Are Supercomputers Still Profitable in the Age of Cloud Computing?

  • Thread starter Thread starter Phylosopher
  • Start date Start date
  • Tags Tags
    business
Click For Summary

Discussion Overview

The discussion revolves around the profitability of supercomputers in the context of the rise of cloud computing services. Participants explore the relevance and necessity of dedicated supercomputers for various industries, contrasting them with cloud-based solutions like AWS. The conversation touches on theoretical, practical, and economic aspects of supercomputing.

Discussion Character

  • Debate/contested
  • Technical explanation
  • Conceptual clarification

Main Points Raised

  • Some participants question the profitability of supercomputers, suggesting that with cloud computing's rise, dedicated supercomputers may no longer be necessary.
  • Others argue that supercomputers are justified for specific applications, such as numerical simulations that require close coupling of processors, which cloud solutions may not provide.
  • A comparison is made between the profitability of supercomputers and that of tractor-trailer trucks, indicating that profitability depends on the use case.
  • Several participants note that while manufacturing supercomputers is profitable, the potential for selling computational time on supercomputers is less clear, particularly in comparison to cloud services.
  • Some mention that academic institutions are primary users of supercomputers and often prefer owning their systems rather than purchasing computational time.
  • There is mention of less publicized uses of supercomputers in sectors like weapons labs, banking, and oil and gas exploration, which may not be widely known due to proprietary interests.
  • Concerns are raised about the distinction between supercomputers and cloud computing services, suggesting that clarity on this point is necessary for further discussion.

Areas of Agreement / Disagreement

Participants express a range of views on the profitability and necessity of supercomputers versus cloud computing, indicating that there is no consensus on whether dedicated supercomputers remain profitable in the current technological landscape.

Contextual Notes

Limitations include varying definitions of what constitutes a supercomputer, the ambiguity surrounding the profitability of selling computational time, and the dependence on specific use cases that may not be universally applicable.

Phylosopher
Messages
139
Reaction score
26
TL;DR
Are supercomputers profitable?
I am wondering whether supercomputers are profitable! I am no computer engineer/scientist, but it seems that it could be. I tried to search on google, on whether supercomputers are in fact profitable, but I couldn't find a suitable source. Also, with the rise of cloud computing such as AWS, I feel that supercomputers, even if they were profitable in some way in the past, they are not anymore! Large companies like amazon and microsoft seem to dominate the scene. So what niche businesses actually need dedicated supercomputers, who would pay?

Note: Not sure if this is a suitable sub-forum
 
Engineering news on Phys.org
Phylosopher said:
I am wondering whether supercomputers are profitable!
It depends on what you mean by a supercomputer, and who makes the profit.
There are numerical simulations based on finite element analysis that justify supercomputers.
The close coupling between the processors is not possible with distributed processors or the cloud.

The people who run the programs seem to be happy with the results.

The people that make the supercomputers seem to be happy manufacturing and selling supercomputers.
 
Sounds a little like asking if tractor-trailer trucks are profitable.
If you are taking the family laundry to the laundromat, probably not.
If you are shipping 40 000 lbs of steel to the next town every week, probably.
 
  • Like
Likes   Reactions: etotheipi and PeroK
Baluncore said:
It depends on what you mean by a supercomputer, and who makes the profit.
There are numerical simulations based on finite element analysis that justify supercomputers.
The close coupling between the processors is not possible with distributed processors or the cloud.

The people who run the programs seem to be happy with the results.

The people that make the supercomputers seem to be happy manufacturing and selling supercomputers.

When I look at the largest supercomputers, I see governmental institutions and electronic companies. That is, no service providing companies:

(wiki)
1- Riken (Funded by the Japanese government)
2- Oak Ridge National Laboratory
3- Lawrence Livermore National Laboratory
4- National Supercomputing Center in Wuxi
5- Eni (oil and gas company)
6- Nvidia
...

Of course, for those who make supercomputers it is profitable, first and foremost they can sell them, secondly, they can use them for their own companies. But can someone establish for example, a company that is based on selling supercomputer computational time? Just like how AWS work, to some extent! What motivated me to write the OP is that, if such thing is possible, people would have already done it; but I couldn't find details on the web regarding this specific services.

Tom.G said:
Sounds a little like asking if tractor-trailer trucks are profitable.
If you are taking the family laundry to the laundromat, probably not.
If you are shipping 40 000 lbs of steel to the next town every week, probably.

It is of course profitable to manufacture tractors, the profit will come by selling them. Also, it is profitable to buy tractors and lend them for money. Does the same thing apply to supercomputers?

Of course, manufacturing supercomputers is profitable in itself. But is buying supercomputers, and then lending the computational time profitable?
 
Phylosopher said:
But can someone establish for example, a company that is based on selling supercomputer computational time? Just like how AWS work, to some extent! What motivated me to write the OP is that, if such thing is possible, people would have already done it; but I couldn't find details on the web regarding this specific services.
Phylosopher said:
Of course, manufacturing supercomputers is profitable in itself. But is buying supercomputers, and then lending the computational time profitable?
Supercomputers' use is mainly for scientific computations, so their main customers are in the academic world. They tend to want their own systems than to buy computer time.

I don't know of any private company that sells time on supercomputers, but I know that some academic/research institutions sometimes sell supercomputer time to private companies.
 
  • Like
Likes   Reactions: Phylosopher
Phylosopher said:
But can someone establish for example, a company that is based on selling supercomputer computational time? Just like how AWS work, to some extent! What motivated me to write the OP is that, if such thing is possible, people would have already done it; but I couldn't find details on the web regarding this specific services.
That is possible. I don't know of any.

However, there are many companies who sell cloud computer time. They have many CPUs, and you can buy time on as many as you want, but in the strict sense they are not supercomputers. Might that be want you want?
 
  • Like
Likes   Reactions: Phylosopher
DrClaude said:
Supercomputers' use is mainly for scientific computations, so their main customers are in the academic world. They tend to want their own systems than to buy computer time.

Not exactly.

There is some of that, but there is also supercomputer use that is less public, so you don't hear as much. There is also:
  • Weapons lab use
  • Banking
  • Oil and gas exploration
Which you don't hear as much about because the owners of these machines have interests in keeping exactly what they are doing proprietary.

Phylosopher said:
Also, it is profitable to buy tractors and lend them for money. Does the same thing apply to supercomputers?

Many - probably most - are leased, not bought.

Phylosopher said:
But is buying supercomputers, and then lending the computational time profitable?

How is a supercomputer different than AWSParallelCluster? I think that's probably necessary to answer for this thread to productively continue.
 
  • Like
Likes   Reactions: Phylosopher and rbelli1
Phylosopher said:
Just like how AWS work, to some extent!

If you think the owner of such a company being worth almost $200 billion is not profitable then no.

BoB
 

Similar threads

  • · Replies 14 ·
Replies
14
Views
3K
  • · Replies 6 ·
Replies
6
Views
2K
Replies
1
Views
2K
  • · Replies 1 ·
Replies
1
Views
3K
  • · Replies 1 ·
Replies
1
Views
3K
  • · Replies 6 ·
Replies
6
Views
2K
  • · Replies 1 ·
Replies
1
Views
1K
  • · Replies 10 ·
Replies
10
Views
6K
  • · Replies 1 ·
Replies
1
Views
1K
  • · Replies 3 ·
Replies
3
Views
3K