The DCF Verdict on AI Datacenter Stocks: Why NVDA, AVGO, and SMCI Valuations Diverge by 1,000%

Our DCF roundup on AI datacenter names last week hit a 5.26% engagement rate — the highest of the week. You asked for the full analysis with real numbers and an explanation of the methodology. Here it is.

We ran three separate valuations on NVIDIA, Broadcom, and Super Micro Computer using live data from the EvidInvest database and FMP. The results do not just disagree — they diverge by hundreds of percent. NVIDIA alone ranges from $275 (Wall Street consensus) to $588 (our normalized model) to $1,654 (raw CAGR extrapolation).

That spread is not a bug. Understanding why the same framework produces such different outputs tells you more about these three stocks than any single price target.

---

The Numbers at a Glance

Prices as of May 6, 2026. EvidInvest model uses 20% annual EPS growth × 50x forward PE. CAGR extrapolations apply historical growth rates to current EPS/FCF over a 5-year horizon with 10% discount rate. Data: EvidInvest database, FMP.

---

Step 1: The Historical CAGRs — What These Businesses Actually Did

Before any model, look at the actual compounded growth rates from financial statements.

*SMCI FCF was negative in FY2024 due to working capital buildup. FCF CAGR uses FY2023→FY2025 (2yr) to avoid the negative year.

NVIDIA grew revenue 8× in four years — from $26.9B in FY2022 to $215.9B in FY2026. Free cash flow grew even faster: from $8.1B to $96.7B, a 4yr CAGR of 86%. SMCI grew 4×. Even AVGO, the steadiest of the three, doubled revenue in three years.

These are not normal growth rates. They are the reason every valuation model that plugs them in produces extreme outputs.

---

Method 1: Raw CAGR Extrapolation — What the Math Produces

The simplest approach: take the historical CAGR, apply it to current EPS for five years, multiply by a target PE, and discount back at 10%.

Formula: Fair Value = EPS × (1 + CAGR)^5 × Target PE × discount factor

NVDA at 68% revenue CAGR applied forward comes out at $1,654. SMCI at 62% comes out at $304. AVGO at 24% comes out at $363 — the only one that lands near its current price.

No analyst puts $1,654 on NVIDIA. These numbers are not price targets. They are the mathematical output of assuming that exceptional historical growth rates compound uninterrupted for five more years. No business does that. But the math tells you something important: at these growth rates, standard valuation frameworks have to be rewritten. The entire debate for NVDA and SMCI is not whether they are cheap — it is whether the growth rate persists.

The AVGO result is the most instructive. A 24% revenue CAGR applied forward gives $363 — below the current $427 price. Even raw extrapolation of historical growth says AVGO is near fair value. That is a different kind of signal than NVDA or SMCI.

---

Method 2: EvidInvest Normalized Model — The Reality Check

EvidInvest's valuation model applies a uniform 20% annual EPS growth rate across all three stocks, regardless of historical CAGR. This is a deliberate normalization — an explicit assumption that exceptional growth rates mean-revert toward a sustainable long-run rate.

Formula: Fair Value = Current EPS × (1.20)^5 × 50x PE

Is 20% a reasonable assumption? The answer depends on the company:

For AVGO, 20% is close to reality — the actual revenue CAGR is 24% and FCF CAGR is 18%. The normalized model and actual growth rates agree, which is why AVGO shows a modest +37% upside rather than the extreme numbers produced for NVDA and SMCI.

For NVDA, 20% is a major haircut from the actual 68% revenue CAGR. It is explicitly pricing in a sharp deceleration from the AI GPU cycle. Even at that conservative assumption, the model still shows nearly 200% upside — which tells you how far ahead of its historical earnings growth the stock is trading.

For SMCI, 20% is again well below the actual 62% CAGR. But this is arguably appropriate — and the reason is not the growth rate itself.

---

Method 3: Wall Street Consensus — What Analysts Price In

Source: FMP price target consensus as of May 6, 2026.

Consensus has NVDA at +40%, AVGO at +4%, and SMCI at +44%. These are 12-month price targets from institutional research, not multi-year DCF outputs. They are a useful near-term anchor but tell a different story than any intrinsic value model.

The most striking number is SMCI at $40.20 consensus versus $211 on EvidInvest's model. That gap is not a modeling disagreement. It is the subject of the next section.

---

Why SMCI Is a Different Kind of Problem

Super Micro Computer grew revenue at 62% annually for three years. Its quarterly revenue hit $12.7B — against a total market cap of less than $17B. A price-to-sales ratio below 0.4x for a company growing 47% annually is, on its face, extraordinary.

Yet Wall Street's consensus target is $40. Not $211. Not $238. $40.

The reason is not the growth rate. It is the trust discount.

