Miss Out on Nvidia? Two More Innovative AI Chip Stocks Hiding in Plain Sight VIEW IN BROWSER Tom Yeung here with your Sunday Digest. It would have been easy to overlook Nvidia Corp. (NVDA) in 2022, the year ChatGPT was launched. Chipmakers were notoriously cyclical beasts, and Nvidia had its own history of losing money. Any investors who bought Nvidia in 2021 would have been sitting on 50% losses by the time ChatGPT hit the market. Or in 2000… 2004… 2009… 2011… or 2018. In fact, Nvidia fell at least 50% in 13 of the 26 years since the firm went public. PC demand is extremely lumpy. So, much like autos and airlines, the chipmaking industry was marked by high investment costs, cyclical demand, and a regular parade of high-profile bankruptcies. However, ChatGPT has created a lasting significant change in chip appetite. Unlike regular PC users, data centers that run these AI models demand more… more… more computing power every day and are willing to pay for it. Nvidia’s latest GB200 Blackwell Superchip now sells for up to $70,000 apiece… provided you can get your hands on one. Even its last-generation H100 chips routinely exchange hands for over $20,000 on the secondary market. No PC gamer – Nvidia's pre-2022 main clients – would ever spend those sums on a graphics card for their home computer. In addition, every cloud computing giant has become terrified of falling behind in the AI race. America’s (and China’s) largest tech firms are paying top dollar for the best hardware, pushing prices of AI chips to sustained record prices. That’s caused operating margins at Nvidia to quintuple to 62% from its pre-ChatGPT average. When you’re selling high-end chips for $70,000 that only cost $17,000 to produce, it’s hard not to print money. Analysts are projecting profits to triple by 2028, which could send Nvidia’s justified value per share into the $250 range. (I had previously projected a $160 split-adjusted value by 2027 when the stock was still at $72.) Still, I know that many of you have not yet jumped in on Nvidia. The stock is wildly expensive on a traditional basis, and besides, who wants to buy a stock that’s already risen 1,050% with only another 32% upside to fair value? Any investor over 40 will remember what happened to 1999 tech stock valuations, 2007 Florida real estate prices, and 2014 energy companies. That’s why Luke’s latest presentation is so interesting. On January 27, he unveiled his Genesis Portfolio, an eight-stock portfolio designed to ride the next leg of AI-led innovation even higher. Nvidia might only have 32% upside from here… but Luke's eight picks are only getting started. In fact, some of these are household AI names that are sitting right under Wall Street's noses. You can catch up with that free broadcast here. In the meantime, I’d like to illustrate this potential with two of my own picks that are leading the way in semiconductor innovation. And much like Nvidia at $72, these two have incredible upside that Wall Street hasn’t fully realized yet... The Broadcom Alternative Broadcom Inc. (AVGO) is often talked about as “the next Nvidia.” The high-profile company has become a leader in custom AI accelerator chips, and Wall Street analysts are quick to point out that AI data centers would be impossible to build without Broadcom's networking chips. It doesn’t matter how quickly a GPU can process data if there’s a data bottleneck preventing it from reaching the processor. However, Broadcom’s visibility on Wall Street has pushed shares to unusual heights, limiting future gains. The stock has risen 500% since ChatGPT’s launch, and my models calculate just another 52% upside to $506 if things go the company’s way. You would need cash-flow assumptions that border on fantasy to get much higher. Instead, my award for fantastic potential returns is Marvell Technology Inc. (MRVL), a Broadcom rival that’s starting from a far smaller base. The company trades at just a third of its larger rival on a price-to-sales basis, and my models project a 76% upside to a $147 justified value on relatively conservative assumptions. Using more aggressive figures suggests potential 100% upside. In short, Marvell has assembled a broad portfolio of networking and processing chips. Its optical chips (which use light waves instead of electrical signals) are considered world class, and its processing products are among the best in the business. Microsoft Corp.’s (MSFT) North American data centers currently source 100% of