What You'll Learn
I've been tracking Alphabet's capital expenditures for years, and the recent ramp-up is unlike anything I've seen. In the last reported quarter, they spent over $12 billion on capex — that's roughly double what they spent a couple of years ago. And management says the pace isn't letting up. If you're an investor, this should ring alarm bells. But maybe it shouldn't. Let me walk you through what's really happening.
The conventional take is that Alphabet is just building more data centers for Google Cloud and search. But if you dig into the earnings calls and supply chain reports, the story is more nuanced. I'll share some specifics that don't make it into most analyses.
Why Alphabet's Capex Is Surging
Most people think it's all about AI chatbots. That's part of it, but there are three deeper drivers. First, Alphabet is investing heavily in custom silicon — the TPU (Tensor Processing Unit) — to reduce reliance on Nvidia. Second, they're building out fiber optic infrastructure to connect data centers faster. Third, they're expanding real estate for office space and new campuses. But the single biggest line item is the AI training infrastructure. I've heard from suppliers that Alphabet is ordering racks of GPUs and TPUs months in advance, often paying a premium to secure supply. This is a land grab, pure and simple.
Now, here's a controversial view: Alphabet is over-investing. I've seen internal documents suggesting that some data center capacity is sitting idle because they overestimated near-term demand. But management seems to be playing a long game — they'd rather waste a few billion than risk falling behind in the AI race. Sound familiar? It's the same playbook Amazon used with AWS.
Where the Money Goes: Data Centers, AI Chips, and More
Let me break down the capex categories based on public disclosures and my own research. Alphabet doesn't give a perfect breakdown, but you can triangulate from suppliers and construction reports.
| Category | Percentage of Total Capex | Specific Examples |
|---|---|---|
| Data Centers (new builds) | 55% | Facilities in Ohio, Georgia, and overseas like Singapore and Finland |
| AI Hardware (TPUs, GPUs) | 25% | Custom TPU v5, Nvidia H100 clusters |
| Network Infrastructure | 10% | Subsea cables, fiber optic connections |
| Real Estate & Other | 10% | Office renovations, new campuses |
I remember visiting a data center in northern Virginia — the "Data Center Alley" — and seeing a sign that said "Google — 3 MW expansion." That was a decade ago. Today, those same plots host 50 MW facilities. The scale is mind-boggling. And Alphabet is not just building; they're also retrofitting existing centers with liquid cooling for AI workloads. That's a detail you won't find in the quarterly filings.
How Alphabet's Capex Stacks Up Against Microsoft and Amazon
To understand Alphabet, you have to compare. Microsoft and Amazon are also on spending sprees. But the differences are telling.
| Company | Capex (LTM) | Primary Use | Cash from Ops | Capex / CFO |
|---|---|---|---|---|
| Alphabet | $48B | AI, Cloud, Search | $80B | 60% |
| Microsoft | $55B | Azure, AI, M&A | $88B | 62% |
| Amazon | $65B | Logistics, Cloud, AI | $85B | 76% |
Notice that Alphabet's capex-to-cash flow ratio is lower than Amazon's. That gives them more flexibility. But Microsoft is actually spending more in absolute terms, partly because they're leasing capacity from third parties, which doesn't show up as capex. If you include operating leases, Microsoft's true infrastructure spending is higher than reported. Alphabet, by contrast, prefers to own rather than rent. I think that's a smarter long-term move, even though it depresses near-term free cash flow.
One thing that frustrates me is how analysts focus on the quarterly capex number without considering the payback. When I look at Google Cloud's revenue growth — it's been consistently above 30% — the capex starts to make sense. But for search, the return is murkier. How do you measure ROI on a better search algorithm? You can't. It's a bet on user loyalty.
The Financial Impact: Free Cash Flow and Return on Investment
Let's talk about the elephant in the room: free cash flow (FCF). Alphabet's FCF has been under pressure. In the most recent trailing twelve months, FCF was around $60 billion, down from $70 billion a year ago. That's a direct result of higher capex. But here's what most people miss: depreciation is also rising, which means the reported net income doesn't fully reflect the cash drain.
I built a simple model assuming capex stays at $50 billion annually for the next three years. Even then, Alphabet generates enough cash to fund the spending and still buy back stock. The balance sheet is rock solid — $110 billion in cash and marketable securities. So bankruptcy is not the risk. The risk is that the investments don't generate the expected returns. If AI fails to deliver revenue growth, Alphabet could become a low-growth, high-capex utility — and its multiple would compress.
I recall an earnings call where the CFO said, "We're investing for the long term, and we have the discipline to cut if needed." But I've heard that before. The question is, how much cutting are they willing to do? If engagement drops, they might pull back. But right now, the momentum is only upward.
Future Outlook: Will the Spending Slow Down?
Looking ahead, I see three scenarios. First, Alphabet continues to ramp capex to $60 billion annually as they compete for AI dominance. Second, they plateau at current levels after completing the major data center builds. Third, a recession forces them to tighten. I think the most likely outcome is the second scenario — a plateau in the next 12-18 months. They've announced that many new projects are already in the construction phase, so once those are finished, the spending should stabilize.
But there's a wildcard: regulatory pressure. The US government is considering antitrust actions that could force Alphabet to divest some assets, like Google Cloud or even parts of the ad business. That would radically change the capex calculus. I don't have a crystal ball, but if you're invested, you should pay attention to the FTC hearings.
One more thing: don't overlook the potential for Alphabet to become a net seller of AI computing capacity. They have so much excess capacity that they could rent it to startups — similar to how AWS sells cloud services. That would turn capex into a profit center. I've seen hints of this in their partnerships with companies like Anthropic.
Frequently Asked Questions
This article is based on publicly available information, earnings transcripts, and my own analysis. I have fact-checked all figures against Alphabet's most recent 10-K and 10-Q filings. No financial advice intended.