Listen to the audio version of this article (generated by AI).
Tom Yeung here with your weekly Sunday Digest.
In March 2025, I recommended three blue-chip companies to buy for the AI Revolution:
- Monolithic Power Systems Inc. (MPWR), a leader in power management chips for AI devices.
- Workday Inc. (WDAY), an AI-enhanced cloud-based HR platform.
- Xometry Inc. (XMTR), a marketplace for AI-powered manufacturing.
The trio has since performed exceptionally well, which isn’t surprising given that we’re in the middle of the AI Revolution. Despite a pullback in Workday from a broader “SaaSpocalypse” panic, these three stocks have returned 105% on average, more than doubling the returns of the tech-heavy Nasdaq Composite index.
However, this presents a new problem. Much like mining a gold seam dry, these stunning gains mean there’s not much left for latecomers. You either got in early and enjoyed high returns, or came in late when only dry rocks remained.
Fortunately, InvestorPlace Senior Analyst Luke Lango believes there’s still one last rich AI seam that’s been overlooked:
Physical AI: taking artificial intelligence out of computers and putting it into cars, robots, factories, and other machines operating in the real world.
Luke argues that shift is arriving faster than most investors expect, and that it will take four layers to make it work. And in a first-of-its-kind InvestorPlace workshop, he maps them onto Elon Musk’s businesses:
- Data. X, a real-time feed of human behavior.
- Compute. Colossus, the supercomputer xAI built in Memphis to train Grok.
- Connectivity. Starlink, SpaceX’s satellite network.
- Robots. Optimus, Tesla’s humanoid robot.
No conglomerate builds all four by itself. Not even Elon Musk’s empire.
That’s Luke’s point: Every layer depends on a network of outside suppliers, the firms he calls “Chosen One” companies during his special broadcast. These make the parts that Musk’s businesses can’t produce themselves. In his presentation, Luke reveals one free pick that feeds these layers, then points to where you can find seven more.
Today, I’ve been given special permission to share Credo Technology Group Holding Ltd. (CRDO) – one of these seven picks to give you a glimpse of how essential (and overlooked) Luke’s picks are to the Physical AI Revolution.
Wiring Up the Physical AI Revolution
Credo Technology runs a straightforward business:
It creates high-tech cables known as active electrical cable (AEC) that run inside AI datacenters.
You see, most datacenters up to this point have been wired up using an old technology known as direct-attach copper (DAC). That’s just a fancy way of saying “copper wire,” the same tech that elementary school children use when building their first electrical circuit. DACs are cheap and easy to produce.
However, copper is not perfect. Some energy is always lost as heat, and electrical signals get distorted and “attenuated” as they pass through the wire. It’s why you sometimes hear power lines buzz, and why professional DJs insist on buying hundred-dollar Monster Cables instead of using the $1 spools from the local hardware store.
The issue is even more problematic for AI datacenters, which require far higher precision.
Credo’s AEC cables fix this problem by adding tiny digital signal processing (DSP) chips along a copper wire. These little devices do the following:
- Read the incoming, distorted signal
- Figure out what the signal was trying to say
- Retransmit a brand-new, clean signal to the next DSP
These active cables work fantastically well. They can handle far more data than traditional copper wires, send information along further distances, and prevent the dreaded “link flap” where a network connection drops because the signal has become so garbled.
Why Credo? Why Now?
AI datacenters today require much more data than ever before. For example, a single Blackwell AI chip from Nvidia Corp. (NVDA) can move around 8 terabytes of data per second… or roughly 680 DVDs in the time it takes someone to blink an eye. These Blackwell chips are then typically run together in clusters of 72, and medium-sized AI datacenters can run thousands of these clusters.
To keep these chips, servers, and datacenters supplied with fresh data, researchers have turned to using higher frequencies to transmit information. Think of it like talking very fast: It’s much easier to say things quickly in a squeaky high-pitched voice than in a rich, gravy-like baritone. (Try it yourself!)
And because we humans tend to take technologies to their extremes, today’s AI datacenters use frequencies that are over a million times higher than what humans can hear.
This presents an issue for traditional copper wire. Distortion becomes worse at high frequencies, and data becomes more garbled the longer it travels down a wire. An ultra-fast AI chip might transmit the best information in the world… and no one will understand what it’s saying if it’s connected by the wrong cable.
Now, this frequency issue was not a hurdle for older datacenters. Most only required several chips to be closer together and data frequencies were lower back then. In fact, Microsoft Corp. (MSFT) in 2023 allegedly decided to cut back on AECs because they were so expensive. Shares of Credo fell 46% in a day following that announcement.
But new AI datacenters need AECs not only because frequencies are now higher… but because AI servers are physically so large (since there are so many chips) that a single cluster of them might run 3 to 7 meters from end to end. That’s too far for traditional copper to reach, and wastefully close for fiber optics. AECs serve this “sweet spot” in between.
That’s why growth at Credo has suddenly accelerated. Revenue growth in fiscal 2026 hit 206% (up from 127% a year earlier), and GAAP net profits rose ninefold. Credo’s products sit at the perfect 3 to 7 meter wire lengths, and the company is now enjoying being in the right place at exactly the right time.
Credo: An Essential AI Player
Luke and I expect demand for AECs to continue rocketing higher. The entire Physical AI Revolution will require more computing power than ever before, and that means more datacenters… more chips… and more specialty wires that connect everything together.
In addition, Nvidia’s Vera Rubin next-generation AI chips will use a new communication standard that requires communication frequencies of 53 GHz. That’s double what the current Blackwell generation uses, and will cut the effective range of DAC copper wires to barely 1 meter (3 feet). Imagine an IT manager being given only 3-foot cables to wire up a datacenter the size of a football field!
That’s why Credo is so essential to the Physical AI Revolution. It has over a hundred active patents on its AEC technology (plus another 80 pending), decades of experience in building the technology, and numerous legal wins where it successfully defended its intellectual property. The company also co-designs its AECs with customers, creating a “lock-in” effect.
Now, I must mention there is an alternative known as active copper cable (ACC) that can boost signals to roughly 2-3 meters. This technology has spooked some investors, since it is cheaper than the AECs that Credo produces. But ACC technology faces the same problem as copper wire because it lacks the multiple signal boosters that AECs have. Once we get to the next high-frequency standard after Vera Rubin, ACCs will face the same wall that pure copper does today.
The Value Behind the Tech
Shares of Credo are extremely attractive at current prices. They are down almost 50% since peaking in June, even though fiscal 2027 guidance has been revised up. The customer list has also broadened, with AI datacenter companies like Meta Platforms Inc. (META) and “neocloud” firms taking up more production. Even Microsoft has reversed course and returned as one of Credo’s largest customers.
CRDO shares now trade for just 26X forward earnings – the lowest 1% of Credo’s post-IPO range
Given this, the share price could easily double – and that’s without the additional demand from Physical AI. The world is going to need a lot more high-tech cabling to overcome copper’s physical limitations, and most regular investors haven’t figured out that a wave of demand is heading our way.
As I mentioned, this is just one of Luke’s “Chosen One” companies. To find out how to access all seven – plus his free pick – click here to watch his Vertical AI broadcast.
Until next week,
Thomas Yeung, CFA
Market Analyst, InvestorPlace
