Alphabet's Capex Doubled, Its Free Cash Flow Didn't: A Fresh Look at Google's Valuation

Back in December we published a deep dive into the AI capital-spending boom (the concern Michael Burry had put on the table) and argued it would eventually pressure the valuations of the hyperscalers doing the spending. We made that concrete with Microsoft, concluding it was overvalued at $490 and taking plenty of grief for it. Six months later Microsoft played out almost exactly as we described: the stock is down about a third from its peak, not because earnings missed but because the market re-rated it for permanently elevated capex.

Microsoft was only the worked example. The same logic applies to every hyperscaler pouring money into AI, a point we came back to when the whole group's capex plans landed at the start of the year. Alphabet is squarely in that camp and spending as hard as anyone, so here we turn the same capex lens on Google's own valuation. The honest answer our model produces is uncomfortable: even on generous assumptions, Alphabet looks expensive. That is a strange thing to say about a stock trading at roughly 27 times earnings, well below Microsoft's old multiple, so it is worth walking through carefully.

Alphabet's capex situation

Start with the cash flows, because that is where the story is. Alphabet's capital spending as a share of revenue sat in the low teens for years. Over the last two it has erupted.

Table 1. Alphabet capital spending and cash flow, FY2020-FY2025.

Fiscal year	Capex / revenue	Operating cash flow	Free cash flow
FY2020	12.2%	$65.1B	$42.8B
FY2021	9.6%	$91.7B	$67.0B
FY2022	11.1%	$91.5B	$60.0B
FY2023	10.5%	$101.7B	$69.5B
FY2024	15.0%	$125.3B	$72.8B
FY2025	22.7%	$164.7B	$73.3B

The two right-hand columns are the whole story. In FY2025 operating cash flow surged 31% in a single year, from $125.3 billion to $164.7 billion, a staggering step-up for a company this size. Free cash flow did not follow: it went from $72.8 billion to $73.3 billion, essentially flat. All $39 billion of that extra operating cash, and then some, disappeared into capital spending, which nearly doubled from about $52.5 billion to $91.4 billion in a single year and climbed from 15.0% to 22.7% of revenue. Alphabet is generating far more cash than ever, and keeping almost none of it.

And it is accelerating. Alphabet has guided 2026 capital spending to roughly $175-190 billion, more than double the 2025 figure (these guidance numbers are from company outlooks and earnings coverage; the FY history above is from Alphabet's filings as we hold them). The trend is industry-wide: across 2026 the four big hyperscalers (Microsoft, Alphabet, Amazon and Meta) have collectively guided to roughly $725 billion of capital spending, up about 77% year over year.

This is what matters for valuation. On earnings, Alphabet's roughly 27x multiple looks cheap. But earnings capitalise the AI build-out: the $91 billion of capex sits on the balance sheet and reaches the income statement only slowly, as depreciation. Free cash flow is where the full bill lands, and free cash flow is what a DCF values. A cheap-looking P/E and an expensive-looking DCF are not a contradiction; they are two views of the same capex, one that defers it and one that does not.

How our model treats capex

Our model handles capex in two stages, and the second one does most of the work:

1. In the explicit forecast years, capex glides from the company's recent (elevated) normalized rate down toward maintenance, rather than sitting flat at today's level. Maintenance capex is roughly equal to depreciation: it is what replacing worn-out assets actually costs, as opposed to building new capacity.
2. In the terminal value, capex is set to maintenance. A company cannot spend 20%-plus of revenue on capex forever; at a long-run growth rate near 3%, its asset base would eventually outgrow the entire economy. In a genuine steady state, capital spending converges toward replacement.

Crucially, this glide-down is the generous treatment. It assumes the current AI surge is temporary and that Alphabet's capital intensity falls back toward a normal level over time. We think that is the right base case. But it is the assumption that flatters the valuation, not the one that punishes it. You do not have to take our word for which view is right: every DCF page on the platform carries a "Hold Capex Elevated" control that flips between the two, so you can set the assumption yourself and watch the fair value move.

What the model says now

Holding everything else constant and flexing only the long-run capex assumption, here is where Alphabet's intrinsic value lands. The latest price is $360.

