(NVDA), (AAPL), (NFLX), (AMD), (IBM), (AMZN), (GOOGL), (META)
I was sipping an overpriced airport cappuccino when Jensen Huang strutted onto the GTC 2025 stage. My seatmate didn't recognize him, but I'd been at a private dinner with one of Nvidia's VPs the night before.
After 40 years covering tech from Tokyo to Wall Street, you build a network that pays dividends in boardroom whispers. "Tomorrow's announcements will make portfolio managers drool," he'd hinted, remembering when I'd steered his family into Apple (AAPL) back in 2003.
Two minutes into Jensen's keynote, my neighbor's jaw dropped faster than Netflix (NFLX) stock after missing subscriber estimates. By our boarding call, he was frantically buying Nvidia (NVDA) shares.
"Too late, kid," I muttered. "You're about $2 trillion late to this party."
While investors obsess over Fed rates, Jensen has built the infrastructure for the biggest gold rush since the internet. Nvidia is selling the only shovels that actually work.
The Blackwell Ultra processor, due late 2025, isn't just an upgrade but a leap in AI reasoning that makes current models look primitive. The Vera Rubin server, arriving in 2026, is expected to be 3.3x faster.
A Stanford classmate who designs power systems texted during the keynote: "Been testing prototypes for months. Had to sign blood-oath NDAs."
When your college buddies design the circulatory systems for the world's digital brain, you get texts analysts would trade their Bloomberg terminals for.
Jensen projected global data center spending reaching $1 trillion by 2030. When I repeated this to my bartender at the St. Regis, he asked if I was having a stroke.
"No one spends a trillion on computers." I reminded him no one used to spend billions streaming cat videos either.
Who benefits from this tsunami?
Nvidia is obvious, with 20-25% projected annual growth looking conservative.
Advanced Micro Devices (AMD) has been gaining ground. At a San Jose conference, I caught an engineer who "accidentally" left next-gen chip data visible.
When you've covered tech since before AMD made math coprocessors for IBM (IBM) clones, you spot when a "mistake" is actually a strategic leak.
Amazon (AMZN) remains the cloud king. During an investor site visit, an AWS director let slip their AI buildout is "significantly ahead of guidance."
My history covering AWS when analysts dismissed it means their PR people don't hover during these visits. They forget I reported on Amazon when Bezos was still shipping books from a garage.
Google (GOOGL) keeps their best tech garaged. In an Uber after a Palo Alto event, a tipsy engineer boasted: "The public sees maybe 20% of what we're running internally."
Twenty years of moderating "Future of Computing" panels puts me in countless rideshares with people building that future, who forget I write for fund managers controlling billions.
Meta (META) is investing heavily in generative AI. At an industry roundtable, their researcher, after his third wine, sketched neural networks on napkins.
Having covered companies from garage stages to trillion-dollar valuations gives you invisibility at these events – they see you as furniture rather than media.
The generative AI market is projected to reach $62.72 billion by 2025, but at a Bloomberg roundtable, fund managers confided they're allocating "multiples" of their usual positions.
After decades of running a hedge fund that spotted Japan's rise, the dot-com boom, and the fracking revolution, you get invited to side conversations where real money moves are discussed.
My advice: don't pick a single winner, think long-term, and watch innovation closely.
My nephew recently asked for stock tips for his summer job earnings. "Buy Nvidia and delete your trading app for 10 years," I advised. He looked offended. "Everyone knows Nvidia. I want something undiscovered."
I patted his shoulder. "Kid, sometimes the biggest mistake in investing is trying to be clever when the obvious answer is staring you in the face."
In the AI revolution, it's not about finding the next Nvidia. It's about not overthinking the Nvidia that's already here.
