For the better part of two decades, owning the largest technology companies in America, overconcentrating in them, and sipping margs on the beach was a winning investment strategy. The Magnificent Seven enjoyed some of the most wonderful moats in the history of capitalism. Each had a well-defined niche: Google in search, Apple in premium hardware, Amazon in e-commerce, Microsoft in enterprise software, Meta in social, Nvidia in chips. They veered into each other’s lanes occasionally (cloud computing), but for the most part each defended a fortress the others did not seriously besiege.
That gave them the fortress balance sheet and free cash flow to simply buy any startup that might out-innovate them: Instagram and WhatsApp before they could threaten Facebook; YouTube and DoubleClick folded into Google. Truly head-to-head competition was surprisingly limited. Détente. Everyone made money.
Détente Gives Way to a Brawl
A summer camp counselor at Geneva Glen once told me: “God forbid you ever get in a knife fight, but if you do, remember that the winner is the guy who bleeds less.” The Magnificent Seven now look remarkably like Sharks and Jets squaring off in West Side Story. All of that cash, all of those brilliant people, all of those fortress balance sheets are going head to head in pursuit of AGI and superintelligence. For the first time, they are not defending separate castles. They are in the same alley, and several of them are already bleeding. Increasingly, the Mag 7 are becoming one correlated bet on AI with a binary outcome return profile, closer to venture than large-cap stocks.
The Magnificent Seven now look remarkably like Sharks and Jets squaring off in West Side Story.
ROI, Meet ROS
This capex cycle is unlike any before it, and I think the market is misinterpreting it with an outdated model. Analysts judge this spending by ROI (return on investment). The CEOs are running a different calculus: ROS (return on survival). When they approve another tranche of tens of billions, they are not running a calm discounted-cash-flow calculation. They are spending against an existential fear: that being second to superintelligence is the same as being dead. Whether or not they are right, that is how they see it — and it changes how we should read the spending, and how we judge the rationality of the spenders.
The numbers are staggering. The four largest AI spenders — Microsoft, Alphabet, Amazon, and Meta — laid out roughly $360 billion of capex in 2025, up from about $219 billion the year before, and their own guidance points to north of $700 billion in 2026. That is not a capex cycle. That is a national-scale industrial mobilization, financed largely by the cash flows of a handful of advertising and software businesses.
The Price of Stupidity
My concern is that for some of these companies, this reinvestment is in the wrong place at the wrong time. It is fundamentally changing what these companies are: decimating the free cash flow I have adored for years, and sucking up cash that could have bought the inevitable disruptors using this tech to innovate on their models. You might say these businesses are run by some of the smartest people in the world. They have a great track record of reinvestment! But do they? Are you living in a different universe — the Metaverse, perhaps?
This is a wonderful example of what happens when a CEO decides he has found the next big thing and pours our money into it without pausing to ask whether it is wise. I have said it before, and I will say it again: a bad manager can tank a great business. Mark Zuckerberg is a bad manager. He just happens to manage such a great business that it can shrug off an $80 billion mistake.
Meta is probably the most obvious beneficiary of AI, already earning a return in the form of growing advertising revenue. That is not my argument. My argument is that some of these CEOs lack the discipline to size, scope, and price the ROI of these investments in the face of a new technology and a potentially imminent threat to their business. I think Zuck is one of them. So if he overdoes it, chasing superintelligence instead of targeted investment in AI that widens his moat in advertising, we could have what we experts like to call — a problem. More broadly: if AI does not return in line with the historic ROIC these businesses earn, can the core operating business absorb the loss the way Meta absorbed the Metaverse? Probably, but it would be a far costlier mistake:
I am using Reality Labs’ operating loss as a proxy, but it gives a sense of the scale. If this bet does not pay off, it will mean more than a mere blip for the stock price.
So What?
When Mark Zuckerberg announced his pivot to the metaverse and the staggering capital it would consume, I sold my entire position, which I had held for many, many years. For a while I looked like a genius: I sold at a blended $260 across October 2021 to February 2022, and Meta then collapsed 72% to $89 by November 2022.
But look at it now. Meta is such a cash-flow machine that it essentially shrugged off the entire misadventure. Reality Labs, the metaverse division, has run up roughly $83 billion in cumulative operating losses since 2020, about $19 billion in 2025 alone. And yet, set against more than $200 billion of revenue and north of $40 billion of free cash flow last year, the market barely blinked. A staggering sum of misinvestment; a blip for the business as a whole.
So was selling the right call? In this case, yes: I used the proceeds to add to NVIDIA that May at $18 a share (split-adjusted). Holding Meta would have produced a gain too, albeit much smaller. But tooting my own horn about a prescient trade is not the point. I could easily have been wrong about what I bought; I knew I was right about what the Metaverse spending signaled about management. That is why I sold, and reinvested with a more disciplined CEO, Jensen.
Don’t Get in a Knife Fight
So should you sell the Magnificent Seven? I think that comes down to your honest assessment of each core business, and to the tax bill, which for a decade of compounding gains is no small consideration. I am not ready to pull the plug on these names entirely. But I have been steadily trimming the ones I own, and I am focused on the discipline and reinvestment track record of management for the names I continue to hold. In subsequent earnings calls, I want to hear a narrative around how they are using targeted investment in AI to strengthen their competitive position and their plans for monetization. Otherwise, I may be forced to exit my positions entirely. After all, there is only one way to win a knife fight… don’t get in a knife fight.