The performance per dollar of AI chips purchased each quarter has grown by an average of 49% per year | Epoch AI

Data Insight<br>Aug. 13, 2026

The performance per dollar of AI chips purchased each quarter has grown by an average of 49% per year

Cite

By Venkat Somala

Since 2023, the average dollar spent on AI chips each quarter has yielded about 49% more performance each year. At that rate, the performance a dollar buys doubles every 1.7 years. Growth has come in spurts, as spending shifts to the newest generation of chips. Price-performance was nearly flat through mid-2024, then roughly doubled as Blackwell-generation chips grew to a majority of new spending.

Performance per dollar of AI chips bought each quarter (2025 USD)QuarterChipPerformance per dollar (relative to an H100 at its 2025 price)Spending in quarter (billion 2025 USD)2023 Q1A1000.63$0.762023 Q1H100/H2000.80$1.832023 Q1Other1.28$1.192023 Q2A1000.64$0.832023 Q2H100/H2000.81$4.862023 Q2Other1.10$1.972023 Q3A1000.64$0.702023 Q3H100/H2000.82$8.292023 Q3Other1.05$2.932023 Q4A1000.65$0.192023 Q4H100/H2000.82$13.382023 Q4Other1.43$1.952024 Q1H200.31$1.752024 Q1H100/H2000.89$16.142024 Q1Trainium24.19$0.062024 Q1Other1.34$4.492024 Q2H200.31$2.792024 Q2H100/H2000.89$18.892024 Q2Trainium24.21$0.132024 Q2Other1.57$5.132024 Q3H200.31$3.412024 Q3H100/H2000.90$22.962024 Q3Trainium24.23$0.332024 Q3TPU v6e5.68$0.172024 Q3Other1.70$5.562024 Q4H200.32$4.022024 Q4H100/H2000.90$19.622024 Q4GB2001.72$7.702024 Q4Trainium24.26$0.782024 Q4TPU v6e5.72$0.732024 Q4Other1.75$5.682025 Q1H200.32$4.432025 Q1H100/H2000.98$8.712025 Q1GB2001.73$19.762025 Q1Trainium24.30$0.862025 Q1TPU v6e5.77$1.702025 Q1Other1.48$6.532025 Q2H200.32$1.862025 Q2H100/H2000.99$6.012025 Q2GB2001.74$16.472025 Q2GB3002.30$9.402025 Q2Trainium24.32$1.022025 Q2TPU v6e5.80$2.762025 Q2Other1.30$5.422025 Q3H200.32$0.212025 Q3H100/H2000.99$3.322025 Q3GB2001.75$13.012025 Q3GB3002.32$23.432025 Q3Trainium24.35$1.172025 Q3TPU v6e5.84$3.042025 Q3Other1.81$6.762025 Q4H100/H2001.00$0.612025 Q4GB2001.76$15.512025 Q4GB3002.33$32.032025 Q4Trainium24.38$1.342025 Q4TPU v6e5.88$1.632025 Q4Other2.42$9.00

Enable JavaScript to see an interactive visualization.

Using data from our AI Chip Sales Hub, we compute the performance per dollar of the chips sold in each quarter since 2023: their combined computing performance divided by the dollars spent on them, in constant 2025 dollars. Widely sold chips contribute more to each quarter’s figure than chips produced in smaller volumes.

Epoch's work is free to use, distribute, and reproduce provided the source and authors are credited under the Creative Commons BY license.<br>Learn more about this graph<br>We estimate the aggregate performance per dollar of the AI chips purchased in each quarter. Performance per dollar rose from about 5.6 × 1011 bit-operations per second per dollar in the first quarter of 2023 to about 1.4 × 1012 by the end of 2025, in constant 2025 dollars. Fitting an exponential trend gives an average growth rate of about 49% per year (90% CI: 36% to 66% per year).

The gains are largely due to chips getting exponentially more powerful with each new generation. While recent chips have become more expensive, they have grown more powerful at an even faster rate. In 2025 dollars, NVIDIA’s GB300 costs about 5.5 times the P100’s 2016 launch price, yet it delivers roughly 200 times the performance, making it about 37 times more cost effective.

While chips are available with performance per dollar as high as 3.6 × 1012 bit-operations per second per dollar (Google’s TPU v6e), the average across each quarter’s purchases is lower. Most AI hardware spending currently goes towards Nvidia’s GPUs, which command a relatively high margin and thus have a lower price-perf than some custom ASICs.

Data

Our analysis draws data from several Epoch datasets:

Chip sales : Quarterly units of each accelerator, from our AI Chip Sales Hub. We begin the series in 2023 because coverage before then is sparse, and end it in Q4 2025 because sales estimates for later quarters are not yet complete. Note that the Chip Sales hub, and thus this analysis, models the amount of units and compute sold rather than deployed each quarter.

Performance : We use each chip’s Total Processing Performance (TPP), a precision-normalized throughput metric, equal to peak operations per second multiplied by the bit width of that number format.

Price : We use one representative purchase price per chip. These are the purchase prices rather than the rental prices, and they cover the accelerator card or module rather than a full server. Most of these chips have no single MSRP, so we estimate what buyers were paying around the time each chip started shipping in volume. For Nvidia, AMD, and Huawei we use reported transaction prices, executive statements, and order values, and take a geometric mean when we only have a range.

Google and Amazon did not sell their chips directly, so we use what it costs them to procure the...