These firms are producing the “Golden Rivets” that the technology desperately needs…
Tom Yeung here with your Sunday Digest.
Last week, I wrote about two “future-proof” stocks to help shield your portfolio against the threats of AI. Artificial intelligence was causing mayhem in the software industry, and former SaaS superstars like Salesforce Inc. (CRM) and SAP SE (SAP) were suddenly looking less shiny. Shares of these firms have fallen 40% or more from their peaks.
I offered fertilizer company The Mosaic Co. (MOS) and Australian whiskey maker Lark Distilling Co. Ltd. (LRK) as defenses from the disruption.
But what about having a good AI offense as well?
That’s more of a challenge. The best time to buy AI chip makers was six months ago when prices were still reasonable.
Since then, memory-chip maker Micron Technology Inc. (MU) has risen 262%, while energy storage firm Fluence Energy Inc. (FLNC) has jumped 111%. The “obvious” plays are now trading at valuations that would make venture capitalists uneasy.
So, it’s challenging… but not impossible.
In a new special presentation, InvestorPlace Senior Analyst Eric Fry says there’s still time to get in on a second AI wave. These are smaller companies that are building the “Golden Rivets,” which are the essential components that power first-wave firms like Micron and Fluence. Without them, it’s like trying to run a race car without fuel.
During the presentation, Eric reveals a whole host of AI bottlenecks – raw materials, digital memory, energy – and identifies the specific companies that are now seeing unprecedented demand thanks to being in the right place at the right time.
You can click here now to watch his FutureProof 2026 event.
Now, to give you a sense of these “Golden Rivet” makers, I’d like to highlight two companies at the forefront of the AI bottlenecks… and that have been overlooked by Wall Street almost entirely so far. You’ll quickly see why I’m so excited for Eric’s picks.
A Dominant Company in the Background
Few people have ever heard of copper sulfate plating – the technology that creates the ultrathin wiring in modern chip packages.
Even fewer people have heard of JCU Corp (TYO:4975), the under-the-radar Japanese firm that dominates that industry.
In 2019, the Tokyo-based firm estimated it had a 70% global market share in copper sulfate plating for smartphones and tablets. That figure has likely grown, as shown by JCU’s sky-high operating margins that continue to rise. It’s now on track to earn 40% operating margins this year, up from 27% in 2019.
That level of profitability is rare in manufacturing. It’s even more remarkable given Japan’s traditionally low-margin corporate environment. To give you a sense, JCU earns higher profit margins than Apple Inc. (AAPL) does from selling premium iPhones.
That’s because JCU’s products sit inside a crucial, failure-sensitive part of chipmaking. Here’s a simplified version of how it works…
A chip package starts as a specialized insulating material with tiny holes drilled into it with lasers. The material is then cleaned, chemically treated, and dipped into a copper sulfate bath to deposit an ultrafine copper layer exactly where it’s needed. (This is the step JCU allows.)
After that, the package is sent through an etching process, where excess copper is removed, and the finished piece has silicon components-attached in a die-bonding process. And if a package needs multiple layers, the process starts over again.
This matters because packaging defects can ruin an entire chip. If copper wiring is defective, the final product could perform poorly, degrade over time, or simply not work at all. Data centers would end up with expensive paperweights.
That’s where JCU’s technology comes in. The company offers a precise recipe and control system for copper sulfate plating used to create wiring of 0.8 micrometers and less –almost 10 times finer than what conventional methods can achieve.
In addition, JCU’s historical dominance means that it’s baked into the process of its customers. Chipmakers know how to precisely etch the copper from JCU’s recipe for the right outcome each time. Why should they risk rolling the dice on a new copper sulfate plating system when the current system works so well?
Two factors are now putting JCU on a high growth path.
The first is the rise of 2.5D and 3D chips. Stacked chips require multiple rounds of copper plating. They also have connection areas called “vias” that send electrical signals between layers, which requires a specialized form of plating. JCU has launched a brand called TIPHARES to deal specifically with stacked chips and anticipates strong demand.
The second is that AI data centers have created a shortage in virtually every computer component. GPUs, hard drives, NAND flash memory, and DRAM have seen their prices rise uncontrollably, and some makers have already sold out their entire 2026 inventory.
That means we should expect a ramp-up of production across the entire semiconductor industry, benefiting JCU at every turn. Virtually every modern semiconductor requires packaging of some kind, which all feeds into the demand for this Japanese firm. JCU is a natural bottleneck because the company is so dominant in its niche.
