How to Answer “How would you forecast Capex and depreciation in a DCF model?” in Investment Banking Interviews

Start by stating what Capex represents and how it should behave over time. In a DCF, Capex is a core reinvestment driver, so I first separate the narrative into growth Capex (capacity expansion, new sites, step-ups for new product lines) and maintenance Capex (spend needed to sustain current revenue and asset condition).

Practically, I look at historical Capex as a percentage of revenue and compare it to depreciation (Capex/D&A) to understand whether the company has been investing above replacement levels. If management guidance exists, I use it for the near years and then fade to a steady-state assumption consistent with margins, market growth, and asset intensity. For a mature business, I’d expect maintenance Capex to trend closer to depreciation over the long run; for a scaling business, Capex can exceed depreciation for several years.

I call out any step-changes explicitly (new plant in year 2, data centre build-out, regulatory spend) rather than hiding them inside a flat percentage-of-sales assumption.

Then I pick the most defensible method given the case and time. Common approaches:

• Percent of revenue (or percent of incremental revenue): Fast, interview-friendly, and often good for high-level valuation interview prep. I’ll usually start here and sanity-check against history and peers.
• PP&E roll-forward / asset-intensity approach: Forecast revenue, assume an asset turn (Revenue/Net PP&E) or Capex-to-sales ratio implied by the target PP&E base, then back into Capex. This is useful when capacity constraints matter.
• Project-based Capex schedule: For discrete build-outs (mines, telecom networks, manufacturing lines), I model known projects explicitly and then add a maintenance layer.

Whichever method I use, I make sure Capex is timed realistically (lumpy for projects) and I separate Capex from acquisitions unless the prompt specifically includes inorganic growth.

For depreciation, I avoid treating it as an independent plug. The clean approach is to link it to the asset base:

• Build a simple PP&E schedule: beginning Net PP&E + Capex − depreciation (and optionally disposals).
• Forecast depreciation either as a percent of beginning/average gross PP&E, or more commonly in interviews, as a percent of beginning Net PP&E based on the historical D&A rate.

If I have enough detail, I translate this into a rough useful life: if depreciation is ~10% of gross PP&E, that suggests ~10-year average life. Then I ensure the implied relationship between Capex and depreciation makes sense over time.

In many DCFs, depreciation is mainly needed to get EBIT, taxes, and unlevered free cash flow right; the key is that it stays consistent with Capex and doesn’t drift to unrealistic levels (e.g., D&A rising while PP&E shrinks).

Next, I connect the lines correctly in the DCF build:

• In unlevered free cash flow: NOPAT = EBIT × (1 − tax rate), then add back D&A (non-cash), subtract Capex (cash outflow), and subtract increases in NWC.
• Capex and depreciation should be internally consistent so the DCF isn’t artificially inflated or deflated.

I also explain what this means for valuation: higher sustained Capex reduces free cash flow and can lower value even if EBITDA looks strong; higher depreciation lowers EBIT and taxes (increasing cash taxes shield), but it doesn’t directly reduce cash flow because it’s added back.

When the interviewer broadens into “walk me through a DCF,” I explicitly note that these reinvestment assumptions will materially impact the implied cash flows going into discounting at WACC and, ultimately, the terminal value.

I close with quick checks that show modelling maturity:

• Capex vs D&A in steady state: In the terminal period, maintenance Capex often trends toward depreciation for stable businesses. If Capex is permanently below D&A, PP&E will shrink indefinitely, which usually contradicts stable revenue.
• Implied asset intensity: Check Revenue/Net PP&E (or Capex as % of sales) vs history and peers.
• Terminal year consistency: Ensure the terminal-year reinvestment rate supports the terminal growth assumption (if using a perpetuity method) and doesn’t imply impossible economics.

If numbers look off, I’ll adjust the driver (Capex % sales, depreciation rate, or asset turns) rather than forcing a plug. That’s the difference between a mechanical model and an interview-ready, explainable one.