Rocket-Powered Frisbees and Barbed Wire Telegraphs
We've all had bad ideas. Some ideas sound fun at the time, but, for reasons, are still bad ones. Strapping rockets onto a frisbee sounds awesome. But there's burns, maybe missing digits or, possibly, decapitation and things to worry about.
Having a tin can phone strung between houses so you can talk to your best friend next door? Solid kid plan. Repurposing barbed wire to carry a telegram to your best friend half a mile away? I can see potential problems with that. And what if your rocket-powered frisbee gets tangled in it? What then?
But the kind of mind that cycles through enough of these truly terrible ideas might, eventually, land on a few really great ones.
Working in Bell Labs in the 1940s, Claude Shannon hit upon the idea that information needs to be novel. To be novel, information needs to be unexpected — not something that was known before. He came up with this in the context of making better telephones and well after he had built rocket-powered frisbees and a barbed wire telegraph.
Shannon also gave us the "bit" — the way computers quantify information as “1” or “0.” His various contributions to communication, computation and information theory earned him a special kind of immortality. Anthropic named their LLM “Claude” after Shannon. Now, millions of people invoke him every day.
What's new?
Shannon's framework for information laid foundations for computer science, but also had implications for natural science like biology and physics. The quest for novelty is also a useful approach for investing.
Prices and news come at investors through a firehose. Most of this, though, doesn’t count as truly novel. Things that the market already broadly knows or expects are difficult to act on. The market mostly does a pretty good job of pricing this in. Investors look for novelty — surprises.
They often talk about surprises relative to some existing expectation. “Apple printed an earnings surprise. It beat eps by a dollar” — something like that. Those can move prices. They carry a real investing signal. Investors also expect these kind of surprises from time to time. Sometimes companies do better than expected, sometimes they do worse.
Then there are surprises that the market hasn't really thought about. Sometimes they seem small but might carry a large signal.
Here's Rambus reporting earnings. The computer memory component supplier had a great quarter. Revenue: solid. Earnings: great. It also announced that, for the first time, it inked one of the hyperscalers — Amazon, Microsoft, Google or Meta — as a customer. Management declined to say who.
Computer memory has grown into one of the most oligopolistic of oligopolies. Three memory makers, Micron, SK Hynix and Samsung, supplied nearly the whole market. Those are Rambus’ customers. China's CXMT has made huge inroads the last few years, pushing the market to four players. It IPOed in China in July at around a $60 billion value which investors quickly bid up to $240 billion. That value crowned it the most valuable Chinese public company.
The future of memory, though, looks a bit different. In addition to creating vast demand for compute, AI created vast markets for specific kinds of compute. NVIDIA has been the biggest winner from this specificity.
In the pre-AI days, one kind of memory was good enough. Sure, some folks at the cutting edge would love for memory to work a certain kind of way. But those were a few researchers working at CERN or down the road from Lamplighter at CalTech or doing high frequency trading. Given the huge fixed costs to produce memory, it didn't make sense to spin-up customizable solutions.
AI changed that. The compute market is much, much bigger. That means that relatively small corners of the memory market are also bigger. Some of those are big enough to support new, more diverse branches of memory. Rambus' presser that it signed on a hyperscaler customer could be a signpost in that direction.
Surprise!
For Rambus, the announcement highlights two things:
It diversified its customer base. It's not perfect diversification, because all its customers still hinge more or less on AI, but it does provide an off-ramp for some customer-specific risks.
It opened a new market for its IP and, if you squint hard enough, a potential new market for its products.
Rambus' announcement checks the box for something novel. It will probably take a while to test whether this particular piece of novel information is important. We're going into year five since LLMs stormed the market and AI is still a bit Wild West. In the meantime, this new market is not something that investors had paid for, so makes the company's shares just a bit more attractive.
Disclaimer: None of this is investment advice. It's meant to illustrate ways LCM thinks about investing. Things that LCM decides are good investments for LCM and its clients are based on many criteria, not all of which are covered here. Some or all of LCM's ideas may not be suitable for other investors. LCM does not recommend investing either long or short any position mentioned. LCM may own positions in some of the companies mentioned. Some of its ideas will lose money — investing entails risk. See full disclaimer here.