What's the secret to great stock-picking? Obviously, discipline and patience are important. Perhaps the most important trait, however, is also the one that's most difficult to develop. That's a constant awareness of how industries are changing before those changes become evident.
Take Amazon as an example. It's obvious now, but it wasn't obvious back in the 1990s -- when the internet itself was still relatively new -- that a punchy little online bookseller would eventually become the e-commerce titan it is today. Only a small handful of investors were actually in on the ground-floor opportunity when the company went public back in 1997.
All industries are constantly evolving, of course, sometimes dethroning their leaders, and sometimes not. Artificial intelligence (AI) is no exception to this dynamic. In fact, there's one AI trend underway right now that could prove incredibly disruptive to the current norm, turning thousands of forward-looking investors into millionaires in the process.
Image source: Getty Images.
The next chapter of AI's story is already underway
To fully appreciate the opportunity ahead, you have to go back to AI's past.
Modern AI has actually been around for a while now. It just wasn't very good, largely due to a lack of high-performance processors that make platforms such as OpenAI's ChatGPT and Google's Gemini possible. Conventional computer processors like those made by Intel and Advanced Micro Devices are good. They're just not capable of handling the massive data loads that most AI platforms require.
Then in 2016, artificial intelligence hardware powerhouse Nvidia (NVDA +4.94%) turned a technological breakthrough into a game-changing commercial AI product. Leveraging several years' worth of experience in the cryptocurrency mining space, the company utilized the same high-capacity technology used in computer graphics cards into the DGX-1, the world's first true deep-learning supercomputer that would eventually evolve into the architecture that ushered ChatGPT into existence. And for years, Nvidia was the go-to source for such technology, which is why it's powering the majority of most modern artificial intelligence platforms today.
As could have been expected though, the industry wasn't going to let Nvidia maintain its near-monopoly forever. Aside from steep pricing for its systems, AI data center operators are finding they've got very specific needs. Their solution? Sidestepping Nvidia altogether and custom-building their own artificial intelligence processing chips.
Now, after a few years of tinkering, this is not only a viable option, but an option that's being increasingly utilized.
The ASIC evolution is underway
They're called application-specific integrated circuits, or ASICs. Just as the name suggests, this is silicon that's custom-built for a specific purpose.
Technically speaking, lots of electronics and even large-scale data centers have used custom-built semiconductors for years now. The difference in the newer ones is their raw computing power. Modern-day ASIC chips are capable of performing as well as any of Nvidia's top-of-the-line data center hardware. Just ask Alphabet's Google, which is using custom-designed processors called Tensor Processing Units in a range of ways, including an artificial intelligence data center service offered to its institutional customers. Amazon's using them, too. Its so-called Graviton processors custom designed by Arm and manufactured by Taiwan Semiconductor Manufacturing for general cloud computing data center use are 60% more power efficient, in fact, lowering customers' total cost of use by as much as 20%.
And yes, Microsoft is on board as well. Microsoft Chief Technology Officer Kevin Scott even went as far as saying recently that the company would like to mostly use its own artificial intelligence data center chips going forward.
This is just anecdotal evidence of a much bigger paradigm shift in how the next generation of AI data centers will be built, of course. Credence Research expects the AI-specific ASIC market to grow at an average annualized pace of nearly 19% through 2032.
The thing is, against this backdrop, such optimism for this sliver of the semiconductor business makes plenty of sense.
Where to start the search
Great, but how does someone invest specifically in the artificial intelligence ASIC market? It's easier said than done. There aren't many pure plays in the business, and even fewer publicly traded ones that you might actually want to own.
Two of the bigger names in the business are worth a look, though.
The first of these is a very familiar Broadcom (AVGO +4.77%), which now makes custom AI processors for the aforementioned Google, OpenAI, and Apple. Although custom-built ASIC silicon isn't nearly as big of an AI profit center as the rest of its dominant data center connectivity solutions currently are, the longer the company remains in the business, the more its tech turns into a complete package.
NASDAQ: AVGO
Key Data Points
Then there's much smaller -- but arguably scrappier -- Marvell Technology (MRVL +12.56%). Although its $80 billion market cap means it's only a fraction of the size of its current top ASIC rival, its smaller size means it's got the advantage of not being bogged down by too many legacy businesses and aging assets; it can build itself from the ground up while eyeing the AI industry's proliferation in real time.
The only potential stumbling block with Marvell? Its small size is an advantage, but could also be a liability. A bigger and better-funded player could step into the AI ASIC business.
Intel may have just done it, in fact, announcing the creation of a new Central Engineering Group in September that's specifically tasked with helping the company design and deliver customized processing solutions. Then in mid-October, the company unveiled its own inference-optimized data center processor, directly putting it in a business it's never quite been in.
Now read between the lines to see the bigger picture. Intel's interest in this custom chip space underscores the scope of the opportunity here. More importantly to investors, just keep your finger on the pulse of this still-young AI trend. You don't necessarily have to deploy your investment dollars today. You just don't want to wait too long. Marvell and Broadcom are your best current bets, each with their own risk and reward profile. That, however, could certainly change in the future as the ASIC movement evolves.
