John Ballard, The Motley Fool
5 min read
In This Article:
-
Nvidia is a monster chip stock, but it's not capturing the insatiable demand for custom AI accelerators.
-
Broadcom expects demand for its custom AI chips to accelerate through 2026.
-
Taiwan Semiconductor shares offer solid value relative to the company's earnings growth prospects.
Nvidia (NASDAQ: NVDA) continues to dominate the lion's share of the chips going into data centers for artificial intelligence (AI). The company just reported another monster quarter of growth, and the stock is closing in on new all-time highs.
There's still a case for buying Nvidia stock, but investors shouldn't put all their chips in one basket (no pun intended). Nvidia is not capturing all the demand for AI semiconductors. Hyperscalers like Alphabet's Google and Amazon are designing custom chips for specific AI workloads in their cloud services, and this is creating tremendous growth for other leading semiconductor companies.
Shares of Broadcom (NASDAQ: AVGO) and Taiwan Semiconductor Manufacturing (NYSE: TSM) recently were up 73% and 29%, respectively -- both outperforming Nvidia's 18% return over the last year. These chip stocks can broaden your exposure to the growing demand for AI chips across the data center market.
Broadcom is a top supplier of custom AI accelerators (XPUs) and networking solutions for data centers and other markets. Its networking business supplies components that help move data at high speeds, which is vital as hyperscalers shift to more advanced AI workloads.
Broadcom has reported five consecutive quarters of 20% or more revenue growth. Specifically, AI-related revenue jumped 46% year over year in the most recent quarter. Broadcom is turning this growth into strong profits, with free cash flow reaching $6.4 billion last quarter, representing a high 43% margin on revenue.
AI networking revenue, including Broadcom's Ethernet networking products, grew 170% year over year, representing nearly half of its total revenue from AI. This incredible growth in networking signals a massive ramp in computing power to create the next wave of AI applications and services.
Broadcom is also supplying XPUs, which are more cost-efficient for specific workloads than Nvidia's general-purpose chips. Broadcom sees at least three customers deploying 1 million custom AI-accelerated clusters by 2027, and it reported on the last earnings call that these large hyperscalers are "unwavering" in their plans to continue investing in the near term.
