Last month I argued that market concentration creates risk. Two weeks ago I argued that your own psychology makes that risk worse. Intellectual honesty demands the third piece: the strongest version of the argument I'm skeptical of. Here's the best case for sustained tech dominance — and the precise list of what would have to stay true for it to hold.
The best investors don't just know why they might be right. They understand exactly how they might be wrong.
So this piece argues against my own recent writing — not because I've changed my mind, but because the case for sustained tech dominance deserves its strongest telling, not a strawman version that's easy to knock down. If you're going to hold a concentrated index anyway (and if you own an S&P 500 fund, you are), you should know precisely what you're betting on.
Here is the steelman.
The bull case: why this time might actually be different
These aren't speculative dot-coms. They're cash machines. The 2000 bubble was inflated by companies with no earnings, no sustainable business models, and no moats — Pets.com was worth billions, and revenue was optional. Today's leaders are categorically different. Nvidia's most recent quarter (Q3 of its fiscal 2026, reported in November) showed record revenue of $57.0 billion, with data center revenue of $51.2 billion — up 66% year over year. Microsoft's Azure has been growing north of 30% year over year. Apple generates over $100 billion in annual services revenue alone. These companies aren't promising future profits; they're generating historic ones right now.
The software moat is real — and possibly durable. Nvidia's advantage isn't just silicon. It's CUDA, a software ecosystem built over nearly two decades that has become the de facto operating system of AI development, with more than four million developers trained on it. For an enterprise, migrating workflows, codebases, and talent off the CUDA standard is enormously costly and complex — which is exactly what Warren Buffett means by a moat. AMD, Intel, and the hyperscalers' custom chips are making genuine progress on hardware, but hardware was never the whole moat.
The infrastructure commitments are locked in. This demand isn't a consumer fad that can evaporate in a quarter. OpenAI has committed to deploying at least 10 gigawatts of Nvidia systems. Anthropic plans to adopt a gigawatt of next-generation Nvidia infrastructure. Microsoft, Amazon, and Google are pouring capital expenditure into AI data centers on multi-year schedules. These are contractual, industrial-scale commitments that sustain demand regardless of quarterly sentiment.
Laid out this way, the bull case is genuinely compelling. I want to be honest about that. Now comes the discipline.
What would have to remain true
A portfolio concentrated in these companies for the next ten to twenty years is a bet on all of the following staying true, simultaneously:
AI spending must keep growing. Enterprises are investing ahead of proven returns. If the productivity gains disappoint — if AI turns out to be transformative but not that transformative, on that timeline — capital spending moderates, and the demand story with it.
Margins must stay elevated. Nvidia's gross margins run above 70% — software-like margins on hardware. Competition from AMD, Intel, and the hyperscalers' own chips exists precisely to compress them.
No technological regime change. The bet assumes the next computing paradigm — quantum, neuromorphic, something not yet imagined — either doesn't arrive or happens to favor the incumbents' architecture. Paradigm shifts have a poor record of asking permission.
Valuations must be sustained. Today's leaders trade at multiples far above the broader market — priced, in other words, for something close to perfection, for a long time.
Any one of those failing doesn't mean the companies fail. It means the stocks reprice — and repricing is what a retirement portfolio experiences.
What history says about "permanent" moats
Every era has its unassailable leaders, and the language never changes.
In the early 1970s, the Nifty Fifty — IBM, Xerox, Polaroid, Avon, Coca-Cola — were "one-decision" stocks: buy and hold forever. Great businesses, real earnings, genuine moats. They traded around 42 times earnings, more than double the market. From their 1972–73 highs to the 1974 lows, Xerox fell 71%, Avon 86%, Polaroid 91%. Most of the companies were fine. The stocks took a decade or more to recover, if they recovered at all.
In 2000, Cisco was the backbone of the internet — roughly 40% share of high-growth networking equipment, genuinely transformative technology, a competitive position everyone described as unassailable. All of that was true. The stock peaked near $77 in March 2000, and as of late 2025 — a quarter century later — it has never regained that level. The internet won. The bet on the internet's most dominant company, made at that price, still lost.
That's the pattern worth memorizing: the technology succeeding and the stock succeeding are two different bets, and valuation is the difference between them.
The bottom line
Maybe this time really is different. The bull case above is the strongest I can construct, and parts of it may well hold.
But if you're within ten years of retirement, "maybe" isn't a strategy — because the consequences are asymmetric. If tech dominance continues, a diversified portfolio still participates meaningfully in it; you give up some upside, not the outcome. If it doesn't, a concentrated portfolio suffers disproportionately, at exactly the moment sequence risk makes losses hardest to recover. Heads you win slightly less; tails you lose what you can't rebuild.
The best investors don't bet on certainty. They build portfolios that survive multiple scenarios — including the scenario where the smartest-sounding argument of the decade turns out, like its predecessors, to have been priced for a perfection that didn't arrive.
If you'd like to see which scenario your own portfolio is actually built for, that's what I'm here for. You can grab a time on my calendar — no pressure, just clarity.
Dan
Sources
• Nvidia Q3 FY2026 results (November 2025) — revenue and data center figures from contemporaneous earnings coverage.
• CUDA ecosystem and developer base: Klover.ai analysis; moat and infrastructure-commitment analysis.
• Nifty Fifty history and declines: Border to Coast; Bridgeway Capital; Wikipedia.
• Concentration-peak performance research: Silvercrest.
This article is educational and general in nature. It is not individualized investment, tax, or legal advice, and it does not account for any individual's specific circumstances. Company figures are as reported for the periods cited; past performance is not indicative of future results. Investing involves risk, including possible loss of principal. Nothing here is a recommendation regarding any security. Current as of December 2025.
Dan Fagan, MSPFP, MPAS™, AIF®
Managing Partner, Sr. Wealth Management Advisor, Radiant Wealth Management
Author, Getting to Emeritus (Johns Hopkins University Press, August 2026)
dan.fagan@radiantwm.com | Office: 475.234.5965 | Cell: 860.748.2552
Scheduling: app.greminders.com/c/danfagan
Radiant Wealth Management is a fee-only practice serving faculty, physicians, and academic professionals. Advisory services offered through NewEdge Advisors, LLC, a registered investment adviser.
About Dan Dan Fagan, MSPFP, MPAS™, AIF®, is a Senior Wealth Management Advisor and Managing Partner at Radiant Wealth Management in Connecticut. With over 25 years of experience, including 17 years at TIAA as a Senior Wealth Management Advisor, Dan specializes in helping academics and physicians retire smarter, plan with clarity, and align their financial lives with what matters most.
This article is for educational purposes only and does not constitute financial, tax, or legal advice. Individual circumstances vary. Please consult qualified professionals for advice specific to your situation. Advisory services are offered through NewEdge Advisors, LLC, a registered investment adviser. Advisory services are only offered to clients or prospective clients where NewEdge Advisors and its representatives are properly licensed or exempt from licensure.
