Key takeaways
- ~$700-750 billion — the estimated combined 2026 AI-heavy capital spending of Amazon, Alphabet, Meta, and Microsoft (working midpoint near $725B). Treat it as an estimate, not an audited total.
- Roughly +75-80% year over year — the approximate increase over an estimated 2025 combined total near $410 billion (about +77% at the midpoint).
- Amazon appears to lead the ranking, with Alphabet and Microsoft close behind and Meta somewhat lower — but the top is tight and can reshuffle each quarter.
- 4 companies account for the bulk of the spending, concentrating the AI buildout in a handful of balance sheets.
- The capex-vs-revenue gap is the key risk: spending appears to be scaling far faster than directly attributable AI revenue.
- Every figure here is an estimate as of 2026, compiled from public filings and trackers. Verify against primary filings before citing.
The defining number in the technology economy right now is not a valuation or a user count — it is a capital-expenditure line, and ai capex 2026 is the phrase investors keep circling. As of 2026, the four largest US hyperscalers appear to be on track to spend somewhere in the region of $700-750 billion — a working midpoint near $725 billion — building the infrastructure behind artificial intelligence: data centers, custom chips, networking, land, and power. That figure is a best-available estimate, not an audited total, but even the low end represents one of the fastest corporate spending surges on record. This tracker breaks the range down by company, sizes the year-over-year jump, and confronts the question that will not go away — whether the revenue is anywhere close to catching up. Last updated: July 2026.
How big is ai capex 2026, really?
Roughly $700-750 billion, combined
Adding up the 2026 capital-spending guidance from Amazon, Alphabet (Google), Meta, and Microsoft produces a combined figure in the neighborhood of $700-750 billion, with a working midpoint near $725 billion. Not all of that is strictly "AI" — hyperscalers also build conventional cloud, storage, and office infrastructure — but the marginal growth is overwhelmingly driven by AI data centers and accelerator chips. In rough terms, these four companies are deploying capital at a scale that rivals the annual infrastructure budgets of mid-sized national governments.
Why a range, not a single figure
Companies do not publish a clean "AI capex" line. Reported capital expenditure blends AI and non-AI investment, and firms differ in how they treat finance leases for data centers — some include them, some disclose them separately. That is why the honest version of this number is a range. Anyone quoting a precise total to the dollar is implying a certainty the disclosures simply do not support.
The company ranking: who spends the most
Amazon appears out front, the pack close behind
On current guidance, Amazon looks like the single largest capital spender, reflecting both its AWS cloud buildout and its own AI-chip and model ambitions. Alphabet and Microsoft appear to follow closely, and Meta sits somewhat lower in absolute dollars — though Meta's spend is arguably the most aggressive relative to its size, since it monetizes AI indirectly through advertising rather than by renting out cloud capacity. The ordering is tight enough at the top that a single quarter of revised guidance can reshuffle it, so treat the ranks below as approximate.
The core data: estimated capex by company, 2025 vs 2026
Read this table as estimates, not ledger entries
The table below is the centerpiece of this tracker. Figures are approximate, expressed in US dollars, and reflect full-year capital-spending guidance or reasonable projections as understood in 2026. They will be revised as actual quarterly results land — so use them to gauge scale and direction, not for a precise citation.
| Company | 2025 capex (est.) | 2026 capex (est.) | YoY change (approx.) | Rank 2026 (approx.) |
|---|---|---|---|---|
| Amazon | ~$100-110B | ~$190-210B | ~+90% | 1 |
| Alphabet (Google) | ~$80-90B | ~$165-185B | ~+100% | 2 |
| Microsoft | ~$90-100B | ~$160-180B | ~+80% | 3 |
| Meta | ~$65-75B | ~$120-140B | ~+85% | 4 |
| Others (Oracle, xAI, CoreWeave, etc.) | ~$50-60B | ~$50B+ | n/a | — |
| Big Four total | ~$355-410B | ~$675-725B | ~+75-80% | — |
A note on the totals: the individual company midpoints above sum to roughly $675 billion for the Big Four alone, and once finance leases, smaller cloud players, and rounding are folded in, the headline lands near the $725 billion range used throughout this report. The "others" row is deliberately incomplete — a reminder that the buildout extends well beyond the four giants.
The growth story: an estimated ~77% year-over-year jump
One of the steepest capex surges on record
Moving from an estimated 2025 combined total near $410 billion to roughly $700-750 billion in 2026 implies a year-over-year increase of about 75-80% — call it ~77% at the midpoint. For context, capital spending at large, mature companies typically grows in the single digits to low double digits in a normal year. A jump this steep across four of the world's most valuable firms at once is not a normal year — it reads as a coordinated bet that AI demand will be durable enough to fill the capacity being built.
