Eh. That's the religion. You have to grow and fight for first mover advantage or something.

Does nvidia pay dividends, or they're also keeping those (indirect) VC money for "growth" ?

They invest heavily in startups and r&d.

Alright but what is the benefit to society as a whole?

If it is positive: thank you AI industry :D

GPU is infrastructure investment, while revenue is flow.

17x doesn't seem high to me.

I agree, with one caveat: computing hardware becomes obsolete much faster than e.g. an oil refinery.

It's profitable for investors, as they'll take later investors' money, unless you are the last of them. No one cares its real value.

Selling pickaxes during a gold rush

https://cdixon.org/2011/02/05/selling-pickaxes-during-a-gold...

> There is a saying in the startup world that “you can mine for gold or you can sell pickaxes.” This is of course an allusion to the California Gold Rush where some of the most successful business people such as Levi Strauss and Samuel Brannan didn’t mine for gold themselves but instead sold supplies to miners – wheelbarrows, tents, jeans, pickaxes etc. Mining for gold was the more glamorous path but actually turned out, in aggregate, to be a worse return on capital and labor than selling supplies.

NVidia is great at looking for and promoting new markets for their GPUs. The whole "GPGPU" market exists because of how heavily they invested in that area once they saw the initial promising research coming out of Stanford, etc.

A 1 year timeframe isn't nearly long enough. Capital purchases normally take time to show a positive return. Maybe the 17x is abnormal, but it's hard to know what's going to be normal in a new area like this.

1905, probably: "The oil industry spent 17x more on drilling equipment than it brought in in revenue"

Or 1848: “Prospectors spent 17x more on picks and shovels than they recouped by finding gold.”

Iirc some got lucky & struck gold. Most didn't, and moved on empty handed.

But businesses that did prosper long-term, were the ones that catered to the gold seekers: sellers of equipment, food stores, bars & room-for-rent type facilities.

So you could say that (on average) the 'meta-suppliers' did better long-term than the gold seekers themselves.

Whats the definition of long-term?

A lot of mining towns became ghost towns (i.e. Cerro Gordo [1]) and in those situations even the merchants are out of business. It's not like we stopped mining silver though; the prospectors moved on to other areas.

Like if TPUs replace GPUs then Nvidia is looking at a decrease in growth. The horse shoe sellers go out of business just as fast as the horse shoe maker does.

[1]: https://en.wikipedia.org/wiki/Cerro_Gordo_Mines

Wrong industry.

With nVidia being the leading investor in many AI companies ( https://news.ycombinator.com/item?id=38602930 ) , is this just an ouroboros of the same dollars making a lot of noise with little of value produced, other than inflating all the involved stock prices and valuations?

Likely a majority of those investments are earmarked to be spent on NVidia hardware. It is common for companies to "invest" in startups by giving the free service, is both marketing and helps lock in new customers.

Also, Microsoft investing billions of dollars into OpenAI, many of that in form of Azure credits, with OpenAI then spending those credits and more and bringing in revenue for Microsoft.

> ... In a presentation earlier this month, the venture-capital firm Sequoia estimated that the AI industry spent $50 billion on the Nvidia chips used to train advanced AI models last year, but brought in only $3 billion in revenue.

> Source: WSJ (paywalled) [1]

I dunno, OpenAI's revenue is apparently 2+B [2] so I really doubt the whole industry is only 3B. Just googling "AI industry reveune" [3] brings up a lot of claims of market size (not revenue ...) being well beyond 3B.

There's probably some qualifier on the AI that is being lost here. It's not like Google's adverts have served for years by marketeers in cubicles... it's all AI.

[1]: https://www.wsj.com/tech/ai/a-peter-thiel-backed-ai-startup-...

[2]: https://www.reuters.com/technology/openai-hits-2-bln-revenue...

[3]: https://www.google.com/search?q=AI+industr+reveune

I think they mean the AI generated content market, there aren't many today who spend a lot on pure AI generated content, the big market is for human arrangements of AI generated content not the raw product.

CharGPT made it mich easier for normal people were the journey is going.

Obvious investment and partially absolutely necessary in case your competitor is getting ahead of you.

The more interesting question is probably how this plays out.

But ai is no longer a hype, it's clear it has and will still continue moving goal posts.

Now do blockchain

I dint think there are any big blockchain networks left that use GPU mining. Bitcoin miners don't use GPUs and Ethereum moved to proof of stake.

Ah the Sun Microsystem days are back.

How many years before the GPUs bought in the last year become cost inefficient for commercial AI workloads? 3?