Hyperscaler Capex 2027 Forecast: Guidance, Estimates and the Run Rate

Wall Street expects Alphabet, Amazon, Microsoft and Meta to spend about $934.5 billion on capex in 2027, roughly 28% more than the $732.5 billion they guide to for 2026. Not one of them has given a 2027 number.

What they have given is direction. Alphabet expects capex to "increase significantly in 2027." Microsoft expects fiscal 2027 capex to grow. Amazon says it won't have enough capacity in 2027 either. Meta says it isn't providing a 2027 outlook yet.

The surprising part is how little the forecast asks for. To hit their 2026 plans, the four have to spend at an annualized pace of about $862 billion by the second half of this year. Consensus for 2027 is only about 8% above that pace. Below: each company's own words, the analyst estimates kept separate, and the worked math.

The short version

• No 2027 guidance exists yet. Every company-level 2027 figure you see is an analyst estimate, not guidance.
• Estimates have been rising. In June, Goldman Sachs put consensus at $920 billion for 2027. By late September, the consensus it cited, for five companies including Oracle, was $1.1 trillion. Goldman's own forecast is $1.2 trillion.
• The forecast is mostly the current pace, held. Microsoft, Amazon and Meta's 2027 estimates sit within about 7% of their implied second-half 2026 run rates. Alphabet's needs a further step-up of about 19%.
• First real numbers come in late October. Alphabet, Microsoft and Meta are expected to report on October 28 and Amazon on October 29. The dates aren't confirmed yet.

What each company has said about 2027

Company statements only, quoted from their most recent earnings calls.

Sources: Alphabet Q2 2026 transcript, Microsoft fiscal Q4 2026 call, Amazon Q2 2026 transcript, Meta Q2 2026 transcript, Oracle fiscal Q1 2027 transcript.

Oracle is the only one with a number that runs into 2027, because its fiscal year ends in May. "Net cash capex" is spending after customer prepayments, which Oracle increasingly uses to fund buildouts.

What analysts expect for 2027

These are estimates, not company guidance. Consensus figures come from I/O Fund's August 5, 2026 analysis, which cites MarketScreener. Estimates have moved since, so treat them as a snapshot.

2026 guidance is from each company's July report; see the 2026 hyperscaler capex tracker for ranges and sources. Growth is our arithmetic: the estimate divided by the guidance midpoint.

Bank forecasts sit higher than consensus:

• Goldman Sachs, September 25: the five largest US hyperscalers, including Oracle, lift capex from about $800 billion in 2026 to $1.2 trillion in 2027 and $1.4 trillion in 2028, against a Street view of $1.1 trillion for 2027. Growth slows to 54% in 2027 and 12% in 2028, as reported by Bloomberg.
• Goldman Sachs, June: consensus then implied $920 billion in 2027, growth of just 22%. Goldman's base case was about $1.1 trillion, with an upside case of $1.4 trillion, per Investing.com.

The June and September consensus figures may not cover exactly the same companies. The direction is still clear: forecasts for 2027 have climbed with every round of earnings.

Worked example: the 2027 forecast vs the current pace

How much of an increase does the 2027 forecast really assume? Compare it with the pace each company must hit in the second half of 2026 to meet its own guidance.

Step 1: subtract first-half 2026 spending from full-year guidance. Step 2: split what's left over the two remaining quarters. Step 3: multiply by four for an annualized pace.

Combined, the annualized pace is $861.6 billion, and consensus of $934.5 billion is about 8% above it.

First-half figures are the reported quarters in the 2026 tracker, plus Amazon's second-quarter cash capex of $53.1 billion from its July 30 call. Amazon's two quarters come from slightly different measures: its 10-Q line for Q1 and the call's "cash capex" for Q2.

Here's what that means:

• For Microsoft, Amazon and Meta, the 2027 estimate is roughly "keep spending at the late-2026 pace." It isn't a forecast of another big acceleration.
• Alphabet is the exception. Its estimate requires a further step-up of almost a fifth on top of an already raised plan. That matches management's "increase significantly" language, but it puts the most weight on Alphabet's next update.
• The pace is a simplification. Spending rarely splits evenly between Q3 and Q4. If Q4 is heavier, the exit rate is higher and the 2027 estimates look even more conservative, which is Goldman's argument.