That’s why I believe estimates for JCU’s growth are far too conservative. Analysts are currently estimating just 31.5 billion yen in 2027 revenues (a 5% annual growth rate), which is roughly what JCU was guiding for in 2024… well before the semiconductor shortages began.
To put that into perspective, revenues already rose 14% in 2025 and operating profits surged 31%.
It’s also noteworthy that markets have not yet fully recognized JCU’s value. Shares trade at just 16X forward earnings, which is already ludicrously low for a company with 40% operating margins. Though shares may be difficult for American investors to buy, JCU’s dominance of its industry could make them worth it.
Rolling the Dice
Those seeking a higher risk/reward “Golden Rivet” company will find one in Cohu Inc. (COHU).
Cohu is a semiconductor test and inspection equipment maker that competes directly with industry giants Teradyne Inc. (TER) and Advantest Corp. (ATEYY). The two larger firms control over 80% of the overall chip testing equipment market and spend roughly as much on research and development (R&D) annually as Cohu generates in total sales.
Traditionally, that’s left Cohu with scraps. The San Diego-based firm focuses on the less desirable midrange market and on test handlers – the robots that physically transport the chips being tested. Margins in both are lower and far more cyclical, because customers can delay purchases without fear of technologically falling behind.
Since 2000, Cohu has posted 16 years of positive operating income and 10 years of negative income. Automotive, industrial, and mobile manufacturers are notoriously tough customers.
COHU operating margin %
Source: Refinitiv
This cyclicality means Cohu’s shares now trade 40% below their 2021 peak. Revenues have shrunk 48% since 2022 on a cyclical downturn, and net income turned negative starting in 2024. In an earnings call last year, CFO Jeffrey Jones admitted that customers were delaying shipments, forcing the firm to cut 2025 forecasts.
However, insatiable demand for AI chips is changing that picture. Last month, Cohu’s management announced that annual revenue growth had turned positive again, and that margins were on the rise. This was driven by both a cyclical uptick in mid-end customers, as well as strong demand from customers working with AI data centers, high-bandwidth memory, and physical AI applications. System orders (the higher-margin type) rose 47% quarter-on-quarter, and analysts now expect net income to flip positive again this year. Wall Street forecasts profits to double again in 2027.
Cohu has also seen some early success with its new Eclipse platform, which is designed specifically to test AI data center chips. Two major customers have now adopted Eclipse for AI testing, and Cohu’s management recently said they now expect to achieve the “upper end” of revenue forecasts for their high-performance computing (HPC) segment this year. They foresee Eclipse shipments accelerating in the second and third quarters.
This is all excellent news for this traditionally cyclical firm. Hyperscalers like Microsoft Corp. (MSFT) and Amazon.com Inc. (AMZN) are projected to spend trillions of dollars through at least 2030 building out AI data centers, and these big spenders have already triggered shortages in the AI chip testing market. In January, Advantest said it was speeding up its expansion plans to keep up and boosted its profit forecast by 21%. Teradyne has reported similarly bullish outlooks.
This is particularly bullish news for Cohu, since its larger rivals are now having trouble keeping up with demand. Customers may switch to the smaller supplier simply to access the AI chip testing they need.
That makes Cohu’s stock worth considering. The lows of cyclical companies might be very low, but that also makes their highs almost stratospheric.
The Golden Rivets
The two companies I mentioned here both have some downsides. JCU is potentially a value trap, because it receives virtually no Wall Street coverage and is difficult for American investors to buy. Meanwhile, Cohu is a cyclical play with far higher downside risks. Only active traders should consider such investments.
That’s why I want to make sure you tune in to Eric’s latest presentation, where he talks about 15 separate “Golden Rivet” picks before homing in on his top choices. These are companies like Nvidia Corp. (NVDA), Advanced Micro Devices Inc. (AMD) and Broadcom Inc. (AVGO) that have solved AI bottlenecks… except Eric’s new picks have yet to see the 10X gains those companies have because they’re still early in the cycle.
But don’t wait long. We’ll only be replaying this free presentation until midnight on Wednesday, so be sure to watch his special talk before then.
Until next week,
Thomas Yeung, CFA
Market Analyst, InvestorPlace
P.S. I will be revisiting my top picks for 2026 in the coming weeks as we enter the second quarter. In the meantime, Larimar Therapeutics Inc. (LRMR) is added to that list.
