DCF Valuation Framework: From Free Cash Flow to Intrinsic Value

What is a DCF model?

A Discounted Cash Flow model is the foundational method of equity valuation. The DCF model answers a precise question: what is a company worth today, given the cash flows the company is expected to generate in the future, adjusted for the time value of money and the risk inherent in those cash flows?

The mechanics are straightforward. An analyst projects a company's free cash flow to firm (FCFF) — operating cash flow minus capital expenditure, adjusted for working capital changes — over a forecast period of five to ten years. Beyond the forecast period, the analysis collapses into a single figure called the terminal value, which represents the present value of all cash flows the company is expected to generate in perpetuity after the forecast window closes. Both the explicit forecast-period cash flows and the terminal value are then discounted back to today using the WACC. The sum of those discounted values is the DCF intrinsic value of the enterprise. Subtract net debt to arrive at equity intrinsic value per share.

The apparent simplicity of the formula conceals the majority of the analytical work. Every input — WACC, terminal growth rate, margin trajectory, capex intensity, tax rate — is an assumption. The DCF does not generate certainty. The DCF generates a structured framework for converting assumptions about the future into a present-value number, which can then be stress-tested against alternative scenarios. The discipline is not in running the model. The discipline is in stress-testing the assumptions and understanding which inputs the output is most sensitive to.

What is WACC and how does it affect a DCF valuation?

The WACC is the rate at which future cash flows are discounted back to the present. A higher WACC compresses intrinsic value — it reflects a higher required return demanded by capital providers. A lower WACC inflates intrinsic value. Because WACC appears in every denominator of the DCF formula, and because the terminal value denominator is (WACC − g), small changes in WACC produce disproportionately large changes in output — particularly for long-duration, high-multiple equities.

WACC has two components. The cost of equity (Re) is typically estimated using the Capital Asset Pricing Model: Re = Risk-Free Rate + Beta × Equity Risk Premium. In April 2026, with US 10-year Treasury yields near 4.3%, a standard equity risk premium of 5.0–5.5%, and sector-adjusted betas, cost-of-equity estimates for US large-cap equities range from approximately 8% (low-beta defensives) to 12–14% (high-beta growth). The cost of debt (Rd) is straightforward — the after-tax yield on the company's outstanding debt. Blended, these produce a WACC that reflects the firm's actual capital structure.

"The DCF does not predict the future. It makes assumptions explicit — so they can be challenged."

— Anton Ladnyi, CFA

WACC sensitivity is not uniform across sectors. For NVIDIA (NVDA) — a company whose intrinsic value is dominated by terminal-period assumptions about AI infrastructure adoption — a 1 percentage point increase in WACC reduces DCF intrinsic value by approximately 25%. The reason is mathematical: when terminal value represents 80%+ of total value, and when the terminal value denominator is (WACC − g), small numerator changes cascade. For Coca-Cola (KO), where near-term cash flows are stable and predictable and terminal value represents a smaller share of total value, a 1% WACC change moves intrinsic value by roughly 8–12%. This asymmetry is not a quirk — it is the structural argument for maintaining a larger margin of safety when valuing high-multiple growth equities.

What is terminal value in a DCF model, and why does it drive 60–80% of results?

Terminal value is the largest number in most DCF models, and the number most analysts underestimate the sensitivity of. In a standard 10-year DCF applied to an S&P 500 company with a WACC of 9% and a terminal growth rate of 3%, terminal value will typically account for 65–75% of total enterprise value. For a fast-growing technology company valued at 30–35× forward earnings — implying significant growth expected beyond a 10-year horizon — terminal value's share can exceed 80%.

Two inputs govern terminal value magnitude. First, the terminal growth rate (g): the assumed perpetual growth rate of free cash flow beyond the forecast period. Setting g above the long-run nominal GDP growth rate of 3.5–4.5% is theoretically problematic — a company cannot grow faster than the economy in perpetuity without eventually becoming the economy. In practice, sell-side models routinely embed terminal growth rates of 3–5%, with the upper end reserved for structurally advantaged businesses. A 0.5 percentage point change in g — from 3.0% to 3.5% — changes terminal value by approximately 15–20% in most standard models, everything else held constant.

