You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
OpenAI’s Sam Altman once stunned TSMC executives with a $7 trillion plan to build 36 chip fabs. A year later, he’s pursuing an even grander goal: nearly 30 gigawatts of AI compute, billed as the largest infrastructure project in U.S. history. Yet Taiwan’s suppliers paint a more grounded picture. From T-Glass substrate shortages to transformer supply gaps, the physical limits of the AI boom are becoming clear.
This week’s TechTaiwan report follows the trail from TSMC to Fortune Electric, revealing the hard bottlenecks behind Altman’s trillion-dollar vision — and why one Taiwanese executive put it bluntly: “No materials, no deal.”
It is laughable but also serious business -- "adult content" was a lot of the driving force behind rapid growth of new technology -- VCRs (1980s), and the Internet (1990s onward). Some of the Y2K bubble was building infrastructure for such use.
On a technical forums for running AI models locally, you'll see people write "I need an AI for Roleplay", which hints strongly at a demand for alternative uses of AI.
It is laughable but also serious business -- "adult content" was a lot of the driving force behind rapid growth of new technology -- VCRs (1980s), and the Internet (1990s onward). Some of the Y2K bubble was building infrastructure for such use.
On a technical forums for running AI models locally, you'll see people write "I need an AI for Roleplay", which hints strongly at a demand for alternative uses of AI.
Not surprising at all. AI will keep people in front of their screens for a long time and we are just at the tip of that iceberg. I just can't wait until AI takes over Hollywood. It costs $20+ to see a movie in a theater out here!
AI really is a gold rush moment and OpenAI is selling the pickaxes and shovels.