AI Inference Silicon
Inference silicon runs trained models in production rather than training them, and rewards throughput per watt over raw peak performance. It is a distinct design problem from training, and has drawn in custom accelerator designers alongside the incumbent chip vendors.
7 stocks
| # | Name |
Funds holding | Growth | Growth % |
Fund value |
|
|---|---|---|---|---|---|---|
| 1 |
1
Qualcomm
QCOM
$184B
| 2,514 | +269 | +12% | $121B |
|
| 2 |
2
Ambarella
AMBA
$2.71B
| 258 | +5 | +2% | $2.04B |
|
| 3 |
3
GSI Technology
GSIT
$163M
| 38 | 0 | 0% | $11.1M |
|
| 4 |
4
Blaize Holdings
BZAI
$35.3M
| 27 | 0 | 0% | $70.2M |
|
| 5 |
5
Defiance Inference AI Chip ETF
AINF
| – | – | – | – |
|
| 6 |
CGPT
6
VegaShares AI Inference ETF
CGPT
| – | – | – | – |
|
| 7 |
KSMH
7
xETFs Korea AI Semiconductor
KSMH
| – | – | – | – |
|