Looks like AI is speed-running Moore’s Law like it’s trying to glitch through a speedrun—except this time, it’s breaking the game for real.
Here’s the deal: computing power used to double every couple of years, thanks to Moore’s Law.
But now AI-driven architectures are hitting the accelerator, pushing performance gains way past what traditional chips could ever do.
The game’s changed, and the graph above basically says we’re entering beast mode.
 Also try:
The Future Called: It’s Got AI, Robots, and a Blockchain Wallet Waiting for You
AI Compute Just Hit Turbo Mode—Here’s Why That’s a Big Deal
Historically, we’ve been doubling performance at a steady pace—8 doublings in 1975, 16 in 1983, 24 in 1998, and so on.
| |
|
|
| |
Compute power per dollar is set to improve over 1000x by 2030. Translation: AI is about to get way cheaper and way more powerful. |
|
| |
|
|
But then, boom: Google drops the TPU in 2018, and suddenly we’re at 40 doublings. Fast forward to now, and we’re at 48, with projections hitting 56 by 2027 and a wild 64 by 2030.
That’s the “end of the chessboard” moment, where performance is so insane that AI capabilities start feeling sci-fi level.
The reason this is happening is that AI-specific hardware is outpacing traditional computing improvements.
Instead of waiting around for silicon to get smaller and faster, companies are engineering chips designed specifically to run AI workloads more efficiently.
Think of it like upgrading from a flip phone to an iPhone overnight—except instead of a better camera, you get an AI that can think faster than ever.
Compute power per dollar is set to improve over 1000x by 2030. Translation: AI is about to get way cheaper and way more powerful.
We’re talking next-gen automation, smarter AI assistants, and machine learning models that can crunch data like never before. The future’s looking wild, and if this trend holds, AI’s just getting started. |