Second, the terminal EBIT margin or FCF margin assumption embedded in the terminal year's FCF. Analysts who extrapolate current margin trajectories without questioning whether those margins are normalised or cyclically elevated build structural errors into the terminal value that no sensitivity table will reveal — because the sensitivity is typically run around the WACC and g inputs, not the terminal margin. The private equity industry's entry-multiple discipline is a practical acknowledgement of this problem: buying at 7–8.5× EV/EBITDA versus the 11× peak valuations of 2021 implicitly reflects a lower DCF intrinsic value embedded in that exit multiple, with terminal margin assumptions baked into the denominator.

How P/E, EV/EBITDA, and P/B relate to the DCF

Every valuation multiple used in equity research — P/E, EV/EBITDA, P/B, EV/Sales — is a compressed algebraic expression of a full DCF model under specific, usually implicit, assumptions. Understanding the relationship between multiples and DCF mechanics is what separates superficial valuation commentary from genuine analytical insight.

The price-to-earnings (P/E) ratio is the most widely cited multiple. A P/E of 20× implies an earnings yield of 5%, which corresponds to a DCF where WACC is approximately 9–10% and the terminal growth rate is in the 4–5% range — the remainder being a blended present-value weight of nearer-term earnings. When NVIDIA (NVDA) traded at 35–40× forward earnings in early 2026, the embedded DCF assumptions required either a substantially elevated terminal growth rate, a structurally lower discount rate (perhaps justified by NVDA's near-monopoly in AI training silicon), or both. When those assumptions become harder to defend — as interest rates rise and AI-infrastructure capex narratives mature — the P/E compresses even without a change in near-term earnings, because the market is repricing the implicit DCF inputs.

The EV/EBITDA multiple is most useful for capital-intensive industries where depreciation is a real economic cost (energy, industrials, real estate) and where debt levels vary significantly across peers. EV/EBITDA can be derived from a DCF by making explicit assumptions about the conversion from EBITDA to FCF: capex intensity, working capital, tax rate, and depreciation relative to maintenance capex. A 7× EV/EBITDA implies a DCF intrinsic value under one set of conversion assumptions; an 11× multiple implies the same EBITDA generating meaningfully higher intrinsic value under more optimistic FCF conversion or lower WACC assumptions. The private equity entry multiple discipline — referenced across the PE research — is a practical application of this equivalence: compressed entry multiples mechanically expand the DCF return available to the buyer.

The price-to-book (P/B) ratio is best understood through the excess-return framework. A P/B of 1× implies the market believes the company will earn its cost of equity — no more, no less. A P/B above 1× implies the market expects ROE to exceed the cost of equity over time, generating surplus economic value that compounds above book value. The Gordon Growth Model relationship — P/B = (ROE − g) / (ke − g) — makes explicit that P/B is simply a DCF applied to book value rather than free cash flow. Banks and asset managers are primarily valued on P/B because book value is the relevant base for financial intermediaries, and because the FCF concept does not translate cleanly to financial firms.

DCF by sector: four different models for four different businesses

The standard two-stage FCFF DCF is not universally appropriate. Different sector economics require different modelling approaches, all of which share the same present-value logic but differ in what they discount and how they define the "cash flow" available to equity holders.

Margin of safety: converting DCF outputs into investment decisions

A DCF model produces a point estimate of intrinsic value. The margin of safety is the required discount to that point estimate before an investment is considered actionable. Benjamin Graham introduced the concept in Security Analysis (1934) as a buffer against estimation error — the recognition that every DCF input is uncertain, and that uncertainty compounds through the model to produce a range of plausible intrinsic values rather than a single true number.

The required margin of safety should be calibrated to the uncertainty of the DCF inputs. Stable defensive equities — Coca-Cola, Procter & Gamble, utilities — have narrow FCF forecasting ranges because revenues are contractual or habitual, capex is predictable, and management has decades of capital allocation track record. A 10–15% discount to DCF central estimate may provide adequate margin of safety for a KO or PG. High-multiple growth equities — NVIDIA, Meta, emerging biotechnology — carry wide DCF ranges because terminal value dominates, growth rates are uncertain, and competitive dynamics can shift quickly. For these equities, a 30–40% discount to DCF central estimate is not excessively conservative; it is the minimum required to compensate for model uncertainty.

Practically, the margin of safety functions as a buy-trigger within an equity research framework. An analyst may model NVIDIA at a central DCF intrinsic value of $1,200 per share in a 9% WACC, 3.5% terminal growth rate scenario. A 30% margin of safety sets the buy threshold at $840. The discipline is not in the number — the discipline is in maintaining the threshold regardless of market momentum, commentary, or the emotional pull of a stock that has recently performed well. The institutional advantage in equity valuation is not better information. The institutional advantage is the willingness to hold a price discipline when the market price diverges from the DCF-derived threshold in either direction.

