Broadcom has a huge opportunity in front of it with custom AI chips.

Alphabet’s custom TPUs give it a big structural cost advantage.

TSMC wins with both GPU and ASIC growth.

10 stocks we like better than Broadcom ›

Artificial intelligence (AI) stocks helped lead the market higher in 2025, and so far in 2026, their momentum has continued. Let’s look at three top AI stocks to buy this month.

Image source: Getty Images.

After a strong performance in 2025, Broadcom (NASDAQ: AVGO) looks poised to repeat it this year. It has become the go-to company to help clients design custom ASICs (application-specific integrated circuits) for AI, which are preprogrammed chips designed to handle specific tasks.

As large tech companies look to diversify away from Nvidia’s graphics processing units (GPUs) and save costs with AI workloads, they are increasingly turning to Broadcom, which helped Alphabet (NASDAQ: GOOGL) (NASDAQ: GOOG) design its highly regarded tensor processing units (TPUs).

Customers provide their own chip architecture, and Broadcom gives them the building blocks to design the chips through its intellectual property (IP) portfolio. It is also able to secure capacity and advanced packaging technologies from **Taiwan Semiconductor Manufacturing **(NYSE: TSM), ensuring companies are able to have their chips manufactured at scale.

Custom AI chips are a huge opportunity for Broadcom, with Citigroup analysts predicting that its AI revenue could rise fivefold over the next two years from around $20 billion this past fiscal year to $100 billion is fiscal 2027.

The company furthest along with custom AI chips is Alphabet, which first designed its TPUs over a decade ago. Since then, the company has used these chips to run most of its internal workloads and has integrated them into its hardware and software systems. Its chips are battle-tested and have continued to evolve, giving the company a huge advantage.

Alphabet’s TPUs have allowed it to train its world-class Gemini large language model (LLM) and run inference at a cost cheaper than competitors. Its chips are so well regarded that Anthropic has placed a $21 billion order with Broadcom for its TPUs to run through Google Cloud for some of its AI workloads.

Analysts at Morgan Stanley predict that for every 500,000 TPUs that customers deploy, Alphabet generates around $13 billion in revenue. The analysts are looking for the company to deploy around 5 million TPUs in 2027 and 7 million in 2028. At the same time, Alphabet is aggressively budgeting for capital expenditures to build more data center infrastructure to try to meet the strong demand for its cloud computing services.