Table 2. Alphabet intrinsic value under different long-run capex assumptions.

What you assume long-run capex does	Fair value	vs price
Glides down to maintenance (our default model)	$252	-30%
Glides from the FY2026 pace (37%) down to maintenance	$213	-41%
Stays permanently elevated at 18% of revenue	$160	-55%
Glides from the FY2026 pace (37%) down to that elevated 18%	$126	-65%

Every row sits below the current price. Our default model, which assumes the AI surge normalizes over time, puts fair value near $252, about 30% under the $360 price. Start capex at Alphabet's own FY2026 guided pace, roughly 37% of revenue, and still glide it all the way down to maintenance, and you get $213. Assume capex never falls back to bare maintenance but settles permanently at its recent run-rate of around 18% of revenue, which is what the platform's "Hold Capex Elevated" switch does, and fair value is $160. Run both the 2026 spike and that permanently-elevated 18%, arguably the most realistic bearish path, and the two compound to $126. Whichever way you turn it, Alphabet comes out worth less than it trades for; the disagreement is only about how much.

Because every capex path here is bearish, it is worth being explicit about the other inputs, none of which are doing the bearish work. Revenue growth is the analyst consensus, not our guess: roughly 17% a year from FY2025 revenue to the Street's 2030 estimate, which is a generous path. The discount rate is a market-derived 9.6% WACC, built from a 1.15 beta (our own regression against the broad market), a 4.5% risk-free rate off the current 10-year Treasury, and a 4.5% equity risk premium; Alphabet carries almost no debt, so that is essentially its cost of equity. Terminal growth is 3.5%, about long-run nominal GDP. Nudge growth higher or the discount rate lower and the fair value rises, but the capex assumption still dominates. It is the single input that moves Alphabet from expensive to very expensive.

How this compares with other DCF models

Our number is not an outlier because we discount harder or lowball growth. It is where a cash-flow model tends to land once it takes the capex seriously. Here is how our fair value sits against other public DCF-based estimates, all captured in early July 2026 at a price near $360.

Table 3. DCF-based fair value estimates for Alphabet from various providers, early July 2026.

Source	DCF fair value	vs price ($360)
Stockoscope (our default model)	$252	-30%
MiniValuator (base case)	$259	-28%
Alpha Spread (base case)	$308	-14%
Simply Wall St (2-stage FCF)	$361	roughly fair

Two things stand out. Every one of these cash-flow models lands at or below the current price; none of them calls Alphabet a bargain on cash flow. And the ranking follows the assumptions: ours and MiniValuator's, which both use around 17-18% growth and a roughly 10% discount rate, cluster near $255, while the more generous models drift up toward the price. Which number you trust is really a single judgment, whether the AI spending converts into free cash flow or just keeps compounding.

So where does that leave us

A DCF is one lens, not a verdict. Alphabet is an extraordinary business, with search, YouTube, a fast-growing cloud, the Waymo and DeepMind option value, and a balance sheet stuffed with cash. None of that is captured fully by a free-cash-flow model that is, deliberately, pricing the capex bill in full. If the AI build-out earns strong returns, today's spending is an investment that the model's "maintenance" framing understates, and the bears (us included, on this measure) will look too cautious.

But that is precisely the bet embedded in the price. At $360 the market is paying up for a company whose free cash flow has gone sideways for two years while capital spending doubled, and whose own guidance is for spending to roughly double again. The roughly 27x P/E that makes Alphabet look reasonable is computed on earnings that have not yet absorbed that cash outflow. On the cash itself, our model cannot find a capex scenario that justifies the current price.

So our honest take: on the numbers we hold, Alphabet screens as roughly 30% overvalued on our default model, and overvalued across every capex scenario we ran. That does not make it a short, and it does not mean the stock cannot keep rising; great businesses trade above intrinsic value for years. It means the margin of safety is negative right now, and the thing that would change that is either a lower price or evidence that the AI spending is about to convert into a step-change in free cash flow rather than just more capex.

As always, this is a candidate for your own research, not a recommendation. The model on Alphabet's valuation page recomputes from live data, and the capex switch is right there for you to test your own assumptions. Flip it, and see where you land.