It's all over the news: Sam Altman has been seeking $5 to $7 trillion from investors in the Middle East, to fund an AI chip venture. This business would be a direct competitor with Nvidia, which produces all the GPUs used for AI training right now. We discuss why Sam might be seeking this funding, as well as why he is asking for such a massive amount.

There are a lot of geopolitical reasons why the semiconductor industry has evolved the way it has. Thanks to globalization, and a lot of unchecked monopolies, all advanced manufacturing is done by one company in Taiwan, called TSMC. Although the US is not fond of the current situation and has been trying to reshore chip production through the CHIPS act, the US created this globalized system in the first place. It's a great way to make sure that IP profits make their way to the US, but it lends itself to fragile global supply chains that cannot survive shocks such as pandemics or climate disaster. The real issue is monopolies, but they are currently not being addressed.

Sam Altman isn't the only one trying to create new types of chips for AI. Cerebras has monstrously large chips that are extremely good at AI training, compatible with standard machine learning tools like PyTorch. Other companies are producing hardware focused on model inference, and even the big tech companies like Microsoft are getting involved in specialized hardware production. We might be looking at the beginning of an AI hardware renaissance, regardless of whether or not the $7 trillion comes through.


OpenAI CEO Sam Altman seeks as much as $7 trillion for new AI chip project: Report
https://www.cnbc.com/2024/02/09/opena...

Sam Altman Seeks Trillions of Dollars to Reshape Business of Chips and AI
https://www.wsj.com/tech/ai/sam-altma...

Sam Altman's $7 trillion chip dreams are way off the mark, says Nvidia CEO Jensen Huang
https://www.businessinsider.com/sam-a...

5 Reasons For Altman's Bold $7 Trillion Bet
https://thealgorithmicbridge.substack...

The CHIPS Act treats the symptoms, but not the causes
  / https-pluralistic-net-2024-02-07-farewell-...  

The CHIPS Act won’t solve the chip shortage
https://www.vox.com/recode/2022/7/27/...

AI Chips: What They Are and Why They Matter
https://cset.georgetown.edu/publicati...

New Cerebras Wafer-Scale Cluster Eliminates Months Of Painstaking Work To Build Massive Intelligence
https://www.forbes.com/sites/karlfreu...

The Cerebras CS-2 wafer-scale engine (850k cores, 40 GB SRAM)
https://news.ycombinator.com/item?id=...

Tachyum announces questionable 96-Core Prodigy-based AI workstation -- $5,000 system also supposedly includes 1TB of memory
https://www.tomshardware.com/pc-compo...

Microsoft is finally making custom chips -- and they’re all about AI
https://www.theverge.com/2023/11/15/2...

The Chip Company In Taiwan You’ve Never Heard Of That Powers The Devices In Your Life --TSMC
https://www.forbes.com/sites/lawrence...

#ai #hardware #nvidia #samaltman

0:00 Intro
0:29 Contents
0:37 Part 1: Why $7 trillion?
0:55 Trillions of dollars in perspective
1:26 Nvidia response to Sam Altman
2:01 Nvidia wants the status quo
2:45 Not Sam Altman's first chip venture
3:13 From billions to trillions
4:15 Bet on exponential growth
4:46 Why the Middle East?
6:00 Part 2: Geopolitical concerns
6:17 Advanced chip manufacturing is only in Taiwan
7:23 How did this situation arise?
7:52 Explanation of tariffs
8:24 Exporting IP laws
9:43 Free trade agreements
10:38 Book recommendation: The Interdependency
11:24 Semiconductor monopolies
12:33 Treating the symptoms, not the causes
13:16 Example: DRAM production
14:14 Is a new approach needed?
14:41 Part 3: AI hardware renaissance
15:32 Example: Cerebras wafer-sized chips
16:34 Wafer processors have 850,000 cores
17:35 Just a few million dollars
18:14 Example: Tachyum trillion-parameter inference
19:06 However, trust in the company is low
19:25 Example: big tech companies doing AI chips
20:03 AMD doesn't control the whole stack
21:13 Not just AI chips experiencing a renaissance
21:55 Return on $7 trillion investment?
22:34 Conclusion
22:28 AI hardware market
24:14 Outro