CapEx and D&A measure two very different things, and the gap between them is the whole story here.
CapEx is a cash outflow that shows up in the investing section of the cash flow statement the moment the company pays for the plant — it never touches the income statement directly. D&A is the opposite: it's a non-cash expense on the income statement that spreads the original cost of an asset over its useful life, and it gets added back on the cash flow statement because no cash actually left in that period.
In Years 1 and 2, the company is paying cash for a plant that doesn't exist yet from an accounting standpoint — it hasn't been placed in service, so it generates zero D&A and zero incremental revenue. Meanwhile, $100M a year in growth CapEx is leaving the building. That's a $100M annual cash drag that EBITDA margin completely hides, because EBITDA is calculated before CapEx and D&A both — a 22% EBITDA margin says nothing about how much of that EBITDA actually converts to free cash.
By Year 3, the plant comes online and starts depreciating — but only $10M a year, because its $200M cost is spread over 20 years. D&A rises from $50M to $60M, while CapEx drops back down to $50M of maintenance spend. The gap flips from a $100M drag to roughly breakeven, but D&A still lags the cash that was actually spent — that catch-up plays out over the full 20-year depreciation schedule, long after the cash left.
Eventually, once the company stops expanding and is just replacing worn-out equipment, maintenance CapEx and D&A converge — that's the textbook “CapEx = D&A” assumption used in DCF terminal value, and it only holds in that mature, no-net-growth state. Assuming it during an active expansion phase like this one would significantly overstate near-term free cash flow.