6 ms·

You seem to be assuming that the full cost of the cluster is recouped by Grok 3. The real value will be in grok 5, 6, etc… xAI also announced a few days ago th

by BluSyn 2y ago

You seem to be assuming that the full cost of the cluster is recouped by Grok 3. The real value will be in grok 5, 6, etc… xAI also announced a few days ago they are starting an internal video game studio. How long before AI companies take over Hollywood and Disney? The value available to be captured is massive. The cluster they’ve built is impressive compared to the competition, and grok 3 barely scratches what it’s capable of.

doctorpangloss 2y ago
> xAI also announced a few days ago they are starting an internal video game studio. Ha ha. I'm sure their play to claim airdrop idle game will be groundbreaking.
Tycho 2y ago
Yes. Why do get these replies on HN that seem to only consider the most shallow, surface details? It could well be that xAI wins the AI race by betting on hardware first and foremost - new ideas are quickly copied by everyone, but a compute edge is hard to match.
- gmerc 2y ago
  DeepSeek just showed the compute edge is not that hard to match. They could have chosen to keep the gains proprietary but probably made good money playing the market instead, quants as they are. https://centreforaileadership.org/resources/deepseeks_narrative_attack/ https://centreforaileadership.org/resources/deepseeks_narrat... If you’re using your compute capacity at 1.25% efficiency, you are not going to win because your iteration time is just going to be too long to stay competitive.
  - Tycho 2y ago
    Sorry, you missed the point - DeepSeek tried some new software ideas, they did not manage to secure the same computation capacity.
    - gmerc 2y ago
      They achieved the same results for 1.25% of the computation cost... If they actually had that computation capacity, it would be game over with the AGI race by the same logic.
  - mirekrusin 2y ago
    Optimisation efforts don’t negate investment in capacity but multiply output.
  - scarmig 2y ago
    Software and algorithmic improvements diffuse faster than hardware, even with attempts to keep them secret. Maybe a company doubles the efficiency, but in 3 months, it's leaked and everyone is using it. And then the compute edge becomes that much more durable.
- niceice 2y ago
  [flagged]
  - ansley 2y ago
    i think it's astroturfing
  - api 2y ago
    Reddit today feels like it's absolutely overrun by bots. So much of the comment content is so superficial and cookie-cutter I find it hard to believe it's all produced by human beings. A lot of it reads like the output of small cheap LLMs of the sort that would be used for spam bots. Of course we know X, Facebook, and probably most other social media is also overrun by bots. I don't think you can assume that humans are on the other end anymore.
    - jaco6 2y ago
      [dead]
  - BluSyn 2y ago
    Noticed this also. It doesn’t feel organic.
  - beepbopboopp 2y ago
    I mean, the honest truth is something closer to: We have no clue how all this is going to play out, what value is captureable and what parts of a lead are likely to stay protected. This race is essentially the collective belief in a generationally big prize and no idea how it unlocks. The problem with that for a comment section is it reduces ALL comments to gossip and guessing, which makes people feel stupid.
  - bnralt 2y ago
