.02% yield hardly qualifies as a dividend talking point.
Nvidia may be the better stock to buy if you're convinced that the AI supercycle will last for the foreseeable future. But those looking for a bit more safety without sacrificing much upside will probably find Broadcom more appealing for their wealth-building efforts.
