AI Explained

Amazon, Alphabet and Microsoft Will Spend About $590 Billion on AI in 2026. Is It Paying Off?

The 3 biggest AI spenders on the Nasdaq: how much they're burning, what it's earning, how many businesses buy their AI, and how far those customers have got.

Key takeaways

  • Amazon (~$220B), Alphabet ($195–205B) and Microsoft (~$170B) plan to spend about $590 billion on capital projects in 2026, mostly AI data centres and chips. All three trade on the Nasdaq.
  • The spending is eating cash: Amazon's free cash flow was –$42.4B over the last 12 months, and Alphabet's was negative last quarter. Microsoft is still cash-positive.
  • The payoff is showing up in cloud sales – AWS +39%, Azure +48%, Google Cloud +103% – but most business customers are still early: only 37% of companies report any profit impact from AI (McKinsey, 2026).

Every tech giant says it’s “all in” on AI. But three companies are spending on a scale nobody has seen before: Amazon, Alphabet (Google’s parent) and Microsoft. All three trade on the Nasdaq, and together they plan to spend roughly $590 billion on capital projects in 2026 – mostly data centres, chips and the power to run them.

So where is that money going, is it coming back, and are the businesses buying their AI actually using it? Here’s what their latest results (July–September 2026) show.

Who spends the most?

“Capital expenditure” (capex) is money spent on long-lasting things – buildings, servers, chips. For these three, the companies themselves say most of it is now for AI.

2026 capex plan Spent in the latest quarter (Jul–Sep 2026)
Amazon about $220 billion $77.3 billion
Alphabet $195–205 billion $51.4 billion
Microsoft about $170 billion (calendar 2026) $45.2 billion (incl. $7.4B of finance leases)

Meta is fourth, with a $135–145 billion plan. Microsoft’s figure was trimmed from about $175 billion because some data-centre capacity it rents from other companies arrived late.

Amazon’s spending more than doubled from the same quarter a year earlier, and it says the increase “primarily reflects investments in artificial intelligence.” Microsoft says about three-quarters of its quarterly spend went on “short-lived” assets – mainly the GPUs and CPUs that do the AI work and wear out (or go out of date) within a few years.

How much cash is being burned?

The clearest measure is free cash flow: the cash a company makes from its business, minus what it spends on capex. For years these were some of the most cash-rich companies on Earth. AI has changed that.

Free cash flow What it means
Amazon –$42.4 billion over the last 12 months (a year earlier: +$14.8 billion) Spending more than the whole company earns in cash
Alphabet –$9.1 billion in the latest quarter; +$19.7 billion over 12 months Went negative for the quarter
Microsoft +$18.6 billion in the latest quarter (down from $25.7 billion a year earlier) Still positive, and expects to stay positive this financial year

So yes, “burning” is fair – at least for Amazon. The companies are betting that today’s cash outflow builds the capacity that earns money for the next decade.

Is it paying off?

None of the three publishes a single “return on AI investment” number. But their cloud businesses – where they rent out that AI capacity – show what the spending is buying:

Cloud revenue (Jul–Sep 2026) Growth Cloud profit Signed-but-not-yet-earned contracts
Amazon (AWS) $46.0 billion +39%, fastest in nearly five years $17.8 billion operating income $578 billion backlog
Alphabet (Google Cloud) $30.7 billion +103%, more than doubled $11.1 billion operating income (from $3.6 billion a year ago) about $561 billion backlog
Microsoft (Azure) $33.2 billion +48% not reported separately —

A few more signals:

  • Amazon says its AI business passed a $25 billion yearly run rate earlier this year, growing at triple-digit rates, and expects AWS to grow more than 40% in both 2027 and 2028.
  • Microsoft says revenue from Foundry, its platform for building AI apps and agents, more than tripled in a year.
  • Alphabet says its own AI models now process nearly 32 billion tokens a minute (tokens are the small chunks of text AI reads and writes) – more than four times a year ago.

The big backlogs matter most. They are contracts customers have already signed, which is why the companies keep spending: they say they don’t have enough capacity to serve the demand they’ve already sold.