Why 2027 is still likely to rise

The companies' words point the same way:

• Capacity is still short. Alphabet says it is "still in a supply constraint environment" and is renting third-party capacity as a bridge. Microsoft says demand "continues to exceed available supply." Amazon expects the shortage to last into 2027.
• Backlogs are already signed. Alphabet's cloud backlog is $514 billion, and it expects to recognize just over half within 24 months. Oracle's remaining performance obligations, contracted revenue not yet recognized, grew by $26 billion in its latest quarter.
• Building takes years. Data centers committed now are what get filled in 2027, so much of the 2027 spend is already in motion.

What could cap it

• Cash flow. Alphabet's free cash flow was negative $5.9 billion in Q2, and Amazon's trailing free cash flow swung to a $7.6 billion outflow. Meta's Q2 free cash flow was $784 million. More capex means more borrowing: Amazon has issued debt this year, and Meta is adding debt and partnerships such as the one it announced with BlackRock in July.
• Costs, not volume. Amazon raised its 2026 plan mainly because memory got more expensive. A 2027 increase driven by component prices buys less capacity than the headline suggests.
• Flexibility. Microsoft says its land and data-center spending is "quite flexible," and Meta wants to keep options open for 2028. Neither is a cut, but neither is a commitment.
• Off-balance-sheet building. Microsoft's shift to operating leases and Oracle's prepayment and bring-your-own-hardware deals move some building outside reported capex. Reported 2027 capex can understate the real build.

Try it yourself

1. Open the free 2026 AI capex dataset and note one company's 2026 guidance and first-half spend.
2. Redo the worked example for that company, using the low and high ends of its range instead of the midpoint.
3. After the late-October reports, plug the new Q3 figure into step 2 and see whether the second-half pace held.
4. Enter the company's latest operating cash flow and your estimated 2027 capex into the capex-to-free-cash-flow calculator. Can it fund 2027 from operations?

Our read

The 2027 forecast is less aggressive than the headlines make it look. For three of the four companies, analysts are mostly assuming the late-2026 pace holds for a year. The real question for 2027 isn't whether spending rises. It's whether revenue arrives fast enough to cover the depreciation that follows, because that bill lands in earnings after the cash has already gone out.

Watch three things on the October calls. First, whether Alphabet puts a number on "significantly." Second, whether Meta gives its first 2027 range. Third, whether anyone's third-quarter spend falls short of the second-half pace in the worked example. A shortfall would be the first sign that 2027 estimates are too high, not too low.

Questions readers ask

Has any hyperscaler given 2027 capex guidance?

Not as a dollar figure. Alphabet expects a significant increase, Microsoft expects fiscal 2027 capex to grow, and Meta has said it isn't giving a 2027 outlook yet. Oracle's fiscal 2027 guide of $90 billion to $95 billion covers June 2026 to May 2027, so it overlaps only part of calendar 2027.

How much will hyperscalers spend in 2027?

Consensus in early August was $934.5 billion for Alphabet, Amazon, Microsoft and Meta. Including Oracle, Goldman Sachs cited a $1.1 trillion consensus in late September and forecasts $1.2 trillion itself. These are estimates and will change after the October reports.

When will 2027 capex guidance come out?

The third-quarter reports are the first chance, expected October 28 for Alphabet, Microsoft and Meta and October 29 for Amazon. Companies often give a full-year range with their fourth-quarter results in late January or early February. Oracle reports its fiscal second quarter in December.

Why don't the estimates match the companies' own capex measures?

The companies count capex differently: Meta includes finance-lease principal, Amazon reports cash purchases net of incentives, and Microsoft's finance-to-operating lease shift lowers its reported figure. Analyst estimates may also follow each company's fiscal year. The AI capex accounting comparison lines up the definitions.

Go deeper with member research

The forecast tells you how much will be spent. Member research asks who gets paid and what it does to earnings:

• Issue 03: The AI bill is arriving late to earnings: four hyperscalers invested almost four dollars in infrastructure for each dollar of depreciation. This issue measures when that gap hits margins.
• Issue 02: The AI Capex Shock: where the next infrastructure dollar goes, which bottlenecks keep pricing power, how the boom is financed and what could break first.
• Issue 05: NVIDIA Is Building the Credit Market for AI: how AI infrastructure is moving into private credit and securitized debt as capex outruns cash flow.

Limits of this data

• No company has given calendar-2027 capex guidance. All 2027 dollar figures here are analyst estimates or bank forecasts.
• The per-company consensus is from August 5, 2026, and estimates have risen since.
• Consensus estimates may use fiscal years and definitions that differ from company guidance, especially for Microsoft.
• The worked example assumes an even split between Q3 and Q4 and mixes two Amazon capex measures.
• Earnings dates for late October are expected, not confirmed.
• This is research, not investment advice.