Let me run the reverse-DCF on NVDA right now.

The verdict from the reverse-DCF: NVDA is actually UNDER-PRICED vs. what analysts expect — but "under-priced" here still requires extraordinary growth. Let me unpack this.

What the current price demands:

At $202.81 and a 63% net margin, the price is only "demanding" a 21.9% revenue CAGR over 5 years to deliver a 10%/yr return to you. That sounds like a lot — but here's the twist: analysts (after fading their own optimism) are modeling 34.3% CAGR. The gap is -12.4 percentage points — meaning the street thinks NVDA will grow nearly 60% faster than what the price already assumes. That's the "under-priced" call.

But the base rate reality check is sobering.

Only 17.5% of companies in history that were asked to deliver 21.9% revenue growth over 5 years actually did it. Even the lower bar — what the price demands — only 1 in 6 companies historically cleared it. NVDA's own 5-year historical CAGR is 66.9% and 10-year is 45.7%, so they've massively beaten that bar before, but sustaining it from a $250B+ revenue base is a completely different challenge.

The scenario grid tells the real story:

If analysts are right (~34% CAGR), you're looking at +21% annualized at the base exit multiple — that's exceptional. But if NVDA merely delivers what the price implies (~22% CAGR — which most companies can't even do — you only earn your 10% hurdle. The downside scenario if growth disappoints: a 5.2% annualized return at a compressed multiple. Basically dead money at best.

The key assumption to stress-test: the 18x exit PE.

The model uses a mean-reverted 18x exit PE — that's being generous in assuming NVDA doesn't trade at a premium forever. NVDA today trades at ~35x forward. If in 5 years the market re-rates it to 25x (still a premium), the numbers look much better. If it compresses to 18x as the model assumes (treating it like a mature cyclical), you need the growth to carry the whole return.

What gives the bull case real fuel:

The research signals are meaningful here — LLM efficiency research has an attention score of 1,032, and HBM/memory stacking at 505. Both map directly to NVDA's GPU and CUDA ecosystem. These aren't lagging indicators — academic and corporate research attention tends to lead product cycles. 298 SEC filings in 90 days mention LLM architecture, meaning every major company is embedding NVDA dependency into their forward plans.

Also: the margin assumption is conservative but risky. The model locks in 63% net margins — NVDA's current level. That's already historically unprecedented for a hardware company. If margins compress at scale (as AMD/Intel compete, as custom silicon from hyperscalers grows), the model is too optimistic on profitability.

My take: The "growth already priced in" question has a nuanced answer here. At $202, the price needs less growth than analysts model — which is the bull case in a nutshell. But you're still betting on a company sustaining 20%+ revenue growth from a quarter-trillion dollar revenue base, something only 1 in 6 companies has historically pulled off. NVDA isn't a typical company (66.9% actual 5-year CAGR proves that), but the law of large numbers is real. This isn't priced for perfection — it's priced for continued dominance, which is more achievable. I'd call it fairly priced for a believer, under-priced if analysts are even half right.

You already hold AVGO (13.2% of your portfolio) which gives you indirect AI infrastructure exposure. NVDA would be a higher-conviction, higher-volatility complement if you wanted to add AI GPU direct exposure.