3 Small Artificial Intelligence (AI) Stocks With More Than a Fighting Chance to Win Big

Not every long-term winner is already an industry giant. Today's business titans were smaller and hungrier once upon a time, and some of today's under-the-radar small-caps will be tomorrow's sector leaders.

That's especially true in the explosive artificial intelligence (AI) market. There is so much going on in the evolving market's rarely covered nooks and crannies that we probably haven't even noticed the biggest sustained winners yet.

So we asked a few of the Fool's tech experts to share their best bets among today's unsung AI heroes. Anders Bylund suggests signal-processing researcher Ceva (CEVA 1.11%). Nicholas Rossolillo proposes cloud-computing service provider DigitalOcean (DOCN 3.30%). Billy Duberstein prefers chip-manufacturing technologist Kulicke & Soffa Industries (KLIC 1.86%).

A tiny cloud infrastructure company suffers (temporary?) setbacks

Nicholas Rossolillo (DigitalOcean): DigitalOcean is a cloud computing infrastructure provider focused on startups and small and midsize businesses (SMBs). It competes against heavyweights Amazon, with Amazon Web Services (AWS); Microsoft, with Azure; Alphabet, with Google Cloud, and Oracle, with Oracle Cloud. But rather than getting steamrolled as some might expect, DigitalOcean has more than held its own.

Data sources: Amazon, Microsoft, Alphabet, Oracle, and DigitalOcean.

Despite its diminutive size, DigitalOcean is also free cash flow positive and is homing in on GAAP net income as well.

Data by YCharts.

Nevertheless, the upstart cloud infrastructure provider has suffered a few setbacks as of late. Last quarter, management downgraded its outlook for growth the back half of 2023, citing global economic conditions affecting its customers. Full-year revenue guidance was slashed by more than $35 million and is now expected to be as much as $685 million, essentially flat with the annualized revenue run rate at the end of Q2 2023. It was also revealed that some deficiencies in DigitalOcean's accounting procedures were discovered. The problem is fixed, and the slip-up wasn't detrimental to the company's earnings.

However, it was announced shortly thereafter that CEO Yancey Spruill will be stepping down after the board of directors finds a new CEO. That has sent the share price down some 50% in the last month, back to where it was at the start of 2023. 

DigitalOcean is a high-risk and only potentially high-reward stock. But I think its recent issues are solvable, and the company still has more than just a fighting chance as it grows with its SMB customers. I recently took another small nibble.

From a few cloud servers to every pocket: Ceva's expanding AI reach

Anders Bylund (Ceva): This is the most direct bet on AI you never heard of. Ceva develops digital signal processing (DSP) solutions to create or read signals in audio, wireless networking, image processing, and mobile device sensors. Name a semiconductor designer or consumer electronics maker, and you're probably talking about a longtime Ceva customer. The company has been providing leading-edge DSP technologies since 2002. However, the nature of their technology licensing business means they often operate behind the scenes, making them one of the industry's best-kept secrets.

But Ceva shouldn't be a secret much longer. The company's DSP patents are absolutely central to generative AI services such as the ChatGPT chatbot and the and DALL-E image-generating platform. And it's one thing to rely on Ceva's patented technology in a handful of high-powered cloud servers, and another when similar technologies roll out to every smartphone in the land. The unit volume is much wider on the consumer-facing side of the street.

Ceva's solutions help chip makers put together the circuits and software that are needed in modern AI systems. The NeuPro-M processor architecture, for example, provides neural network cores to accelerate generative AI and classic machine learning workloads.

Now, Ceva is no newcomer. Your smartphone, wireless earbuds, smartwatch, and robotic vacuum cleaner all depend on this company's technology. If the gadget either interprets or outputs real-world data in the form of audio, images, and wireless signals, Ceva probably had a hand in the process. But that doesn't mean its market is oversaturated. Its AI products carry much higher price tags and more lucrative profit margins than wireless radio processors and smartwatch sensors.

In the second-quarter earnings call a month ago, Ceva CEO Amir Panush highlighted recent licensing deals with four leading carmakers, related to self-driving sensors and in-cabin radar systems.

"And with generative AI and the overall focus on high-performance neural processing units for the AI, those are the types of technology that will help us actually to expand very, very nicely in other regions," he said. "I believe that our ability to support transformer architecture with exceptionally low power consumption and highly efficient positions as very well to exploit this new wave of AI across the full spectrum of end markets from consumer IoT to industrial, automotive, and networking."

Ceva is a cyclical business. Its revenues and profits tend to spike when a large number of semiconductor start-ups enter the market together, and I think we'll see a lot of that action over the next few years in an AI-inspired Silicon Valley gold rush.

But the AI-loving market hasn't caught on to Ceva's artificial intelligence opportunity yet, and the stock price is down by 16% in 2023. You can grab Ceva shares at the modest valuation of 4.2 times sales today. That's a bargain-bin deal for this experienced technology expert.

This cheap cyclical is primed for an AI-powered up-cycle

Billy Duberstein (Kulicke & Soffa): Artificial-intelligence chips will require increasingly more advanced packaging innovation, as packaging complexity is necessary to coordinate all the GPUs, CPUs, high-bandwidth memory (HBM), Ethernet, Infiniband, and optics needed for AI workloads into power-efficient structures. In addition, we're now seeing processors themselves constructed in "chiplet" form, with several optimized parts of the chip made separately, then stitched together in a package.

This is why Kulicke & Soffa could be an unheralded beneficiary of the AI trend. While the bulk of its business is in traditional ball bonding equipment for high-volume electronics, it does have a small but fast-growing thermocompression bonding (TCB) business, which is one of several key new advanced packaging technologies.

On its recent third-quarter conference call, management noted that even though its core business was down significantly year over year because of the downturn, K&S shipped a record number of thermocompression bonding systems. Moreover, the company indicated that it had broadened its engagements with more customers, which could materially grow its TCB revenue from these levels.

CFO Lester Wong noted that the company had shipped more than $40 million of TCB systems over the past two quarters, for a greater than $80 million run rate, which should grow to, in his words, "well over" $100 million in TCB revenue in fiscal 2025, which ends in September 2025. That could then make up a significant portion of K&S revenue, which totaled just $826 million over the past 12 months.

In addition, management noted that it had begun a new packaging technique with an HBM customer, using vertical wires instead of through-silicon-vias to connect the DRAM module stacks, The technique apparently improves performance, and with HBM in high demand for Generative AI applications, that's an area where K&S makes very little revenue now but could grow significantly in the future.

Finally, K&S is starting to see a recovery in its core ball bonding business after the past year's downturn. To capture the opportunity, the company just introduced a new line of ball bonders with higher margins and more capability.

The stock currently trades at 27 times earnings, but those earnings are at a cyclical trough. At $48 per share as of this writing, the stock trades at a paltry 6.8 times 2022 earnings, which was the most recent peak. Moreover, the stock is much cheaper on an enterprise value basis, as K&S has more than $700 million in cash on its balance sheet and no debt. That cash amounts to more than 25% of its market cap.

In all, Kulicke & Soffa is a cheap stock coming off a cyclical bottom, and its recovery should be enhanced by AI, which should enable earnings to eventually make new highs.