Income traders,
Time to step inside my kitchen, where I'll show you some of the special sauce we use in Turbo Income.
Samsung unveiled a next-generation AI memory part Tuesday at the Future of Memory and Storage conference out in Santa Clara. It's a prototype called V10 BV-NAND, over 400 layers, about 58 percent denser than what they're shipping today. The tape immediately hung on the whole memory complex, and Sandisk didn't help when it reported Wednesday night and dropped 12 percent.
So the story writes itself. Bottleneck solved. Sell memory. Right?
Read the Sandisk Print Again
Here's the thing. Sandisk (SNDK) did $8.97 billion in revenue, up 51 percent from the prior quarter, and earned $39.25 a share when the Street wanted $33.38. That's a beat on both lines, and the stock fell anyway because next quarter's guidance came in light.
The part nobody put in a headline is that Sandisk now has 10 long-term supply agreements on the books, worth at least $93.9 billion, averaging north of four years each.
Then Wednesday afternoon CoreWeave (CRWV) signed a multi-year deal with Solidigm for priority access to enterprise SSD capacity (SSDs are flash storage drives, expensive but fast). The release says the quiet part out loud: storage supply is tightening across the industry.
Nobody signs multi-year supply in a glut. You do that when you can't get the part on the spot market and you're scared of what it costs next year.
Those two headlines are not the same kind of information. A prototype on a conference stage is competing for revenue that doesn't exist yet, while a signed contract is competing for inventory this quarter.
Why I Screen for Monopolies
Amazon (AMZN) already spelled this out for anyone paying attention. Andy Jassy raised 2026 cash capex to $220 billion from $200 billion and put the blame directly on what memory costs now. He then told analysts Amazon still won't have enough capacity in 2026, and figures 2027 looks the same.
That's the screen I've been running all week and I want you running it too. In a compute shortage I only want names with monopoly-grade positioning, and memory is the benchmark I measure everything else against.
Three companies control better than 90 percent of DRAM revenue. Fabs take years to build. Samsung told analysts last week that long-term contracts could eventually run 60 to 70 percent of its memory sales, which is what happens when a commodity business stops behaving like one.
Flash is a bit more crowded, five companies covering roughly 90 percent, so I hold it to a looser standard. Lead times run into years either way.
That screen kills most of what gets called "the AI trade." Broad semis are too competitive, everybody has a chip. I'll pass on the hyperscalers too, because the spending isn't ending and the market is pricing a Fed hike before year-end rather than a cut, which makes funding it more expensive as they go.
What's left is a short list. Own the bottleneck, not the application.
How I'd Get Paid on This
For income, a headline-driven dip in a name whose supply curve didn't actually change is the setup. Not a reason to chase. A reason to get paid.
When IV runs hot on a supply-story name after a prototype headline (IV is implied volatility, basically what the options market charges for the odds of a big move), put credit spreads earn their keep. You sell a put, buy a cheaper one underneath it, collect the difference up front, and know your worst case walking in.
What you're really doing is collecting on a fear with a long fuse while the contract cycle says otherwise. Then Turbo the upside if the MTI is loving the name. My Market Temperature Indicator turns green when a trend has real staying power behind it, which keeps me from confusing a bounce with a move.
Keep it 30 to 45 days out. Size so assignment leaves you in a position you wanted anyway. And don't stack it all on one ticker in a complex that trades as one thing.
Class dismissed!
Here for a good time AND a long time,
Hans