Bernstein analysts are still wildly bullish on two AI semiconductor powerhouses, despite calls to slow down the development of the most advanced models and rekindled investor enthusiasm for a competing part of the AI chip ecosystem. "We would still own both Nvidia and Broadcom ," Bernstein wrote to clients in a note on Tuesday. "Both appear to have line of sight for strong growth even in a constrained environment, and both are extremely cheap." The semiconductor research team, led by closely followed chip analyst Stacy Rasgon, has buy-equivalent outperform ratings on both stocks, with a price target of $575 per share on Broadcom (nearly 58% upside to Monday's $363 close) and a $400 price target on Nvidia (upside of $76% versus Monday's $227 close). "Extremely cheap" price-to-earnings ratios almost feels like an understatement. Nvidia trades at just 17 times next 12 months' earnings per share (EPS) estimates versus the 5-year average of 35 times, according to FactSet. Broadcom trades at 19.7 times forward EPS estimates versus the 5-year average of 23 times. Not too long ago, it was unheard of for companies with these kinds of growth rates to only garner a broader market multiple. The forward P/E on the S & P 500 is 19.56 times. Rasgon said that such inexpensive valuations could provide some cushion if AI sentiment does weaken —but notably, his team does not think that will happen. In fact, they pointed to multiyear visibility, as well as tight supply and capacity across virtually all of the components needed for the AI buildout. They noted, "NVDA suggesting 70%+ growth off of a massive number into 2027, and AVGO suggesting potential to double revenues in 2027 and again in 2028 or even more." (NVDA is Nvidia's ticker symbol, and AVGO is Broadcom's.) On the chip side of the AI infrastructure boom, Nvidia does it all now, including designing central processing units (CPUs), but the company's graphics processing units (GPUs) have long been the industry gold standard in training AI and running complex models. Broadcom makes custom chips for specific AI workloads, and counts Alphabet 's Google, Meta Platforms , Anthropic, and OpenAI. Both Nvidia's GPUs and Broadcom's custom chips fall into the category of AI accelerators. The GPU-to-CPU ratio has been narrowing due to the rise of more advanced AI systems capable of handling tasks autonomously, known as agentic AI. The way these systems operate requires CPUs and GPUs working in close coordination. But for certain parts of agentic tasks, CPUs are more efficient than their GPU counterparts. The launch of Meta's new Muse app to build agents has investors piling back into CPU stocks this week — reminiscent of the type of trading seen this spring when agentic AI adoption began to take off. We first bought Arm to play the CPU renaissance, but later switched our preference to Intel , which remains one of Jim's favorite ways to play the AI compute boom. In addition to its CPU business, Intel's third-party manufacturing and packaging business sweetens the bull case. Jim shares Rasgon's views about AI demand, and he made that clear last week when the debate was raging over Anthropic CEO Dario Amodei's essay, arguing that companies need to collectively pace AI models in the name of safety and the government needs to step in with regulations to keep everyone on the same page. Last week, Jim interviewed Nvidia CEO Jensen Huang and Broadcom CEO Hock Tan about Amodei's white paper, and both executives said that they don't see demand slowing. Tan said on "Mad Money" that he is not changing his forecasts. Huang told Jim that AI companies can take as much time as they need to put out safe products, but there is no need for a coordinated effort or government intervention. Rasgon's team also spoke last week with a host of chip companies, including Nvidia and Broadcom, as well as Intel. The investment implications from those meetings, according to Bernstein, were that Nvidia's data center opportunity "remains enormous;" Broadcom's "AI trajectory seems set to markedly accelerate" in the next two years; and for Intel, "server strength is helping the company get back on their feet, and narrative/headlines may fuel the vibe for now." Rasgon turning even a bit more positive on Intel is significant since he has been down Intel for years. Bottom line While favoring Intel, and even Micron on the memory side of the chip complex, over Nvidia, Jim still sees Nvidia as a 1-rated buy, with a $280 price target. Jim is more on the fence on Broadcom, with a hold-equivalent 2 rating and a $430 price target. There is room in both of those targets for upside. However, we cut in half our position in Broadcom in late August to reduce the portfolio's exposure to the AI data center theme in response to growing political pressure. It was a good call, as we were able to book a big profit and Broadcom stock has been fairly flat since. (Jim Cramer's Charitable Trust is long NVDA, AVGO, INTC, GOOGL, META. See here for a full list of the stocks.) As a subscriber to the CNBC Investing Club with Jim Cramer, you will receive a trade alert before Jim makes a trade. Jim waits 45 minutes after sending a trade alert before buying or selling a stock in his charitable trust's portfolio. If Jim has talked about a stock on CNBC TV, he waits 72 hours after issuing the trade alert before executing the trade. THE ABOVE INVESTING CLUB INFORMATION IS SUBJECT TO OUR TERMS AND CONDITIONS AND PRIVACY POLICY , TOGETHER WITH OUR DISCLAIMER . NO FIDUCIARY OBLIGATION OR DUTY EXISTS, OR IS CREATED, BY VIRTUE OF YOUR RECEIPT OF ANY INFORMATION PROVIDED IN CONNECTION WITH THE INVESTING CLUB. NO SPECIFIC OUTCOME OR PROFIT IS GUARANTEED.