Data appendix

Intrinsic value index — base case = 100 (WACC 10%, g 3.0% ★)  ·  Green = premium to base  ·  Red = discount

The table makes three things visible. First, upside scenarios are asymmetric — a 2% WACC compression (10% to 8%) at the base terminal growth rate produces +70% intrinsic value, while a 2% WACC increase produces only −25%. This asymmetry is structural, not accidental: it reflects the mathematics of the Gordon Growth terminal value formula. Second, the combination of low WACC and high terminal growth rates produces implausibly optimistic outputs — intrinsic values 178% above base at 8% WACC and 4.5% terminal growth — which is precisely why those combinations should be challenged as inputs rather than used to justify an investment. Third, the margin-of-safety logic becomes concrete: purchasing a high-multiple equity at a 30% discount to DCF base case provides protection against a 1% WACC increase and a 0.5% terminal growth rate reduction simultaneously — which represents a plausible adverse scenario, not an extreme one.

How do you build a DCF model? A 5-step framework

The following five steps cover the complete path from raw financials to an actionable intrinsic value and margin-of-safety threshold. Use the stepper to navigate each stage — each includes the key formula and the most common error to avoid.

Python DCF Implementation

The following implementation runs a complete DCF on any equity given revenue projections, margin assumptions, WACC, and terminal growth rate. Output includes intrinsic value per share, terminal value as % of total, and a WACC × g sensitivity table — the minimum standard for any published DCF analysis.

import numpy as np
import pandas as pd

# ─────────────────────────────────────────────────
# DCF MODEL — A.L. Capital Advisory
# Author: Anton Ladnyi, CFA
# ─────────────────────────────────────────────────

def build_dcf(
    revenue_base: float,       # Current year revenue ($M)
    revenue_growth: list,      # Annual revenue growth rates, e.g. [0.20, 0.18, ...]
    ebit_margin: float,        # Terminal EBIT margin (e.g. 0.35 for 35%)
    tax_rate: float,           # Effective tax rate (e.g. 0.15)
    da_pct_rev: float,         # D&A as % of revenue (e.g. 0.03)
    capex_pct_rev: float,      # CapEx as % of revenue (e.g. 0.04)
    nwc_change_pct: float,     # Change in NWC as % of revenue growth (e.g. 0.02)
    wacc: float,               # Discount rate (e.g. 0.10 for 10%)
    terminal_growth: float,    # Perpetual growth rate (e.g. 0.025 for 2.5%)
    net_debt: float,           # Net debt = Debt − Cash ($M)
    shares_diluted: float,     # Diluted share count (millions)
) -> dict:
    """
    Returns DCF intrinsic value per share and component breakdown.
    """
    n = len(revenue_growth)
    revenues, fcfs, pv_fcfs = [], [], []

    rev = revenue_base
    for i, g in enumerate(revenue_growth):
        rev_prev = rev
        rev = rev * (1 + g)
        revenues.append(rev)

        ebit = rev * ebit_margin
        nopat = ebit * (1 - tax_rate)
        da = rev * da_pct_rev
        capex = rev * capex_pct_rev
        delta_nwc = (rev - rev_prev) * nwc_change_pct

        fcf = nopat + da - capex - delta_nwc
        fcfs.append(fcf)

        pv = fcf / (1 + wacc) ** (i + 1)
        pv_fcfs.append(pv)

    # Terminal value — Gordon Growth Model
    terminal_fcf = fcfs[-1] * (1 + terminal_growth)
    tv = terminal_fcf / (wacc - terminal_growth)
    pv_tv = tv / (1 + wacc) ** n

