Google’s software engineers say that if money were no object, custom chips like those made by Broadcom would be the default choice for building cutting-edge artificial intelligence models. “When we don’t have resource constraints and need to move quickly and get as much training as possible, we 100% use some kind of custom chip,” said Gabriel Raskin of the Gemini AI team. In an interview with CNBC, he emphasized that “every second of computing matters.” The idea seems to be growing among hyperscalers as the desire for custom chips grows, with Broadcom leading the way. Google successfully trained Gemini 3 using a tensor processor unit (TPU) co-designed with Broadcom. The November launch of Gemini 3 put the Alphabet unit back on the map as the large-scale language model to beat. The TPU’s performance also makes Google a hot topic as an AI chip to replace the industry-standard Nvidia GPU (graphics processing unit). Every second of computing matters. Gabriel Raskin, Google Gemini Software Engineer Custom chips are designed for specific high-volume tasks, something Nvidia’s general-purpose GPUs aren’t built for. However, NVIDIA CEO Jensen Huang recently dismissed custom chips as a threat to the company’s business. “What Nvidia is doing is much more versatile,” he told Jim Cramer in an interview last month. “NVIDIA can serve a much broader market than just chatbots.” Back in November, Nvidia said in a post about X that it was “delighted with Google’s success,” but the post also said, “Nvidia is a generation ahead of the industry. We are the only platform that can run any AI model and run it anywhere computing happens.” To be sure, TPUs and other competing custom chips (also known as application-specific integrated circuits, or ASICs) won’t replace Nvidia. Google is also one of Nvidia’s biggest customers, relying on vast numbers of GPUs to support its cloud infrastructure and provide customers with the computing power to train and run their own AI models. Google, which has been developing TPUs for more than a decade, is also offering use of TPUs to its cloud customers. Google Cloud is the third largest after Amazon and Microsoft. While these companies are also big customers of Nvidia, they are also introducing their own custom chips. Although industry experts largely agree that the risks to NVIDIA are minimal at this point, some believe that NVIDIA’s dominance in the AI chip market is being tested. “In every market, you’re going to lose some market share,” DA Davidson analyst Gil Luria told CNBC. “This is a free market, and profits breed competition. That’s what’s happening here. Nvidia’s customers, especially the biggest customers like Amazon, Google, Microsoft, and Meta, don’t want to be tied down to one vendor,” Luria said. Broadcom’s October custom chip deal with ChatGPT developer OpenAI underscores the trend toward diversification. Despite Broadcom’s traction, Luria said GPUs remain essential. “Even for Google, TPUs are just ancillary. They’ve historically used TPUs primarily for internal purposes. Now they’re starting to offer them externally, even selling them. But they’re still mostly buying Nvidia.”An Apple official familiar with the complexities of chip development pointed out that the barriers to ASIC production are high, which works to Nvidia’s advantage. Pursuing customer chips can be too time-consuming and costly for small companies. Another constraint is chip manufacturers’ limited access to manufacturing capacity. Keep in mind that Broadcom and Nvidia design chips, but are considered fabless. This means that the actual hardware manufacturing is outsourced to fabs like Taiwan Semiconductor Corporation (TSMC). Apple has been developing its own chips since 2010, gradually replacing third-party silicon in its own devices. “Broadcom is a top performer in the custom chip space,” Luria said. Broadcom CEO Hock Tan said in an earnings call last month that the company’s AI sales “up 65% year over year to $20 billion, and the company’s semiconductor sales reached a record $37 billion for the year.” The CEO also confirmed the last time we heard rumors that the fourth customer with a $10 billion order was indeed Anthropic. “But unlike Nvidia, which we can reasonably assume will continue to be a dominant force, Broadcom’s position is more vulnerable,” Luria explained. “The company’s biggest customer so far is Google. If Google were to decide to go directly into TSMC like Apple did, that would be a much bigger risk to Broadcom than what Nvidia faces.” Luria said he expects Nvidia to maintain more than 50% market share for at least the next five years and more than 70% for the next three years. Analysts at DA Davidson rate Nvidia a Buy, with a price target of $250 per share. This represents an increase of nearly 31% from Thursday’s closing price. Others on Wall Street are also watching closely. Morgan Stanley this week reissued Buy ratings on both Broadcom and Nvidia, noting that it “favors Nvidia despite growing enthusiasm for ASICs.” Analysts named the Vera Rubin platform the solution with the industry’s highest return on investment in cloud computing as it becomes fully operational in the second half of 2026. However, Wolf analysts were somewhat bullish on Broadcom. “Google’s willingness to provide TPUs to third parties makes it a real competitor for Nvidia,” the analysts wrote, adding that they expect Broadcom to be a major beneficiary. The company, which upgraded Broadcom to a buy-equivalent rating on Friday, forecasts shipments of about 7 million TPUs by 2028. Analysts have a $400 price target for Broadcom stock, implying a nearly 21% upside from Thursday’s closing price. On Friday’s “Squawk on the Street,” Jim Cramer said the upgrade was a “timely technical recommendation” for investors looking for quality stocks, which are down nearly 20% from their all-time high closing price of about $413 and down 4% year-to-date. “Broadcom’s poor behavior is incomprehensible,” Jim said at the CNBC Investment Club’s January monthly meeting. Mr. Zim’s tendency to buy on the dip mirrors the stance he took when Broadcom fell 11.4% in December after posting a strong quarterly beat that was obscured by misinterpreted comments. AVGO NVDA 5Y Mountain Nvidia vs. Broadcom 5-Year Performance Nvidia stock is a little higher since the beginning of the year, but it has its own problems. Stocks are battling multiple compressions as investors become less willing to pay a premium for every dollar of profit. Continued geopolitical tensions between the US and China are also weighing on stock prices. Reuters reported this week that China has approved the purchase of Nvidia’s H200 chips by ByteDance, Alibaba and Tencent. Jim reiterated his policy that the stock “is mine and I won’t trade it,” but warned that people should be patient until the deal is completed. According to Jim, Jensen’s presentation of the new Vera Rubin chip at Nvidia’s annual GTC conference in mid-March could help move the stock in a positive direction. The club has hold-equivalent ratings of 2 on both Broadcom and Nvidia, and recommends adding to positions on significant declines. Broadcom’s price target is $425 and Nvidia PT’s $230. (Jim Cramer’s Charitable Trust is long GOOGL, AVGO, NVDA, AMZN, MSFT. See here for a complete list of stocks.) As a subscriber to Jim Cramer’s CNBC Investing Club, you will receive trade alerts from Jim Cramer before he makes a trade. After Jim sends a trade alert, he waits 45 minutes before buying or selling stocks in his charitable trust’s portfolio. If Jim talks about a stock on CNBC TV, he will issue a trade alert and then wait 72 hours before executing the trade. The above investment club information is subject to our Terms of Use and Privacy Policy, along with our disclaimer. No fiduciary duties or obligations exist or arise from your receipt of information provided in connection with the Investment Club. No specific results or benefits are guaranteed.