their optical chips from Marvell. The Silicon Valley-based company is also a leader in custom chips – something that put Broadcom on the map. Data centers are increasingly seeking to lower inference costs (the costs associated with running existing AI models) and often build custom chips that are designed specifically for the task. Marvell’s custom silicon has been particularly popular with Amazon.com Inc (AMZN), and management expects the company’s custom chip business to grow 20% next year as more customers sign on. In addition, Marvell plans to acquire Celestial AI for roughly $3.3 billion, which will help the combined firm further expand optical chip products. Together, that means Marvell's profits will grow far faster than Broadcom’s in the coming years. And given Marvell’s discount to its larger rival, it becomes clear which innovator to buy. The AI Chip Builder At this point, virtually everyone knows Taiwan Semiconductor Manufacturing Co. Ltd. (TSM), the company that makes chips for Nvidia, Apple Inc. (AAPL), Broadcom, Intel Corp. (INTC), and more. In fact, the Taiwanese firm is the only company in the world that can consistently produce at the 4-nanometer (nm) node – the microscopic technology that powers every chip in Nvidia’s latest Blackwell series (including that $70,000 GB200 Superchip). It’s a monopoly hiding in plain sight. What most people don’t realize is that TSM shares have risen only 300% since ChatGPT was launched and are only up 180% in the past five years. Shares trade at just 24 times forward earnings, giving TSM a potential 110% upside to $705 in my longer-range models. The investment thesis here is clear: Taiwan Semiconductor makes the world’s best chips, and no company comes close. For instance, the only other firm that can produce 4nm nodes at scale is Samsung Electronics. However, the advanced technology has been a – as one tech writer recently put it – “nightmare for the Korean giant” to implement. According to industry analysts, Samsung barely achieves a 60% yield on 4nm chips, which means 40% of its production must get scrapped. The performance gap is even starker at the more advanced 3-nanometer node. Here, Taiwan Semi achieves a 90% yield, compared to under 50% at Samsung. Alphabet, Advanced Micro Devices Inc. (AMD), and Qualcomm Inc. (QCOM) reportedly switched their 3nm chips to TSM, and other firms are reportedly considering the same. Analysts expect Nvidia’s next-generation “Rubin” GPUs to rely entirely on TSM production. In fact, TSM has already begun volume production of its next generation 2nm technology – years ahead of its closest rivals. These results were on full display during TSM’s latest earnings call on January 15, where financial figures crushed Wall Street forecasts. Management now expects revenue to grow in the mid-20% annually (up from previous forecasts of 20%) through 2029, and that AI revenues should jump 50% annually (up from mid-40%). That’s why it makes little sense for TSM to be valued as a commoditized contract manufacturer. Chip manufacturing has turned into a high-end winner-takes-all market, and investors are belatedly realizing the true value of Taiwan Semi. The Six Sectors of Innovation Last week, I talked about two biotech companies and a drone maker at the cutting edge of American innovation. These companies represent two key areas that the U.S. government has identified as crucial technologies to fund, no matter the cost. Marvell and Taiwan Semi represent another area of intense government investment. Under the Biden administration, TSMC alone received $6.6 billion in grants and $5 billion in low-cost loans to build three chip factories in Arizona. It might receive even more funding indirectly under the Trump administration through Project Stargate, a $500 billion project aimed at building American AI infrastructure. Luke believes this is only the start. In his latest presentation, he outlines six core sectors that should receive enormous amounts of government funding in the coming years as America races to stay ahead: - Artificial Intelligence
- Quantum Computing
- Nuclear Energy
- Biotech
- Semiconductors
- Advanced Manufacturing
And to capitalize on this surge, Luke has created what he’s called the Genesis Portfolio – a group of eight elite companies that focus on these sectors. He outlines all this and more in that free broadcast. Check it out. Until next week, Thomas Yeung, CFA
Market Analyst, InvestorPlace | 
No comments:
Post a Comment