Between 2024 and 2025, SMCI faced: its auditor (Ernst & Young) resigning, delayed 10-K filings, an SEC investigation into accounting practices, and a Department of Justice subpoena. The stock crashed from highs above $100 to under $30 — not because server demand collapsed, but because investors can no longer price a business whose reported numbers are under formal scrutiny.

Wall Street analysts are not modeling SMCI wrong on the growth side. They are applying a risk discount that a CAGR model cannot capture. Remove the regulatory overhang and SMCI's 62% revenue CAGR would command a multiple in line with comparable high-growth hardware names. The implied re-rating upon resolution is one of the larger binary outcomes in the AI infrastructure space.

But it is a binary. Resolution is violent upside. Escalation — restated financials, penalties, management changes — creates no valuation floor that a growth model can defend.

---

The Consensus Revenue Path vs. 20% Normalization

For completeness: here is what Wall Street consensus actually expects for revenue growth at each company — and how that compares to the EvidInvest 20% normalized assumption.

NVIDIA — Consensus is MORE bullish near-term than our model:

Consensus expects NVDA to grow 70% in FY2027 before decelerating sharply. The EvidInvest 20% model is not aggressive relative to consensus in the near term — it is pricing in a faster normalization. Even so, 20% sustained for five years generates a $588 fair value. Which path do you believe?

---

The Three Different Bets

All three methods, laid out together:

NVDA — Every model shows undervaluation at the current price of $196.50. Consensus sees +40%. Our 20% model sees +199%. Raw CAGR extrapolation produces numbers that are not serious price targets but confirm the growth magnitude. The debate is not whether NVDA is a good business — it is whether the AI GPU cycle sustains long enough for the earnings to grow into any of these fair values.

AVGO — The cleanest result. Historical CAGR (24%) matches the 20% model. Consensus ($443) implies 4% upside from $427 — nearly at fair value. All three methods converge. No extreme valuation gap, no regulatory overhang, no binary risk. It is the steadiest AI infrastructure trade, and the valuation math confirms it.

SMCI — The business grew 62% annually. The stock is at $28. The gap between $28 and $211 is entirely explained by regulatory risk, not by any dispute about the growth rate. If the SEC investigation resolves favorably, the re-rating is substantial. If it escalates, no growth model offers protection. This is not a valuation question — it is a risk underwriting question.

---

The Takeaway: What You Are Actually Choosing

When you buy one of these three stocks, you are not just making a bet on a business. You are choosing which set of assumptions to believe.

Raw CAGR extrapolation says: the historical growth rate is the best predictor of future growth. It produces extreme outputs for NVDA and SMCI, and near-fair-value for AVGO.

The normalized 20% model says: hypergrowth always reverts. Even extraordinary businesses mean-revert to a sustainable long-run rate. At 20%, all three still show upside, but the magnitude is anchored.

Wall Street consensus says: near-term earnings estimates drive 12-month price targets. Buy what analysts have modeled, ignore the long-run assumptions.

None of these is right or wrong in isolation. They are frameworks for organizing different beliefs about the future. The right question is not "which model should I use?" — it is "which assumptions do you actually believe?"

Run your own growth scenarios on any stock at EvidInvest →

---

Frequently Asked Questions

Why does NVIDIA's fair value differ so much across models?

Because NVDA's historical growth rate (68% revenue CAGR over 4 years) is so far outside normal ranges that small changes in the assumed forward growth rate produce enormous differences in fair value. A model that uses 68% for 5 years gives $1,654. A model that uses 20% gives $588. The math is sensitive to the growth input precisely because the compounding is so powerful at those rates.

Is EvidInvest's 20% growth assumption conservative for NVDA?

Yes — significantly. Wall Street consensus expects NVDA to grow revenue 70% in FY2027 before normalizing. The 20% assumption is deliberately conservative, pricing in much faster deceleration than consensus. Even so, 20% compounded over 5 years with a 50x PE still yields a $588 fair value, which is 200% above current price.

Why is SMCI priced so much lower than its growth rates suggest?

SMCI's accounting issues are the primary driver. Between 2024 and 2025, the company's auditor resigned, SEC launched an investigation, and DOJ issued a subpoena over accounting practices. The stock fell from above $100 to below $30. The 62% revenue CAGR is real — but investors cannot price a business whose reported numbers are under formal regulatory scrutiny. The valuation discount reflects risk, not growth.

Which of these three stocks has the best risk-adjusted return?

Based on the data, AVGO offers the most straightforward case: historical CAGR matches the normalized model, consensus is roughly aligned, and there are no regulatory or governance overhangs. NVDA has more absolute upside but requires a view on AI cycle duration. SMCI has the highest potential upside if regulatory issues resolve, and binary downside risk if they do not.

---

All prices as of May 6, 2026. Revenue estimates from FMP analyst consensus. Historical financials from FMP income statement and cash flow statement data. EvidInvest fair values from production database (aggressive scenario: 20% growth, 50x PE, 5-year horizon). This is not investment advice — always conduct your own due diligence.