Computers went from room-sized mainframes to desktops to phones in your pocket. You'd think central infrastructure would shrink with them. It did the opposite. Every smaller device created more demand for the bigger buildings behind it.
$DLR
77
Walk into a bank's computer room in 1965: raised floor for cabling, controlled temperature, locked doors, backup systems, round-the-clock technicians. Swap the tape drives for GPUs and you've described a modern AI data center, minus a few gigawatts.
$EQIX
46
On April 7, 1964, IBM unveiled the System/360, a family of compatible machines a company could grow into. The pitch wasn't a computer. It was: build your business around computing. Sixty years later, the hyperscalers are making the same pitch.
$IBM
80
Span is private, so the obvious trade doesn't exist. Every mini data center still needs rugged housings, electrical controls and backup power. When you can't buy the miner, sell the generators.
$GNRC
66
Training needs giant concentrated clusters. Inference, the part that answers every query, doesn't always need to cross the country. The next AI layer may look more like rooftop solar than a warehouse.
$VRT
80
Your house is wired for more power than it draws at any given moment. The AC cycles off, the oven sits idle, the EV is full. Distributed compute turns that dormant capacity into rentable infrastructure.
$ENPH
108
Nvidia partnered with Span to bolt air-conditioner-size AI compute boxes onto individual homes. When the chip king starts routing around the data center, the grid bottleneck stops being theoretical.
$NVDA
87
Big Tech can raise another $100 billion. It can't conjure a transformer with a multiyear lead time or skip an interconnection queue. AI's growth curve now runs through utility permitting offices.
$ETN
135
Reprocessing tailings can recover metal and shrink a remediation liability in the same move. Try getting that from a greenfield mine. One company's trash becoming another's feedstock is a balance-sheet story, not a recycling slogan.
$NEM
47
Shale was useless rock until the technology caught up. Mine waste could be next. The resource doesn't change; the recovery cost does. Wall Street chases discovery headlines. The quieter money is in a pile that gets rerated.
$FCX
74
Beijing isn't panic-selling gold on the pullback. It's shopping. Maybe the world's most patient buyer knows something your broker doesn't.
73
DOE put $134 million behind commercial rare earth recovery from tailings, e-waste and other leftovers. Nobody doubts the atoms are in there. The money is in whoever can separate them profitably and at scale. Chemistry beats geology.
$UUUU
1
225
A new deposit means drilling, permitting, financing, roads, power and years of moving dirt. A tailings pile skips most of that. It's already mined, hauled and often crushed. The chemistry still has to work, but the shovel work is done.
$MP
115
For a century, American miners kept what they came for and dumped the rest. Now USGS is building a National Mine Waste Inventory to catalog those piles. The next strategic mineral reserve may be the stuff we already hauled out and forgot.
$REMX
70
A Treasury bond is a promise from the U.S. government. Gold is nobody's promise. China now holds roughly 2,387 tons, about 9% of its FX reserves, and keeps adding. That's not a bet on rates. It's a vote on counterparty risk.
$PHYS
71
A falling price is not a broken thesis. Gold set a record in January, then dropped as much as 4% in one session to near $4,111. The question isn't where the price went. It's whether the buyers who drove it there left. China didn't.
$GDX
109
The 10-year Treasury yield just hit its highest level since June 2007. That's the real reason gold dropped. Bullion pays nothing, bonds pay more every week. It's a legitimate headwind, and it explains nothing about why Beijing keeps buying.
$TLT
149





