Article on the AI economy
The economic value of AI
How much is AI actually worth to the economy? Here is what the leading estimates say, how they are built, and why every dollar of value rests on compute, with real data.
Key takeaways
- Goldman Sachs Research estimates generative AI could raise global GDP by about 7 percent, roughly 7 trillion dollars, over ten years.
- McKinsey estimates generative AI could add the equivalent of 2.6 to 4.4 trillion dollars in value annually across studied use cases.
- These estimates use different methods, so they describe the same shift from two useful angles.
- All of this value depends on compute, which keeps demand for GPU hardware and infrastructure rising.
Two credible ways to measure AI value
There is no single number for the economic value of AI, because there is no single way to measure it. Two of the most cited estimates come at the question differently. Goldman Sachs Research looks at the effect on total economic output, while McKinsey looks at value created across specific business use cases.
Reading both together is more useful than picking one. They describe the same underlying shift from different vantage points, and the picture they paint is consistent: the potential value is large, and it is uneven.
This article walks through both estimates, explains how each is constructed, and connects the figures to the physical compute that has to exist for any of the value to be real.
What the value estimates show
~7%
Possible rise in global GDP over ten years, according to Goldman Sachs Research.
Source: Goldman Sachs Research, April 2023
$7T
Approximate scale of that GDP increase, according to Goldman Sachs Research.
Source: Goldman Sachs Research, April 2023
$2.6 to 4.4T
Value added annually across studied use cases, according to McKinsey.
Source: McKinsey and Company, June 2023
How these estimates are built
Goldman Sachs Research reaches its figure by modeling how automation and productivity gains feed into economic output over a ten year horizon, suggesting a roughly 7 percent lift in global GDP. It is a top-down view that asks what happens to the whole economy as AI raises productivity.
McKinsey instead totals the value across many concrete use cases, landing on the equivalent of 2.6 to 4.4 trillion dollars annually. That is a bottom-up view that adds up potential value function by function and task by task.
Both note that value is not spread evenly. McKinsey finds it concentrated in customer operations, marketing and sales, software engineering, and research and development. Estimates like these are projections, not guarantees, but they are grounded in detailed analysis rather than guesswork.
Two methods, side by side
| Aspect | Goldman Sachs Research | McKinsey |
|---|---|---|
| Approach | Top-down GDP modeling | Bottom-up use-case totals |
| Headline figure | About 7 percent global GDP over ten years | 2.6 to 4.4 trillion dollars annually |
| What it measures | Effect on total economic output | Value across specific business functions |
| Shared caveat | A projection, not a guarantee | A projection, not a guarantee |
Value the economy can feel
Trillions of dollars is an abstract figure until you translate it into ordinary work getting done faster and better across millions of tasks.
Both estimates are really attempts to put a number on that accumulation of small gains, which is why they describe the same shift even though their methods differ.
What could push the value higher or lower
These figures are scenarios, and several things move them. Faster adoption, better tools, and smarter use push the value toward the higher end. Slow rollout, friction, and limits on trust pull it lower. The authors present ranges precisely because the outcome is uncertain.
Another swing factor is whether enough compute, power, and skilled operations exist to run AI at the scale the estimates assume. A productivity gain that the hardware cannot support does not show up in the economy.
The honest takeaway is a direction, not a promise. Credible analysis points to large potential value, concentrated in a few areas, conditional on execution and infrastructure.
Value rests on physical compute
There is a constant beneath both estimates. AI value is produced by models running on GPUs in data centers that need power and cooling. The trillions of dollars described by Goldman Sachs and McKinsey cannot materialize without a large and growing base of compute.
That link is why some people look past using AI and toward owning the hardware that produces AI compute. One response is managed GPU ownership, where you hold physical NVIDIA-powered hardware and a professional team handles operations inside American data centers.
This is not a promise of any result. Operational benefits are not guaranteed and depend on utilization, uptime, demand, costs, hardware performance, and market conditions.
References and data
- Generative AI could raise global GDP by 7 percent. Goldman Sachs Research. April 2023.
- The economic potential of generative AI: The next productivity frontier. McKinsey and Company. June 2023.
Common questions about the value of AI
Estimates differ by method. Goldman Sachs Research suggests generative AI could raise global GDP by about 7 percent, roughly 7 trillion dollars, over ten years. McKinsey estimates it could add the equivalent of 2.6 to 4.4 trillion dollars in value annually across studied use cases.
They measure different things. Goldman Sachs models the effect on total economic output from the top down, while McKinsey totals value across specific business use cases from the bottom up. Together they describe the same shift from two angles.
Faster adoption, better tools, and skilled use push value toward the higher end, while slow rollout and friction pull it lower. Available compute, power, and operations also matter, since gains the hardware cannot support do not appear in the economy.
No. McKinsey finds value concentrated in customer operations, marketing and sales, software engineering, and research and development, so some functions and sectors capture far more of it than others.
No. They are projections based on detailed analysis, not guarantees. Actual outcomes depend on adoption, execution, and conditions, and all of the value still rests on having enough compute capacity.
Want a real position in AI compute?
Talk through what owning managed NVIDIA GPU hardware would look like, with no pressure and straight answers.
Operational benefits are not guaranteed and depend on utilization, uptime, demand, costs, hardware performance, and market conditions.