    There seems to be a coordinated effort to control the narrative. Grok3's release is pretty important, no matter what you think of it, and initially this story quickly fell off the front page, likely from malicious mass flagging. One thing that's taken over Reddit and unfortunately has spread to the rest of the internet is people thinking of themselves as online activists, who are saving the world by controlling what people can talk about and steering the conversation in the direction they want it to go. It's becoming harder and harder to have a normal conversation without someone trying to derail it with their own personal crusade.
- HarHarVeryFunny 2y ago
  The compute edge belongs to those like Google (TPU) and Amazon/Anthropic (Trainium) building their own accelerators and not paying NVIDIAs 1000% cost markups. Microsoft just announced experimenting with Cerebras wafer scale chips for LLM inference which are also a cost savings. Microsoft is in process of building optical links between existing datacenters to create meta-clusters, and I'd expect that others like Amazon and Meta may be doing the same. Of course for Musk this is an irrational ego-driven pursuit, so he can throw as much money at it as he has available, but trying to sell AI when you're paying 10x the competition for FLOPs seems problematic, even you you are capable of building a competitive product.
  - theckel 2y ago
    Curious where you saw the Microsoft/Cerebras experimentation noted online? That's very interesting.
    - HarHarVeryFunny 2y ago
      It was mentioned in Anthropic Jack Clark's "Import AI" newsletter. https://jack-clark.net/2025/02/17/import-ai-400-distillation-scaling-laws-recursive-gpu-kernel-improvement-and-wafer-scale-computation/ https://jack-clark.net/2025/02/17/import-ai-400-distillation...
  - Tycho 2y ago
    Timing matters. A long term strategy for superior hardware might bear fruit too late.
    - HarHarVeryFunny 2y ago
      I'm not sure about that - I expect AI is going to become a commodity market, so it doesn't matter how late you are if you've got a cheaper price. In terms of who's got a lead on cheap (non-NVIDIA) hardware, I guess you have to give it to Google who are on their 6th generation TPU.
  - vardump 2y ago
- acchow 2y ago
  > but a compute edge is hard to match. xAI bought hardware off the open market. Their compute edge could dissappear in a month if Google or Amazon wanted to raise their compute by a whole xAI
  - Tycho 2y ago
    Not if there’s a hardware shortage.
    - acchow 2y ago
      Ok, 2 months. Remember, the new B200 have 2.2x the performance of xAI’s current H100 “hardware edge”. So it only takes an order half the size. Or you could order the old H100 instead and avoid the B200 shortage.
Larrikin 2y ago
It's not going to be from this unless it's forced upon us by the federal government. All the other companies are ahead and aren't just going to stop.
gmerc 2y ago
None of which explains this massive waste of money for zero gain.
HarHarVeryFunny 2y ago
How quickly will Grok 4/5/6 be released? Of course you can choose to keep running older GPUs for years, but if you want bleeding edge performance then you need to upgrade, so I'm not sure how many model generations the cost can really be spread over. Also, what isn't clear is how RL-based reasoning model training compute requirements compares to earlier models. OpenAI have announced that GPT 4.5 will be their last non-reasoning model, so it seems we're definitely at a transition point now.
- gmerc 2y ago
  At current efficiency? Not nearly as fast as DeepSeek 4 ;)
kmac_ 2y ago
The point is that it is inefficient. Others achieved similar results much cheaper, meaning they can go much further. Compute is important, but model architecture and compute methods still outweigh it.

