What jumps out to me in the raw data is not just scale, but the speed with which NVIDIA has turned scale into near-unprecedented profitability. Annual revenue went from $60.9B in FY24 to $130.5B in FY25 to $215.9B in FY26; net income went from $29.8B to $72.9B to $120.1B over the same span. That is not a normal semiconductor ramp; it is a platform toll-taker phase. The latest quarterly run-rate is even more extreme: revenue of $81.6B and net income of $58.3B imply a 71.5% net margin, which is so high that I treat it as peak economics, not a stable base. Even if you smooth that out, the business is throwing off absurd cash relative to capital needs: $102.7B operating cash flow, $96.7B free cash flow, and only $6.0B of capex. Balance-sheet risk is basically absent with $10.6B cash against $11.0B debt and a current ratio near 3.9. So the core story I see is a company with real monopoly-like economics today, but whose valuation now rests less on proving quality and more on proving durability.

That durability is where I get more cautious than the headline growth suggests. Yes, the quarter-to-quarter revenue staircase from $30.0B to $81.6B is extraordinary, and yes, the most recent YoY growth of 65.5% on such a huge base is still elite. But the stock is already capitalized at $4.77T, about 22.4x trailing sales and 40.2x annual earnings on the canonical numbers. Even if you prefer the lower trailing P/E implied by annualizing recent quarters, you are still paying a megacap premium for margins that are likely above mid-cycle and possibly above sustainable-cycle. FY26 gross margin was 71.1%, operating margin 60.4%, net margin 55.6%; those are phenomenal, but also leave very little room for disappointment. My read is that NVIDIA no longer needs to merely grow fast; it needs to avoid any material reset in pricing power, mix, or hyperscaler urgency. This is not fragile financially, but it is fragile narratively because so much of the equity value is tied to keeping these exceptional margins unusually high.

Opus argues the “quarterly trajectory is actually re-accelerating, not decelerating,” and on the raw quarterly sequence I agree. The latest step from $68.1B to $81.6B is a stronger sequential move than the prior two quarters, so any blanket “deceleration” label is too crude. I also agree with Opus that the insider-selling alarm is not supported by the ledger provided here: a 500K-share gift and repeated award entries are not evidence of bearish insider exits. But I disagree with Opus’s more important conclusion that “at $197 and ~$4.77T, NVDA is not priced for perfection — it’s priced for very good.” I think that understates what is embedded. At 22x sales and roughly 36x EV/EBITDA for a company already doing $216B annual revenue, the market is not paying for merely “very good”; it is paying for several more years of extraordinary economics. If revenue growth settles from 65% YoY toward 20-25% and net margins normalize from the current 55-70% zone toward something closer to 40-45%, the valuation support becomes much thinner than Opus suggests.

I also disagree with Opus’s fair-value band of $180-230 because it leans too heavily on annualizing a blowout quarter and too lightly on cyclicality. Opus says even 25-30% CAGR through FY29 with margin normalization to 55% gets you to fair value around here. That 55% net margin assumption is exactly where I balk. NVIDIA’s FY26 full-year net margin was 55.6%, and the latest quarter was 71.5%; using 55% as a normalized destination is not conservatism, it is close to assuming today’s exceptional economics persist almost intact. In semis, even dominant franchises rarely hold this level once customers optimize spend, competitors improve, and supply constraints ease. On the other hand, I agree with Opus that the real fault line is hyperscaler capex reflexivity. With revenue now so dependent on a small set of giant buyers, one pause in AI infrastructure digestion could hit both growth and margin at once. That is why I would not call the stock a short either; the business quality is too high, the balance sheet too strong, and the earnings power too real. But I also would not call it “fairly valued” in a complacent sense.

A careful skeptic of both my view and Opus’s would say we are both still anchored to old semiconductor intuitions that may not fit this asset. If CUDA/software lock-in and inference demand turn NVIDIA into a quasi-infrastructure utility for AI, then 22x sales on today’s numbers could look reasonable in hindsight, and margin normalization may be far slower than bears expect. That skeptic would also note that trailing valuation metrics can compress very fast when earnings are growing this quickly; a company earning $120B one year and potentially far more the next can outrun “expensive” labels. Fair. But the skeptic should also admit the reverse: once a company is worth $4.8T, even great execution can produce mediocre stock returns if the entry multiple already discounts the next leg.