Jim Keller to Sam Altman's $7T AI chips plan – 'I can do it for under $1T' | Hacker News Reader

Jim Keller to Sam Altman's $7T AI chips plan – 'I can do it for under $1T'

tomshardware.com

Great example of how "anchoring" techniques work, even if by accident.

A month ago talking about raises in terms of Trillions would feel insane. But with Sam coming out of the gate with 7T, suddenly "under 1T" seems cheap.

For my money, they both sound utterly deranged and divorced from the current reality of both what their product is, and the money supply.

Considering the insane impact Altman's company had on the general global economy (S&P500, NVDA, AMD etc. performance being carried by the hype OpenAI created) I can see how he could justify these outlandish numbers.

Crypto had a similar impact for similar reasons, namely enriching NVIDIA, and there was no there there either.

The difference here is the outlook on AI compared to crypto. In the past year alone AI has started shifting the way entire industries are working, and everything points to the trend continuing. Crypto never did that outside of the speculative investment space.

Just to play DA (I generally agree with your take) - lots of corporations invested into "blockchain" before realizing it was kind of useless. We are seeing tons of corporations now invest into AI, but there's a very real argument to be made that the "killer use case" has not really appeared, and it might not ever.

I definitely think AI has more staying power than crypto, and real use cases. But I'm not seeing it "shifting the way entire industries are working" just yet... we could absolutely be oversold on it.

Exactly. It wasn't that long ago that most people (here especially) thought Crypto was the beginning of a new paradigm in money, programming, business, and even government. In the end it's been literally none of those things, and it's hard to estimate just how much energy, time, and money has been wasted or mislaid as a result.

Now we have another NVIDIA enriching tech that everyone agrees is the future of work, art, maybe even programming and government. All of this despite the fact that what LLM's can actually do is... ok I guess? Sometimes useful, often incomplete or deceptive, and certainly not intelligent.

But we can imagine where this might go if trends continue indefinitely, and because so much money and hype is involved, common sense falls away. I fully expect "lessons will be learned" in four or five years, and people who scoff at AI skepticism today will have "always seen the writing on the wall."

Indeed. At least now it's only 3x TSMCs annual revenue.

Jim Kellar is also out to ride the AI bubble, but at least he has a company with product you can buy and a bunch of ASIC engineers with experience (Tenstorrent). $1T in his pocket would produce something, I'm sure. I just don't think you want to king-make to that degree.

The best thing of this is that it completely chips the legs out of Sam Altman in the eyes of government. (hopefully)

> "Start by eliminating the margin stacking," wrote Keller, implying the cost or profit margin added by each participant in the supply chain to deliver a product to the end user. "There are two–three layers.

Is margin-stacking the most correct term what Keller is describing?

I think “greed” has been the more common term used in history

I was thinking something closer towards "total cost at maximum vertical integration", but expressed succinctly in a word or phrase.

Reply on news.ycombinator.com