How many businesses use their AI?

You asked about business customers, beyond everyday users. None of the three gives one total “enterprise AI customers” number, but each shares pieces of the picture:

Microsoft

  • More than 120,000 businesses use Foundry to build their own AI apps and agents.
  • Microsoft 365 Copilot, the AI assistant inside Word, Excel, Teams and Outlook, passed 30 million paid seats in July 2026.

Alphabet (Google)

  • Nearly 90% of the Fortune 100 use Gemini Enterprise, Google’s AI for businesses.
  • Google says it has sold more than 100 million seats of its AI services for office work.
  • Over the past year, more than 2,600 customers each used at least 100 billion tokens – heavy, real-world use, not just trials.

Amazon (AWS)

  • Fortune 500 companies spent nearly 7× more on Amazon Bedrock, AWS’s service for running AI models, than a year ago, and total Bedrock usage grew nearly 17×.
  • Tens of thousands of customers use AgentCore, AWS’s tool for building AI agents, and they deployed twice as many agents in July–September as in the previous nine months combined.

How far have those customers actually got?

This is the honest part. Buying AI and getting value from it are two different things.

Inside one product: Copilot’s 30 million paid seats sound huge, but Microsoft 365 has more than 450 million paid business seats. That’s roughly 7% – fewer than one in ten business users pays for the AI version so far.

Across companies in general, McKinsey’s 2026 State of AI survey of 1,719 business leaders found:

Measure 2026
Use AI regularly in at least one part of the business nearly 9 in 10
Scaling AI across the whole company (not just pilots) 44% (up from 38% in 2025)
Large companies ($1B+ revenue) scaling AI agents 40% (up from 27%)
Report any profit (EBIT) impact from AI 37% – unchanged from last year
“High performers”: AI drives at least 5% of profit about 6%

In plain terms: almost every big company is using AI, and many are rolling it out widely – but only about a third can see it in their profits yet, and only a handful are seeing big gains. Eight in ten respondents say AI makes them personally more productive; turning that into company-wide results is the step most haven’t finished.

The bottom line

  • The spending is real and enormous: about $590 billion this year across three companies, and Amazon and Alphabet have already said they’ll spend more in 2027.
  • The cost is real too: Amazon is now spending more cash than it makes, and Alphabet dipped negative last quarter.
  • The early returns are real: cloud revenue growing 39–103% a year, rising cloud profits, and well over half a trillion dollars each in signed contracts at Amazon and Alphabet.
  • The big question is the customers: most businesses are still between “trying AI” and “profiting from AI”. If they get there, today’s spending looks smart. If they stall, those data centres will be expensive to carry.

This post explains company results and surveys. It is not investment advice. Figures are from company filings and earnings calls for the quarter ended 30 September 2026; capex plans can change every quarter.

Disclosure: TechPrana uses Anthropic’s Claude to help with research and writing. Amazon and Google are investors in Anthropic. We have no financial relationship with any company in this post.

Your turn: does your workplace pay for an AI assistant yet – and does it actually save you time? Tell us in the YouTube comments.

Sources

  1. Amazon: Q3 2026 results (8-K Exhibit 99.1, 29 Oct 2026)
  2. Amazon: Q3 2026 earnings call transcript (StockAnalysis)
  3. Alphabet: Q3 2026 results (8-K Exhibit 99.1, 28 Oct 2026)
  4. Alphabet: Q3 2026 earnings call (Investor Relations)
  5. Microsoft: FY2027 Q1 earnings call (Investor Relations, 28 Oct 2026)
  6. Microsoft: FY2026 Q4 results – 30 million+ paid Copilot seats (8-K, 29 Jul 2026)
  7. CNBC: Microsoft stock jumps as Azure grows 48% (28 Oct 2026)
  8. Yield Theory: Hyperscaler capex 2026 guidance by company
  9. McKinsey: The state of AI in 2026 – On the road to ROI (Aug 2026)
  10. Office 365 for IT Pros: Copilot at 30 million seats is under 7% of Microsoft 365
AI spendingAmazonAlphabetMicrosoftCloud computingEnterprise AI

Keep reading