Amazon CEO Andy Jassy: Cost of AI still more expensive than it should and will be

# Amazon CEO Andy Jassy: AI Costs Will Fall, But Demand Remains Strong

In a recent interview, Amazon CEO Andy Jassy shared valuable insights on the current state and future of AI infrastructure costs. Despite emerging more efficient models like DeepSeek, Jassy maintains that Amazon sees no slowing in demand for data centers and computational resources.

## Key Points from Jassy’s Discussion:

### AI Costs Remain Too High
Jassy was direct in his assessment of the current AI landscape: “The cost of AI today is still more expensive than it should be. And than it will be.” This high cost barrier is limiting how extensively companies can deploy AI solutions.

### Lower Costs Won’t Reduce Overall Spending
Drawing parallels to AWS’s early days, Jassy explained an important economic principle:
– When cost per unit decreases, customers don’t spend less
– Instead, reduced costs “unleash them to do more innovation and they spend more”
– This pattern occurred with AWS’s compute, storage, and database offerings

### Two Major Cost Reduction Drivers
According to Jassy, two key factors will drive AI costs down:

1. **Chip Price Performance**: Amazon is addressing this by developing custom AI chips that deliver “30% to 40% better [price performance] than what is out there in the current GPU instances”

2. **Inference Optimization**: While most AI spending today focuses on training, Jassy notes that at scale, “the overwhelming majority of the spend is on the inference, the predictions of the model”

### AWS’s Mission on AI Costs
Jassy emphasized that AWS teams “feel like it is their responsibility and their mission to make the cost of AI meaningfully less than today.”

## The Bigger Picture

Amazon’s approach to AI costs reflects a strategic understanding that making AI more accessible ultimately expands the market. As Jassy put it: “The lower that we can make the cost of AI, the more customers are going to use it.”

For businesses watching the AI landscape, this suggests that while infrastructure demands remain high, the per-unit economics of AI deployment should improve significantly as Amazon and other providers work to reduce these barriers to innovation.