How Big Is the AI Economy, Really? A $15–20 Trillion Question

Every few months a new headline promises that AI will add trillions to the world economy. The numbers vary wildly — $7 trillion, $15.7 trillion, $19.9 trillion — and it's fair to ask: which is it, and what do these figures actually mean for a company deciding its AI budget next quarter?

Let's line them up.

The big four estimates

PwC: $15.7 trillion by 2030. The most-quoted figure. PwC projects global GDP will be 14% higher in 2030 because of AI — more than the current output of China and India combined. The split matters: $6.6 trillion from productivity gains and $9.1 trillion from consumption effects — better, more personalized products people buy more of.

IDC: $19.9 trillion through 2030. IDC counts cumulatively and lands higher, estimating AI will drive 3.5% of global GDP in 2030.

Goldman Sachs: $7 trillion over a decade. Goldman focuses on generative AI specifically, projecting a 1.5 percentage-point lift in annual productivity growth.

McKinsey: $2.6–4.4 trillion per year. McKinsey's estimate is for generative AI alone, annually — roughly the GDP of the United Kingdom, added every year.

The estimates differ because they measure different things: total AI vs. generative AI, cumulative vs. annual, GDP effects vs. corporate value. But every serious forecast points the same direction, and none of them is small.

The near-term number that matters more

Long-range trillions are abstract. The 2026 spending data is not.

That services number is the one most companies should stare at. It says something the GDP projections don't: capturing AI value is not a purchasing decision, it's an implementation discipline. The models are commoditizing; the scarce resource is the ability to wire them into real processes, measure them, and run them in production.

Where the value actually lands

Productivity first. In PwC's model, more than half of all gains to 2030 come from labor productivity — not from exotic new products. The boring version of AI (the same work, done faster, with fewer errors, around the clock) is where the first trillions live. Our own client work confirms it: response times measured in minutes instead of hours, processes that once took days compressed to a single conversation.

Concentration. The value is not being distributed evenly. Regional analyses show China and North America capturing ~70% of the impact, and firm-level research shows a single-digit percentage of "high performer" companies capturing most of the enterprise value. The AI economy is enormous — and winner-take-most.

The takeaway

You don't need to believe any single trillion-dollar forecast. You need to notice what all of them agree on: the value is real, the majority of it comes from ordinary productivity, and it flows disproportionately to organizations that can actually deploy — not just adopt.

The AI economy's size is a macro question. Your share of it is an engineering question.