AMZN, MSFT, GOOGL Need $1 Trillion in Revenue To Earn 15% Return on AI Capex — Here's What Goldman Says About Their $1.69 Trillion Backlog

Amazon.com Inc. (NASDAQ:AMZN), Microsoft Corp. (NASDAQ:MSFT), and Alphabet Inc. (NASDAQ:GOOG) (NASDAQ:GOOGL) require $1 trillion in combined revenue between 2028 and 2030 to achieve a 15% return on invested capital for artificial intelligence infrastructure.

However, these companies’ existing cloud service backlogs total $1.69 trillion, covering 59% of that required revenue.

Capital Expenditure and Returns

To justify upfront capital expenditures on AI infrastructure, U.S.-Based hyperscalers must meet specific future revenue thresholds. According to a financial framework published by Goldman Sachs Global Investment Research, the path to profitability is supported by current revenue pipelines.

Summarizing the data, Futurum Equites’ Shay Boloor noted on X that “Goldman says $AMZN, $MSFT and $GOOGL only need ~$1T of 2028 to 2030 revenue to earn a 15% ROIC on AI capex.”

This $1 trillion revenue target represents the amount needed over a three-year period to hit the 15% Return on Invested Capital (ROIC) threshold.

Breaking Down the Cloud Backlogs

Financial models indicate that existing contractual commitments provide a foundation for these capital investments. The three companies reported a combined backlog of $1,687.9 billion as of the second calendar quarter of 2026.

According to the graphic shared by Boloor, Microsoft leads with a reported backlog of $678 billion, followed by Amazon at $513.9 billion, and Google at $496 billion. According to Goldman’s analysis, Microsoft would need to convert about 39.6% of its current backlog into revenue, compared with 67.7% for Amazon and 77.7% for Google, to meet their respective shares of the $1 trillion 2028-2030 revenue requirement.

Backlog figures reflect Goldman Sachs’ proprietary Calendar Quarter (CQ2’26) normalization and estimates, which may differ from reported company SEC filings.

Market Speculation and Upside

The ratio of required revenue to current backlog alters the perceived risk of ongoing data center expansions. Analysis of the Goldman Sachs data emphasizes this buffer, with Boloor saying: “That’s just ~60% of the ~$1.7T backlog already on the books, which makes AI buildout look a lot less speculative with even more upside if utilization or pricing moves higher.”

The underlying framework calculates these returns assuming 29 gigawatts of implied capacity added, factoring in utilization rates scaling to 85% by the year 2030.

How Have These Stocks Performed?

Disclaimer: This content was partially produced with the help of AI tools and was reviewed and published by Benzinga editors.

Photo courtesy: Shutterstock