by BluSyn 2y ago

doctorpangloss 2y ago

> xAI also announced a few days ago they are starting an internal video game studio. Ha ha. I'm sure their play to claim airdrop idle game will be groundbreaking.

Tycho 2y ago

Yes. Why do get these replies on HN that seem to only consider the most shallow, surface details? It could well be that xAI wins the AI race by betting on hardware first and foremost - new ideas are quickly copied by everyone, but a compute edge is hard to match.

gmerc 2y ago
DeepSeek just showed the compute edge is not that hard to match. They could have chosen to keep the gains proprietary but probably made good money playing the market instead, quants as they are. https://centreforaileadership.org/resources/deepseeks_narrative_attack/ https://centreforaileadership.org/resources/deepseeks_narrat... If you’re using your compute capacity at 1.25% efficiency, you are not going to win because your iteration time is just going to be too long to stay competitive.
- Tycho 2y ago
  Sorry, you missed the point - DeepSeek tried some new software ideas, they did not manage to secure the same computation capacity.
  - gmerc 2y ago
    They achieved the same results for 1.25% of the computation cost... If they actually had that computation capacity, it would be game over with the AGI race by the same logic.
- mirekrusin 2y ago
  Optimisation efforts don’t negate investment in capacity but multiply output.
- scarmig 2y ago
  Software and algorithmic improvements diffuse faster than hardware, even with attempts to keep them secret. Maybe a company doubles the efficiency, but in 3 months, it's leaked and everyone is using it. And then the compute edge becomes that much more durable.
niceice 2y ago
[flagged]
- ansley 2y ago
  i think it's astroturfing
- api 2y ago
  Reddit today feels like it's absolutely overrun by bots. So much of the comment content is so superficial and cookie-cutter I find it hard to believe it's all produced by human beings. A lot of it reads like the output of small cheap LLMs of the sort that would be used for spam bots. Of course we know X, Facebook, and probably most other social media is also overrun by bots. I don't think you can assume that humans are on the other end anymore.
  - jaco6 2y ago
    [dead]
- BluSyn 2y ago
  Noticed this also. It doesn’t feel organic.
- beepbopboopp 2y ago
  I mean, the honest truth is something closer to: We have no clue how all this is going to play out, what value is captureable and what parts of a lead are likely to stay protected. This race is essentially the collective belief in a generationally big prize and no idea how it unlocks. The problem with that for a comment section is it reduces ALL comments to gossip and guessing, which makes people feel stupid.
- bnralt 2y ago
  There seems to be a coordinated effort to control the narrative. Grok3's release is pretty important, no matter what you think of it, and initially this story quickly fell off the front page, likely from malicious mass flagging. One thing that's taken over Reddit and unfortunately has spread to the rest of the internet is people thinking of themselves as online activists, who are saving the world by controlling what people can talk about and steering the conversation in the direction they want it to go. It's becoming harder and harder to have a normal conversation without someone trying to derail it with their own personal crusade.
HarHarVeryFunny 2y ago
The compute edge belongs to those like Google (TPU) and Amazon/Anthropic (Trainium) building their own accelerators and not paying NVIDIAs 1000% cost markups. Microsoft just announced experimenting with Cerebras wafer scale chips for LLM inference which are also a cost savings. Microsoft is in process of building optical links between existing datacenters to create meta-clusters, and I'd expect that others like Amazon and Meta may be doing the same. Of course for Musk this is an irrational ego-driven pursuit, so he can throw as much money at it as he has available, but trying to sell AI when you're paying 10x the competition for FLOPs seems problematic, even you you are capable of building a competitive product.
- theckel 2y ago
  Curious where you saw the Microsoft/Cerebras experimentation noted online? That's very interesting.
  - HarHarVeryFunny 2y ago
    It was mentioned in Anthropic Jack Clark's "Import AI" newsletter. https://jack-clark.net/2025/02/17/import-ai-400-distillation-scaling-laws-recursive-gpu-kernel-improvement-and-wafer-scale-computation/ https://jack-clark.net/2025/02/17/import-ai-400-distillation...
- Tycho 2y ago
  Timing matters. A long term strategy for superior hardware might bear fruit too late.
  - HarHarVeryFunny 2y ago
    I'm not sure about that - I expect AI is going to become a commodity market, so it doesn't matter how late you are if you've got a cheaper price. In terms of who's got a lead on cheap (non-NVIDIA) hardware, I guess you have to give it to Google who are on their 6th generation TPU.
- vardump 2y ago
acchow 2y ago
> but a compute edge is hard to match. xAI bought hardware off the open market. Their compute edge could dissappear in a month if Google or Amazon wanted to raise their compute by a whole xAI
- Tycho 2y ago
  Not if there’s a hardware shortage.
  - acchow 2y ago
    Ok, 2 months. Remember, the new B200 have 2.2x the performance of xAI’s current H100 “hardware edge”. So it only takes an order half the size. Or you could order the old H100 instead and avoid the B200 shortage.

Larrikin 2y ago

It's not going to be from this unless it's forced upon us by the federal government. All the other companies are ahead and aren't just going to stop.

gmerc 2y ago

None of which explains this massive waste of money for zero gain.

HarHarVeryFunny 2y ago

How quickly will Grok 4/5/6 be released? Of course you can choose to keep running older GPUs for years, but if you want bleeding edge performance then you need to upgrade, so I'm not sure how many model generations the cost can really be spread over. Also, what isn't clear is how RL-based reasoning model training compute requirements compares to earlier models. OpenAI have announced that GPT 4.5 will be their last non-reasoning model, so it seems we're definitely at a transition point now.

gmerc 2y ago
At current efficiency? Not nearly as fast as DeepSeek 4 ;)

kmac_ 2y ago

The point is that it is inefficient. Others achieved similar results much cheaper, meaning they can go much further. Compute is important, but model architecture and compute methods still outweigh it.