The tech industry is in the middle of the most aggressive spending spree we have seen in decades. The four major hyperscalers—Microsoft, Alphabet, Amazon, and Meta—are on track to drop over $700 billion into capital expenditures in 2026. Money is pouring directly into data center real estate, power grids, custom chips, and massive liquid-cooling setups.
If you follow the markets, you’ve probably felt the shift in tone. Wall Street used to cheer every single time a CEO announced a new AI initiative. Now, investors are asking a much harder question: When are we going to see a real return on all this cash?
Here is a plain-English breakdown of what is actually happening behind the scenes, how the accounting works, and when these massive bills will show up on corporate balance sheets.
The Numbers: A Spending Spree Like No Other
To grasp why investors are suddenly nervous, look at how much cash the big four are burning to stay ahead in the AI arms race:
| Company | Estimated 2026 Capex | Where the Cash Is Going | The Main Risk |
| Amazon | ~$200B+ | AWS data centers, custom Trainium chips, power supply | Capex is consuming a massive chunk of operating cash flow |
| Microsoft | ~$190B | Azure AI clusters and OpenAI infrastructure | Piles of hardware waiting to be depreciated |
| Alphabet | ~$175B–$185B+ | Custom TPU chips, Gemini training, external data centers | Posted its first major cash burn quarter in mid-2026 |
| Meta | ~$130B–$145B | Open-source Llama model clusters, data center sites | Huge capital outlays without direct cloud rental income |
The "D&A Lag": Why the Financial Damage Is Delayed
When a tech giant spends $20 billion buying server racks and GPUs today, that $20 billion doesn't instantly show up as a loss on their earnings report.
Instead, corporate accounting rules allow them to capitalize the expense. They put the physical hardware on their balance sheet and spread out the cost over its useful life—usually 4 to 6 years—using a line item called Depreciation & Amortization (D&A).
This creates a serious time delay that can trick casual observers:
The Cash Leaves Now: Tens of billions in cash exit the company’s bank accounts today.
The Income Statement Looks Fine Today: Today’s quarterly report only shows a tiny fraction of that hardware cost.
The Cumulative Wall: As these companies stack hundreds of billions in new hardware year after year, those small quarterly depreciation charges start compounding into massive annual expenses.
If software revenues don't grow fast enough to cover that compounding depreciation wall, profit margins take a direct hit.
Wall Street Is Dividing Tech Into Two Camps
Investors aren't panicking across the board, but they are being much pickier. Wall Street has split big tech into two distinct groups:
The Cloud Monetizers (Amazon & Microsoft): These companies have an immediate way to collect cash. They rent compute power directly to startups, banks, and enterprise clients. Every time a business runs an AI model on Azure or AWS, the meter runs. Because cloud revenue is actively surging, Wall Street gives them more leeway on spending.
The High-Risk Gamblers (Meta & Consumer Apps): Companies that build massive AI models without a direct cloud-rental business have a much harder narrative to sell. They are spending institutional-level money up front, hoping that better ad targeting or smarter chatbots will eventually make up for it down the line.
When Will This Show Up in Quarterly Earnings?
We are already starting to see the early cracks. In mid-2026, Alphabet reported a quarterly cash burn that caught analysts off guard, sending a quick shockwave through tech valuations.
The full impact on quarterly margins will likely peak over the next 12 to 24 months as three things happen:
Depreciation Hits Peak Levels: The massive wave of 2024–2026 GPU purchases will hit their maximum annual depreciation drag, forcing operating margins lower if software sales stall out.
Off-Balance-Sheet Leases Come Due: To keep balance sheets looking clean, several tech giants signed massive off-balance-sheet data center lease agreements. When those facilities go live, rent expenses will hit operating costs directly.
Power Grid Costs Surge: Data centers are running out of cheap power. Tech firms are having to pay high premiums to secure dedicated power contracts, driving up baseline operational costs.
What This Means for Investors
The takeaway isn't that AI is a bubble—the demand for compute power is undeniably real. However, the days of free cash flow running wild while companies spend infinitely without scrutiny are over.
If you are looking at tech stocks today, stop focusing just on top-line revenue growth. Keep a close eye on Free Cash Flow (FCF) and Operating Margins. The winners won't just be the companies with the most GPUs; they'll be the ones that actually figure out how to monetize them before the depreciation bill comes due.