- My Forums
- Tiger Rant
- LSU Recruiting
- SEC Rant
- Saints Talk
- Pelicans Talk
- More Sports Board
- Coaching Changes
- Fantasy Sports
- Golf Board
- Soccer Board
- O-T Lounge
- Tech Board
- Home/Garden Board
- Outdoor Board
- Health/Fitness Board
- Movie/TV Board
- Book Board
- Music Board
- Political Talk
- Money Talk
- Fark Board
- Gaming Board
- Travel Board
- Food/Drink Board
- Ticket Exchange
- TD Help Board
Customize My Forums- View All Forums
- Show Left Links
- Topic Sort Options
- Trending Topics
- Recent Topics
- Active Topics
Started By
Message
re: Markets may be turbulent tomorrow due to contraction in the AI sector
Posted on 1/27/25 at 7:35 pm to NC_Tigah
Posted on 1/27/25 at 7:35 pm to NC_Tigah
quote:
FYI, I sold a bunch of NVDA on Friday at $146. I bought $50K at $117.60 today.
Hope for your sake that the gap between 117 and 146 was the AI hype baked into NVDA and now NVDA can grow from a new baseline. If that proves to be true, you will look like a genius trader on that move.
2
Posted on 1/27/25 at 7:37 pm to GumboPot
quote:
"I know. I canceled my ChatGPT math subscription and opened a DeepSeek account for free."
And that's what 99% of people are doing. It's definitely going to change the AI market going forward.
Posted on 1/27/25 at 8:21 pm to John Barron
So how did China get 50,000 of Nvidia's most powerful chips? Do you think Nvda was complacent or complicit?
Posted on 1/27/25 at 8:30 pm to GumboPot
quote:
Hope for your sake that the gap between 117 and 146 was the AI hype baked into NVDA and now NVDA can grow from a new baseline. If that proves to be true, you will look like a genius trader on that move.
When was the NVDA earnings call that kicked off this AI hype cycle? What was the share price then (adjusted for the split)? The difference is your AI hype.
This post was edited on 1/27/25 at 8:31 pm
Posted on 1/27/25 at 8:37 pm to TrueTiger
quote:
Tech is cannibalizing Tech. It's moving so fast now that the latest developments are obsolete in weeks
... this happened back in the late 1990s
the exact same thing, where tech was cannibalizing tech and things were changing in a second ...
Posted on 1/27/25 at 9:16 pm to John Barron
quote:
And that's what 99% of people are doing. It's definitely going to change the AI market going forward.
This is the biggest short-term effect. This will eat into the demand for the monthly plans and API charges from OpenAI, Anthropic, etc.
Demand for CPU/GPU Hours on cloud platforms will drop also as some of the efficiency techniques get more widely adopted by model makers and fine tuners.
I'm really waiting for 'host your own' to get more widely adopted to drive the subscription costs down even more. Anyone who has privacy concerns about the data being fed into a prompts should be interested....which should be just about everyone!!
This post was edited on 1/27/25 at 9:20 pm
Posted on 1/27/25 at 9:45 pm to John Barron
quote:
uote:
They either used A100s, as suspected, or what they claim, H800.
That's not what's suspected. That's what the people coping keep wishcasting.
That's what the Chinese stated. So let's roll with that.
2000 H800 costs 140,000,000.
That alone blows up your love fest with this nonsense of it only cost them 6 million.
Now you have to
- House it.
- Supply electricity... and that takes a lot of money to run wire.
- Those cards are usless unless they are in a system... Massive amount of money for that.
- Then it has to be cooled.
And the list goes on and on.
quote:
The bottom line is that DeepSeek uses fewer resources to get results faster.
But they didn't. Your own words are proving it. You are the one wish casting hoping it's true.
Posted on 1/27/25 at 9:51 pm to John Barron
quote:
quote:
I'm telling you those are not all truths.
Without any evidence.
Dude.. It's their OWN words! They are flat out claiming they used 2046 H800!
