1. Home >
  2. Computing

TSMC: One Trillion Transistor GPUs Will Be Possible in a Decade

3D-stacking chiplets will be the standard to increase compute power going forward.
By Josh Norem
TSMC wafer
Credit: TSMC

Things are about to get interesting in the semiconductor world, according to a lengthy missive penned by some executives at TSMC. The world's largest chipmaker is embarking on a journey that it says will result in a GPU with one trillion transistors—roughly 10 times the amount found in today's biggest chips—though it will take the Taiwanese company a decade to get there.

TSMC chairman Mark Liu and chief scientist H.-S Philip Wong have penned an editorial for IEEE Spectrum outlining their thoughts on the future of semiconductors. The headline is how the company plans to create a one trillion transistor GPU. The article details how the AI boom is currently the main driver for increased compute power in chips, especially GPUs. It notes that as we reach the end of the traditional node-shrink era, the way forward is clear: chiplets and 3D stacking.

AMD MI300
AMD's MI300 series of accelerators are indicative of what future chips will require as it sports 146 billion transistors across 13 stacked chiplets. Credit: AMD

The pair say we're already at the reticle limit for 2D lithography, roughly 800mm squared. However, vertical-stacking technologies like chip-on-wafer-on-substrate (CoWoS) can allow for up to six reticle fields' worth of chips on a single package. It also touts its system-on-integrated-chips (SoIC) technology, which is used to stack high-bandwidth memory (HBM) chips. Current methods can stack eight layers, with 12 layers coming next. They note the transition from solder bumps between layers to "hybrid bonding" using copper connections will further increase density.

"If the AI revolution is to continue at its current pace, it will need even more from the semiconductor industry. Within a decade, it will need a 1-trillion-transistor GPU—that is, a GPU with 10 times as many devices as is typical today," write the duo. They note that semiconductor manufacturing is about to get much more difficult and complex than previously, as it used to be about shrinking the node. Now that there's not much left to shrink as we arrive at 2nm and beyond, the challenge will be in the form of connecting chiplets both horizontally, like with Nvidia's Blackwell GPU, and vertically, like with AMD's MI300 accelerators.

Tagged In

Semiconductors

More from Computing

Subscribe Today to get the latest ExtremeTech news delivered right to your inbox.
This newsletter may contain advertising, deals, or affiliate links. Subscribing to a newsletter indicates your consent to our Terms of use(Opens in a new window) and Privacy Policy. You may unsubscribe from the newsletter at any time.
Thanks for Signing Up