What a difference a year makes in AI accounting

What happened

Alphabet, Amazon, Meta Platforms and Microsoft now mostly spread the cost of graphics processing units and other data-center equipment over five to six years. Alphabet moved to six years from four for servers in 2023, and Meta changed its timeline to five to 5.5 years a year later.

Confirmed

Global impact / market context

Spreading cost over more years lowers the yearly depreciation charge, which means the reported profit figure stays higher. That accounting choice can flatter earnings just as spending on AI servers and chips is rising sharply, so investors may be comparing companies with different timing assumptions.

Analyst inference

The companies building the largest computing networks used to train and run large-language models face huge spending needs. A small change in how long this equipment is assumed to keep earning cash flow shifts reported profit, so it can affect how investors value these firms and their shares.

Analyst inference

What to watch

1. Alphabet switched server depreciation to six years from four in 2023, while Meta later changed server and networking equipment to five to 5.5 years from four to five years. Confirmed
2. Watch whether the four companies shorten useful-life estimates, because a shorter span raises the yearly depreciation charge that reduces reported profit. Proposed
3. Track the gap between cash spent on AI equipment and depreciation charged to profit, since a widening gap signals reported earnings lean on longer assumed useful lives. Analyst inference

Evidence

• Original source link unavailable