This growth rate will not repeat forever
It is worth stating plainly: a ~77% annual increase is a surge, not a trend line. Even the companies themselves frame current spending as front-loaded — a compressed buildout of chips and data centers. Extrapolating this growth rate several years forward would produce implausible numbers, so treat 2026 as a spike year, not a new baseline slope.
By segment: where the money actually goes
Chips, buildings, and power
The single largest destination appears to be AI accelerators — the GPUs and custom silicon that do the training and inference. Beyond the chips, a large share flows into the physical shell: data-center construction, land acquisition, cooling systems, and increasingly, power. Electricity availability has become a genuine constraint, and some capex now effectively pays for grid connections, on-site generation, and long-term energy contracts. As a rough split, a dollar of "AI capex" in 2026 goes partly to the compute inside the building and partly to the building-and-power around it — the exact mix varies by company and is not cleanly disclosed.
The capex-vs-revenue gap: the number that worries investors
Spending appears to outrun attributable AI revenue
Here is the tension at the heart of the story. Combined 2026 AI-specific revenue across these firms — the money customers pay specifically for AI features, tokens, and services — is widely estimated to be a small fraction of the ~$725 billion being spent. The exact ratio is unknowable from public disclosures, but the direction is not seriously disputed: capital appears to be deployed far faster than directly attributable AI revenue is arriving. The bull case says this is normal for infrastructure — you build the railroad before the freight shows up. The bear case says depreciation on hundreds of billions of dollars of fast-obsolescing chips will eventually pressure margins if demand disappoints.
Why the gap is hard to measure honestly
Neither side can prove its case cleanly, because companies do not break out "AI revenue" any more than they break out "AI capex." Cloud revenue, advertising lift, and productivity gains all blend AI and non-AI contributions. So the gap looks real, but its precise size is an estimate layered on estimates — a reason for caution on both the optimism and the alarm.
Outlook: what to watch each quarter
Guidance revisions, depreciation, and free cash flow
The most useful signals going forward are not the headline capex numbers but the second-order ones. Watch whether companies raise forward guidance (a sign of confidence) or quietly trim it. Watch depreciation expense, which lags the buildout and will grow heavily through 2027-2028. And watch free cash flow — the figure that shows whether these firms can fund the spend from operations or are leaning harder on debt and leases. Those metrics, reported every quarter, will reveal whether the ai capex 2026 bet is paying off long before any single revenue line confirms it.
For more numbers-first breakdowns like this one, see our top 10 lists, or return to Countly for the full library of global-statistics trackers.
Frequently asked questions
How much is Big Tech spending on AI in 2026?
Combined capital expenditure across the four largest US hyperscalers (Amazon, Google/Alphabet, Meta, and Microsoft) appears to be tracking toward roughly $700-750 billion in 2026, with a working midpoint near $725 billion. This is a best-available estimate compiled from company earnings guidance and reputable trackers, not an audited total, and it will be revised as each quarter reports. Verify against primary filings before citing.
How much did ai capex 2026 grow year over year?
The 2026 combined figure appears to represent an increase of roughly 75-80% over the estimated 2025 total (a working estimate near $410 billion), or about 77% at the midpoint. Growth this steep reflects a front-loaded surge in data-center and chip buildout and is not expected to repeat at the same rate every year.
Which company spends the most on AI capex?
On current guidance Amazon appears to be the single largest spender, followed closely by Alphabet (Google) and Microsoft, with Meta somewhat behind the top three. The exact ranking shifts quarter to quarter and depends on how each firm classifies capital spending, so treat the order as approximate.
What is the capex-vs-revenue gap?
It is the concern that capital spending on AI is rising much faster than the incremental revenue directly attributable to AI products. For most hyperscalers, 2026 AI-specific revenue appears to be a small fraction of the capex being deployed, meaning the payback case rests on future demand rather than current sales.
Are these AI capex figures official?
No. They are best-available estimates as of 2026, compiled from public sources such as company filings, earnings-call guidance, and reputable industry trackers. Companies rarely publish a clean "AI-only" capex line, so any AI-specific breakdown involves estimation. Verify against primary filings before citing.
How often does this AI capex tracker update?
The intent is to refresh the figures each earnings quarter, when the major hyperscalers report actual capital spending and update forward guidance. Between reporting periods the numbers reflect the most recent guidance and should be treated as provisional.
Methodology and disclaimer: Every figure in this report is a best-available estimate as of 2026, compiled from public sources — company filings and earnings-call guidance, official statistics, and reputable industry trackers. No single source is asserted as definitive, and companies do not disclose a clean "AI-only" capital-expenditure line, so all AI-specific breakdowns involve estimation; totals are expressed as approximate values or ranges rather than exact figures. These numbers may be revised materially as each earnings quarter reports. Please verify against primary filings before citing.
