Maybe, but it seems that many or most of these things are being built on spec, and that speculation may or may not prove financially viable.
In that respect, I find it to have eerie echos of the dot com boom, same house of cards.
U.S. electricity price increases by state from 2020
- Thread starter GON
- Start date
In that respect, I find it to have eerie echos of the dot com boom, same house of cards.
MolaKule
Staff member
Let's keep this apolitical.
4wd
$50 site donor 2026
Going to be some very expensive competition - and to your point some will outrun their headlights …
I see your point, but the growth of server farms is necessary for the huge growth of the data we capture. Analytics has moved from text, which has a small footprint, to images which has a huge footprint.
Facebook alone handles tens of petabytes of data and generates 4 petabytes of new data daily, according to a 2014 study. And it's only gonna grow exponentially.
I posted a thread on the Vera C Rubin Observatory yesterday; the data capture will be measured in yottabytes. A yottabyte is 10^24 bytes. Yup.
Good conversation.
It's all well to capture that kind of volume, the question is whether Facebook and its brethren can monetize this growth in data captured. I really can't see how they can.
Let's just say, if you are not paying for the product, you are the product.
Agree with you, but is there that kind of value in me as a product for targeted advertising?
I cannot speak to you as an individual, but the Internet has changed everything. If it were not profitable, they wouldn't do it. The amount of insights gleaned from data and from tracking your clicks, is today's gold rush. Only far more valuable.
As a veteran of database application solutions, I can tell you that data by itself is worthless. But turning data into information is priceless.
I'm not sure if you are familiar with the new field of Data Science.
Data science is an interdisciplinary field that extracts knowledge and insights from data using statistics, algorithms, and scientific methods to solve complex problems and make better decisions.
Google pays as much as $250K for new Data Scientist grads to mine their Google Cloud. Amazon makes more money from their Amazon Web Services cloud platform than from their storefront.
Last edited:
Is it any sort of gold rush, or is it more of a nickle and dime proposition?
I need only search for a flight to get hundreds of popups offering to save me money by clicking on them.
Do I? Is anyone credulous enough to do so?
Maybe.
The jobs data center created would mostly be in the construction and R&D of the chips they bought. Consider how many R&D labors are used in the chips design and manufacturing, they would be the ones benefit from more data centers.
Nobody knows about a bubble until it pop, that's always the case.
Compare to dot com, I would say we are more aware of the risk now but there's a perfect storm of inflation, still lowish interest rate, carry trade, trade war, AI boom, etc. It is really a race against what is better: gold, T bill, AI stock, speculating foreign rebound, etc. Maybe nVidia would go Sun Micro System, maybe Oracle would go MCI Worldcom, but maybe they won't. Nobody knows.
In the end every investor has to calculate his own risk tolerance. I bought some of those stocks, but I am not going all in on them.
- Joined
- Nov 23, 2020
- Messages
- 741
The R&D jobs are created no matter where the chips get manufactured. In fact, many of the chips are already manufactured by TSMC in foreign fabs. TSMC is just now building US-based fabs. That was the underlying motivation behind the CHIPS Act. We stopped dominating semiconductor manufacturing a long time ago.
Most of the chip manufacturing cost is the fab cost and most of the fab cost is the equipments' cost.
US have been a if not the leader in almost all but lithography of the latest chip technology, and if you consider Cymer being the core of ASML's technology they are US technology. We just outsourced the funding to build the plant and buy the equipments to Taiwan, and we are still very much a leader in the manufacturing of it.
Our power transmission assets were sold to Spain some time ago, and the voters rejected an idea to take them back under eminent domain.
We NEED micro nuclear power plants on all big data centers clusters.
No one is making money on AI yet. The hyper scalers all make gobs on other things. NVDA sells chips today and the buyers amortize them over 7 years. So far it’s mostly smoke and mirrors on the profit side.
I am positive AI. But a lot of capital will get spent learning how to make it work. There will likely be a handful of winners and lots of losers.
Those would make in vehicle panels more attractive, sadly the actual market is actually moving towards less efficient and cheaper per watt and not more efficient
The newest optical disks are petabyte sized, would make a cheap and slow backup for more rarely accessed data.
Owen Lucas
$100 Site Donor 2023
- Joined
- Sep 5, 2021
- Messages
- 4,838
Can’t the water be reused? I’ve seen pictures of these new Ai clusters where there are large storage tanks on site.
Fill up storage tank > circulate through cluster > cooling tower > recirculate?
Maybe the water rights are a just in case? I must be missing something, I can’t image they would just waste the water after cooling.
Most of them, as far as I know, work through evaporated cooling. The towers on site is probably for emergency use.