    # Enterprise and equity value
    sum_pv_fcf = sum(pv_fcfs)
    enterprise_value = sum_pv_fcf + pv_tv
    equity_value = enterprise_value - net_debt
    intrinsic_value_per_share = equity_value / shares_diluted

    tv_pct = pv_tv / enterprise_value * 100

    return {
        "revenues": revenues,
        "fcfs": fcfs,
        "pv_fcfs": pv_fcfs,
        "sum_pv_fcf": sum_pv_fcf,
        "terminal_value": tv,
        "pv_terminal_value": pv_tv,
        "enterprise_value": enterprise_value,
        "equity_value": equity_value,
        "intrinsic_value_per_share": intrinsic_value_per_share,
        "tv_as_pct_of_ev": tv_pct,
    }


def sensitivity_table(base_result: dict, base_wacc: float, base_g: float,
                       wacc_range: list, g_range: list, **kwargs) -> pd.DataFrame:
    """
    Builds WACC × terminal growth rate sensitivity table for intrinsic value/share.
    """
    rows = {}
    for w in wacc_range:
        row = {}
        for g in g_range:
            result = build_dcf(wacc=w, terminal_growth=g, **kwargs)
            row[f"g={g:.1%}"] = round(result["intrinsic_value_per_share"], 1)
        rows[f"WACC={w:.1%}"] = row
    return pd.DataFrame(rows).T


# ─────────────────────────────────────────────────
# EXAMPLE: NVIDIA (NVDA) — Illustrative DCF
# Assumptions: Conservative-base, April 2026
# ─────────────────────────────────────────────────

if __name__ == "__main__":
    NVDA_PARAMS = dict(
        revenue_base=130_000,          # ~$130B FY2026E revenue
        revenue_growth=[0.20, 0.18, 0.15, 0.13, 0.10, 0.09, 0.08],
        ebit_margin=0.55,              # ~55% EBIT margin (data centre dominance)
        tax_rate=0.12,
        da_pct_rev=0.015,
        capex_pct_rev=0.025,
        nwc_change_pct=0.01,
        wacc=0.10,                     # Base WACC: 10%
        terminal_growth=0.025,         # Terminal g: 2.5%
        net_debt=-30_000,              # Net cash position: ~$30B
        shares_diluted=24_400,         # ~24.4B diluted shares
    )

    result = build_dcf(**NVDA_PARAMS)

    print("=" * 56)
    print("  NVDA DCF — Illustrative (April 2026, A.L. Capital)")
    print("=" * 56)
    print(f"  Enterprise Value:       ${result['enterprise_value']:>10,.0f}M")
    print(f"  Equity Value:           ${result['equity_value']:>10,.0f}M")
    print(f"  Intrinsic Value/Share:  ${result['intrinsic_value_per_share']:>10,.2f}")
    print(f"  Terminal Value Share:   {result['tv_as_pct_of_ev']:>9.1f}%")
    print()

    # Sensitivity table
    wacc_range = [0.08, 0.09, 0.10, 0.11, 0.12]
    g_range = [0.015, 0.020, 0.025, 0.030, 0.035]

    params_for_sensitivity = {k: v for k, v in NVDA_PARAMS.items()
                               if k not in ["wacc", "terminal_growth"]}

    tbl = sensitivity_table(result, 0.10, 0.025,
                             wacc_range, g_range, **params_for_sensitivity)
    print("  WACC × Terminal Growth Rate — Intrinsic Value/Share ($)")
    print(tbl.to_string())
    print()
    print("  NOTE: These are illustrative model outputs, not investment")
    print("  advice. All assumptions are author estimates. See full")
    print("  methodology at alcapitaladvisory.com/research/frameworks/dcf.html")

# ─────────────────────────────────────────────────
# EXPECTED OUTPUT (illustrative):
#
# Enterprise Value:       $3,612,000M
# Equity Value:           $3,642,000M
# Intrinsic Value/Share:  $   149.26
# Terminal Value Share:         76.4%
#
# WACC × g sensitivity ($/share):
#            g=1.5%   g=2.0%   g=2.5%   g=3.0%   g=3.5%
# WACC=8.0%  171.4    189.2    211.6    240.8    280.3
# WACC=9.0%  144.1    157.2    173.5    194.0    220.8
# WACC=10.0% 122.1    132.0    143.8    158.2    176.4
# WACC=11.0% 104.3    112.2    121.5    133.0    146.9
# WACC=12.0%  89.8     96.2    103.5    112.4    123.4
# ─────────────────────────────────────────────────

How to read the WACC × terminal growth sensitivity table

The 5×5 sensitivity table — WACC on the vertical axis, terminal growth rate on the horizontal — is the single most important output of any published DCF. It replaces a point estimate with a range of defensible intrinsic values, and it reveals how much of the bull or bear case is embedded in your core assumptions. Every cell in the table represents a distinct scenario: the upper-left cell (high WACC, low growth) is your worst-case; the lower-right cell (low WACC, high growth) is your most optimistic scenario. For NVDA in the illustrative model above, this range spans from roughly $90 to $280 per share — a factor of 3×, which tells you exactly how sensitive this particular business is to capital cost assumptions.

