Nvidia, Arm or Broadcom: The best AI Semiconductor stock to buy now, according to Wall Street

Artificial intelligence (AI) will eventually reshape most industries either by eliminating the need for human workers or by increasing worker productivity. Analysts generally agree that the transformation will occur in stages. They define and name those phases differently, but infrastructure companies are always the first to benefit.

Training machine learning models and running inferences on AI applications is dependent on powerful semiconductors and networking hardware. As a result, chip makers Nvidia (NVDA -2.25%), Arm holds (ARM -4.96%)and Broadcom (AVGO -4.05%) have already benefited from AI adoption. Year to date, Nvidia stock is up 160%, Arm stock is up 80%, and Broadcom stock is up 49%. Meanwhile, the S&P 500 (^GSPC -0.32%) advanced by 18%.

However, Wall Street believes the ranking will look different over the next 12 months. Listed below are the average price targets (and implied upside) for all three stocks as of August 25.

• Nvidia: $144 per share (with 11% default)
• Arm: $135 per share (1% default downside)
• Broadcom: $196 per share (with 18% implied upside)

Wall Street actually sees a downside to Arm’s stock. And while analysts expect the other two semiconductor stocks to produce positive returns, Broadcom’s median price target implies more upside.

1. Nvidia

Nvidia graphics processing units (GPUs) are the industry standard in accelerating complex data center workloads such as scientific computing and artificial intelligence (AI) applications. “Nvidia’s chips underpin all of the most advanced AI systems, giving the company an estimated market share of more than 80 percent,” according to The Wall Street Journal.

Nvidia also has monetization opportunities beyond GPUs. The company recently introduced its first data center server chip, called Grace, and CEO Jensen Huang says it’s headed for a multibillion-dollar product line. Nvidia also sells InfiniBand and Ethernet networking platforms. That part of its business recently reached a revenue rate exceeding $12 billion annually.

Finally, Nvidia has further expanded its ability to monetize AI by delving into software and services. Nvidia AI Enterprise is a subscription software suite that simplifies the development and deployment of AI applications. And DGX Cloud brings together Nvidia hardware and software to create a complete AI-as-a-service offering. Collectively, Nvidia’s software and services segment recently hit $1 billion in revenue.

Looking ahead, Wall Street expects the company’s non-GAAP earnings to grow 39% annually through fiscal 2027 (ending January 2027). Compared to that estimate, its current valuation of 71.7 times adjusted earnings falls somewhere between cheap and expensive.

2. Supporting the arms

Arm designs central processing unit (CPU) architectures and development tools and licenses intellectual property (IP) to customers. Its clients include the five largest companies in the world. Arm IP is the basis for AppleIts Private Cloud Compute, Nvidia’s Grace processors, Microsofthis cobalt chips, AlphabetAxion’s processors and Amazonhis Graviton processors.

This range is particularly interesting because those chips are used in data centers, a market historically dominated by Intel and AMD. To elaborate, Arm chips have traditionally been limited to mobile devices, as x86 processors from Intel and AMD offered superior computing performance. Companies on both sides of the architectural divide tried to take market share from the others, but Arm was more successful.

The company has defended its dominance in smartphones and other mobile devices while gaining share in data centers by improving the processing power of its chips, particularly in AI. That means Arm is well-positioned to monetize AI on two fronts: personal devices like Apple’s iPhones and Microsoft’s Copilot+ PCs, as well as infrastructure and platform services offered by public cloud providers like Amazon and Alphabet.

The only problem with Arm is the rating. Wall Street expects the company to grow adjusted earnings 25% annually through fiscal 2026 (ending March 2026). That consensus estimate makes the current valuation of 96 times adjusted earnings look outrageously expensive.

3. Broadcom

Broadcom offers IT solutions in two broad categories: infrastructure software and semiconductors. To elaborate, the company sells virtualization, mainframe and cybersecurity software that helps companies manage, optimize and protect their IT infrastructure. Of the three software verticals, virtualization is the most central to Broadcom’s growth narrative. Its subsidiary VMware is the leader in server virtualization, and the market is expected to grow by 18% annually through 2032.

However, Broadcom’s biggest opportunities lie in semiconductor solutions, as that segment will benefit from demand for AI infrastructure. Broadcom is the market leader in data center networking chips and application specific integrated circuits (ASICs). Argus analysts Jim Kelleher recently wrote: “The business most affected by AI is networking, where Broadcom holds the long-standing market lead in enterprise data centers and cloud.” But that could change in the coming years.

Broadcom is also a market leader in next-generation ASICs, custom silicon built for specialized use cases such as artificial intelligence. The company already designs AI accelerators for Google and Meta platformsand recently won a third major client. Broadcom announced the deal in March without naming the client, but Reuters identified the company as TikTok parent ByteDance.

Today, custom AI accelerators account for less than 10% of AI processors, as the market is dominated by GPUs (more specifically, Nvidia GPUs). But Morgan Stanley Analysts believe custom AI chip sales will grow faster than GPU sales over the next few years, so that ASICs account for up to 30% of AI chips by 2027. In this scenario, Broadcom would likely be the biggest winner , as it controls 55% to 60% of the market, according to analysts.

Wall Street expects the company’s earnings to grow 24% annually through 2025. That makes the current valuation of 38 times earnings seem reasonable. Now is a good time for patient investors to buy a small position in this semiconductor stock.