EV to EBITDA

The enterprise multiples decompress into the DCF's firm-side variables. The bridge is the value-driver logic from the DCF forecasting work. In steady state, free cash flow to the firm is NOPAT minus net reinvestment, and reinvestment rate = g/ROIC chains reinvestment to growth. So:

FCFF1 = NOPAT1 × (1 − g/ROIC)          and          EV = FCFF1 / (WACC − g)

⇒  EV = NOPAT1 × (1 − g/ROIC) / (WACC − g)

Every enterprise multiple is this one equation divided by a different denominator.

EV/EBITDA: decompressed

Write NOPAT in EBITDA terms. Let d = D&A/EBITDA (the depreciation intensity), so EBIT = EBITDA × (1−d) and NOPAT = EBITDA × (1−d)(1−t):

Forward EV/EBITDA = (1 − d) × (1 − t) × (1 − g/ROIC) / (WACC − g)

Four drivers, each with a story:

• d, the capital intensity. It's the share of EBITDA that the fixed-asset base eats up. Higher d means a lower multiple, because the market pays for cash after the machines are fed.
• t, the tax take. Two identical firms in 25% and 35% jurisdictions deserve different EV/EBITDA. EBITDA is pre-tax, and value is post-tax.
• g and ROIC, growth held in check by returns. The (1−g/ROIC) term is the reinvestment toll from the DCF forecasting work. Growth costs capital, and the higher the ROIC, the lighter the toll.
• WACC, the rate regime and risk.

What is a good EV/EBITDA multiple?

Try it on a US industrial with d = 0.30, t = 25%, g = 4%, ROIC = 16% and WACC = 8.5%. The reinvestment factor = 1 − 4/16 = 0.75. Justified forward EV/EBITDA = 0.70 × 0.75 × 0.75 / 0.045 = 0.394/0.045 = 8.75×. That's reassuringly close to where average US industrials actually trade. Now move one dial at a time:

Same EBITDA, same growth, multiples from 5× to 11×, and all three correctly priced. It's the most misread fact in markets. A telecom at 6× isn't "cheaper" than a staffing firm at 11×. EBITDA simply means less where D&A eats more of it. So when someone shows you a stock that's "cheap on EBITDA", your first question is now automatic: what is d, and what is the ROIC?

Run the India check too. Put the same base firm under India parameters, WACC 11.5% and g 5%. The reinvestment factor becomes 1 − 5/16 = 0.6875, and the firm justifies 0.70×0.75×0.6875/0.065 = 5.6×. Indian EV/EBITDA multiples carry a structural rate handicap. The ones that beat US peers anyway are leaning hard on growth and ROIC, and the section on India's premium multiples shows how hard.

EV/EBIT versus EV/EBITDA: when D&A is a real cost (always, eventually)

Divide the same EV by EBIT instead and d drops out:

Forward EV/EBIT = (1 − t) × (1 − g/ROIC) / (WACC − g)

That's one fewer distortion. So why does EBITDA dominate practice? There are three reasons you can defend, and one you can't. (1) D&A is hostage to accounting choices (useful lives, revaluations, purchase-price allocations), so EBITDA compares operations across firms whose depreciation policies differ. (2) It's capital-structure-proof further up the P&L. That's handy in leveraged deals where the buyer will re-lever everything anyway. It's LBO country, and it comes back when you build deal models. (3) In a business that really is asset-light, the two multiples nearly coincide because d is small. The one you can't defend is that EBITDA makes bad businesses look better, and sell-side materials gravitate to whichever line flatters.

Munger's verdict ("every time you see the word EBITDA, substitute 'bullshit earnings'") overstates, but in the right direction. Depreciation is a real cost that arrives with a lag. The machines wear out, and the capex that replaces them is as mandatory as wages. Price a trucking company on EBITDA and you've priced it as if trucks were immortal.

So here's the working rule: the heavier and longer-lived the asset base, the further you must move from EBITDA toward EBIT. For any business past its growth phase, replacement capex ≈ D&A, which makes EBIT the truthful operating line. One refinement is worth borrowing from the McKinsey house style. EBITA is earnings before interest, taxes, and amortization of acquired intangibles, but after depreciation. Depreciation is a real recurring cost. Amortization of a customer list acquired in a deal isn't, because no one re-buys the customer list every year. For acquisitive companies, EV/EBITA beats both its siblings. Wherever you see EV/EBIT from here on, treat "adjust out acquired-intangible amortization where material" as a standing instruction.

EV/Sales: a margin forecast wearing a disguise

Divide the master equation by revenue. NOPAT = revenue × EBIT margin × (1−t), so:

Forward EV/Sales = EBIT margin × (1 − t) × (1 − g/ROIC) / (WACC − g)

EV/Sales is the EV/NOPAT multiple times a margin, so every EV/Sales number is a margin expectation in disguise. A steady business with a 20% EBIT margin, t 25%, g 4%, ROIC 16% and WACC 8.5% justifies 0.20 × 0.75 × 0.75/0.045 = 2.5× sales. Keep that as your anchor. A good, mature business is worth roughly 2 to 3× revenue, and every turn above that is a claim about margins or growth that you should force into the open.

The algebra does the forcing. Say a SaaS firm trades at 10× EV/Sales, and assume it matures to WACC 9%, g 4.5% and light reinvestment (1 − g/ROIC ≈ 0.9). What steady-state EBIT margin is priced in? Solve 10 = m × 0.75 × 0.9/0.045, so m = 10 × 0.045/0.675 = 67%. That's a sixty-seven percent EBIT margin, above the best software company that has ever existed at scale. Microsoft runs in the mid-40s.

So 10× sales is not a steady-state price. It's a bet on years of hyper-growth before the steady state. Go and price that bet with the two-stage machinery from the terminal-value work, instead of nodding along to "10× is normal for SaaS." EV/Sales is the multiple of last resort, for loss-makers, young firms and turnarounds where E, EBITDA, and B are meaningless or negative. The habit it demands never changes. State the mature margin you're underwriting, and check it against the best-ever base rate for that business model.

EV/IC and the value-creation plot

Divide the master equation by invested capital. NOPAT1 = ROIC × IC0, so:

EV/IC = ROIC × (1 − g/ROIC)/(WACC − g) = (ROIC − g) / (WACC − g)

That's the enterprise mirror of P/B = (ROE−g)/(r−g), and the same anchors hold. If ROIC = WACC, EV/IC = 1, because a rupee invested is worth a rupee when it earns only its keep. Excess returns push it above 1, and growth amplifies whichever side of 1 the spread puts you on. With ROIC 20%, WACC 8.5% and g 4%, EV/IC = 16/4.5 = 3.6×. The market pays ₹3.6 for every ₹1 inside the business, because each rupee mints 20 paise against an 8.5-paise hurdle.

The practical weapon is the value-creation plot. For a sector, scatter EV/IC on the y-axis against ROIC or ROIC/WACC on the x-axis. Theory says the points should climb a line anchored near (WACC, 1.0×). Firms far above the line are being paid for more value creation than their returns show, so the market is either pricing an improvement or overpaying. Firms below it are cheap, or else the market disbelieves the durability of their reported ROIC. Every dot off the line is a research question. Later, in the work on value drivers, this plot becomes home ground for judging capital allocation. You'll meet the equity edition of the same idea in the comps regression, with P/B against ROE for banks.