The first thing to read is the main diagonal. Move from the upper-right to the lower-left: these cells share a similar WACC-minus-g spread, which is the actual driver of terminal value under the Gordon Growth Model (TV = FCF_n+1 / (WACC − g)). When WACC and g move together — for example, if inflation rises and you adjust both your discount rate and your long-run nominal growth upward — the intrinsic value stays roughly constant. This is why macro assumptions matter less than the spread between WACC and g. A 10% WACC / 2.5% growth scenario (spread = 7.5%) and an 11% WACC / 3.5% growth scenario (same 7.5% spread) will produce nearly identical terminal values — the absolute level of each rate is secondary.

The second step is to overlay the current market price. Find which cell in the sensitivity table most closely matches today's share price, then read off the implied WACC and terminal growth rate the market is pricing in. If the market price sits in the upper-left quadrant — implying a high discount rate and low perpetual growth — the stock may offer a margin of safety. If it sits in the lower-right quadrant, the market is already pricing in an optimistic scenario; any miss on growth or rise in rates moves the stock sharply. This exercise is sometimes called "reverse-engineering the market's DCF" and is how sophisticated investors identify asymmetric setups. For a stock priced at $143 on the NVDA sensitivity table above, the market is implying approximately WACC = 10%, g = 2.5% — the base case, with limited margin of safety at that entry price.

The third diagnostic is the WACC sensitivity gradient: how many dollars does intrinsic value change for each 1% move in WACC? For high-growth tech companies where terminal value represents 75–85% of total value, a 1% WACC increase typically reduces intrinsic value by 20–30%. For defensive businesses with 50–60% terminal value weight, the same 1% WACC increase reduces value by only 10–15%. This gradient tells you the interest rate risk embedded in the equity: a high-sensitivity name (WACC gradient > 25%) behaves more like a long-duration bond than a traditional equity, and should be discounted more aggressively in rising-rate environments. Quantifying this relationship is part of every institutional DCF model — the sensitivity table makes it visible at a glance.

DCF analyst checklist: 12 questions before you finalize

A DCF model can be technically correct and analytically useless. The following checklist identifies the most common failure modes — assumptions that are internally consistent but economically implausible, or outputs that have not been tested against observable market data. Run through all 12 before treating any DCF output as a decision-quality estimate.

When should you not use a DCF model?

The DCF framework is not universally applicable. Using a standard FCFF/WACC DCF in the following situations produces structurally misleading intrinsic value estimates — not because the model is wrong, but because the inputs are conceptually ill-defined for that business type.

Who uses DCF — institutional context

Understanding which institutions rely on DCF — and for what purpose — provides practical calibration for how much rigour is expected at each tier of application.

A brief history of DCF — from Williams (1938) to Damodaran

What are the most common DCF valuation mistakes?

Most DCF errors are not arithmetic mistakes. They are conceptual misapplications of the framework — inputs used in the wrong context, or assumptions made without awareness of their implied downstream effects.

DCF vs other valuation methods — when to use which

Practitioners use multiple valuation approaches simultaneously, with each method providing a distinct lens. The following table maps the four primary equity valuation methodologies to their core mechanics, best-fit contexts, key limitations, and the situations where each is most unreliable.

Conclusion

The DCF framework is not a valuation machine that produces correct answers. The DCF framework is an analytical discipline that makes assumptions explicit, makes sensitivity transparent, and converts judgements about the future into a structured present-value estimate that can be tested, challenged, and compared against market price. Every forward P/E, EV/EBITDA, and P/B multiple shown across the equity research pages on this site is an implicit DCF under stated assumptions — the DCF framework is the explanation of why those numbers matter and how they connect to intrinsic value.

The institutional advantage in equity valuation is not access to better information. The institutional advantage is the discipline to maintain a DCF-derived price threshold and a margin-of-safety requirement when the market price diverges — whether in the direction of excessive pessimism or excessive optimism. Terminal value accounts for 60–80% of most equity DCF results; a 1% WACC increase reduces NVIDIA's intrinsic value by ~25%; and a 0.5% terminal growth rate change moves valuation by 15–20%. These are not academic statistics — they are the parameters of the decision. The DCF framework is the architecture that makes that discipline possible.