The ten largest stocks in the S&P 500 now account for roughly 40 to 45 percent of the index's total market capitalization, according to data compiled by Bloomberg and Charles Schwab, exceeding the peak concentration seen at the height of the dot-com bubble in 2000, when the top ten held about 27 percent. Nvidia trades around 38 times forward earnings on Bloomberg consensus numbers. Ray Dalio told Bloomberg Television in June that his proprietary bubble indicators are rising close to, though not yet matching, the readings he saw in 1929 and 2000.
I built a DCF on Oracle earlier this year, and I want to use it honestly here, including being direct about where that model's scope stops short of the exact question this concentration data raises.
What my Oracle model actually captured
My Oracle valuation was built from a bottom up weighted average cost of capital, 10.16 percent, applied to a cash flow forecast tied to Oracle's cloud ERP transition. The model landed on an intrinsic value of $197.72 a share, about 15 percent above where the stock traded at the time. That gap came from specific, defensible assumptions about margin expansion in a real product transition.
Here's the part I want to be direct about. Oracle now sits inside the exact circular financing arrangement that's driving a meaningful share of current market concern, OpenAI committing hundreds of billions to Oracle and Microsoft for compute, those hyperscalers buying chips from Nvidia, Nvidia investing back into the labs that then spend more on compute. My cloud ERP model was not built to isolate or stress test that AI infrastructure and compute commitment revenue specifically. If a meaningful share of Oracle's current valuation is now resting on that circular arrangement rather than the ERP transition my model priced, then my existing DCF doesn't settle the question of whether today's Oracle valuation is defensible. It answers a narrower, earlier question well. It doesn't answer the current one.
An estimated $1.4 trillion in announced AI infrastructure commitments through 2030 depends on AI revenue that has not yet materialized.Bain & Company, April 2026
Why that gap is the actual lesson, not a problem to paper over
Goldman Sachs' James Covello has been publicly skeptical of the AI trade for a related reason, questioning whether the market's pricing can be supported by underlying cash flows rather than continued capital raising. I don't think the honest answer to whether this is a bubble is a single word. Historical concentration episodes, 1929, the Nifty Fifty in 1973, dot-com in 2000, all corrected somewhere between 40 and 80 percent within roughly two years of peaking, and none were obvious in real time to the people living through them.
What a disciplined DCF gives you in an environment like this isn't a verdict handed down once and reused. It's a method that has to be rerun honestly every time the underlying revenue mix changes, which is exactly what's happened to Oracle since I built mine. My model is genuinely useful evidence that Oracle's core ERP transition was fundamentally sound at the time I priced it. It is not evidence, on its own, that Oracle's valuation today, now meaningfully exposed to AI infrastructure demand that hasn't yet proven durable, is equally sound. The discipline that matters isn't having built one model. It's being willing to say plainly when that model no longer covers the question being asked of it, and rebuilding it rather than stretching its conclusion to cover ground it was never built to reach.