That, if they are being honest, would have cost them well over 100 MILLION alone
Posted on 1/27/25 at 10:16 pm to BCreed1
The $6M figure is definitely misleading - it only accounts for the compute cost for the final training run - almost 3M GPU hours at $2 or so an hour (approximate per GPU hr cost for an H800).
For comparison to competition, a GPU hour comparison would be better, but maybe not as impactful to laypeople.
For example - Meta needed 30M GPU hrs to train Llama 3 with 405B parameters. So DeepSeek managed to train more parameters (670B) with 1/10th of the GPU hours. Still impressive, but not "Tony Stark trained this model in a cave with scraps and $6M" impressive...
For comparison to competition, a GPU hour comparison would be better, but maybe not as impactful to laypeople.
For example - Meta needed 30M GPU hrs to train Llama 3 with 405B parameters. So DeepSeek managed to train more parameters (670B) with 1/10th of the GPU hours. Still impressive, but not "Tony Stark trained this model in a cave with scraps and $6M" impressive...
This post was edited on 1/27/25 at 10:18 pm
Posted on 1/27/25 at 10:42 pm to BCreed1
quote:
Dude.. It's their OWN words! They are flat out claiming they used 2046 H800
I know. I am not arguing that point. I am arguing with the people claiming they used 50,000 A100s like that Wang CEO said on CNBC. That's not true and is cope and wishcasting. The 6 million figure was for the training on the 2048 H800s. I understand buying those were not calculated in the cost in their report.
quote:
That, if they are being honest, would have cost them well over 100 MILLION alone
That is not true. A company buying that many H800s would get a bulk price close to 20k each.
Loading Twitter/X Embed...
If tweet fails to load, click here. This post was edited on 1/27/25 at 10:51 pm
Posted on 1/28/25 at 4:35 am to GumboPot
quote:Nah, it's a small position.
If that proves to be true, you will look like a genius trader on that move.
Posted on 1/28/25 at 5:18 am to Decatur
quote:Right.
When was the NVDA earnings call that kicked off this AI hype cycle? What was the share price then (adjusted for the split)? The difference is your AI hype.
I think the catalyst was NVDA's Q1 call.
Prices took off from the low 100's. We missed that run up, but bought in Aug after the Q2 call dropped prices. Just felt a $45 run up from Aug was too far too fast. Sold about 1/2 in Nov and the rest last week. Same premise (too far too fast) in reverse yesterday. But it's still 8-10% up from the Q1 call.
This post was edited on 1/28/25 at 5:31 am
Posted on 1/28/25 at 5:22 am to stout
quote:
DeepSeek is opensource AI built by 200 people for $6 million and has a chance to disrupt the AI bubble where the US is investing trillions in chips, data centers, and development.
It was built in just a few months and just released
It wouldn’t surprise me if the code was written by some other AI.
Posted on 1/28/25 at 6:59 am to John Barron
quote:
And that's what 99% of people are doing.
B2C was never the future for these basic AI prompts
B2B is the where the real revenue is, but that will take the compute that Deepseek isn't being honest about publicly.
Posted on 1/28/25 at 7:32 am to SlowFlowPro
Chinese lie about everything. Remember that Chinese drilling rig in Oklahoma, rated by the Chinese for deep drilling? When used for deep drilling it collapsed killing about 4 people….Chineese metal failed substantially under the Chinese rated strength…
Also these benchmarks for AI can be easily faked….they have the answer ahead of time, they just claim they got to the answer in x time…doesn’t mean they did….
Chinese are feeling the heat and making shyyyte up..
Also these benchmarks for AI can be easily faked….they have the answer ahead of time, they just claim they got to the answer in x time…doesn’t mean they did….
Chinese are feeling the heat and making shyyyte up..
Posted on 1/28/25 at 10:20 am to John Barron
quote:Please feel free to predict any impending market crash like the one you bragged about happening yesterday.
I remember starting a thread a month or so ago warning about a market crash and Warren Buffet selling almost 50% of his Apple shares. Some of these same people in this thread attacked me and said I was fear mongering. So complain all you want, I will keep posting valuable information
I'm making a lot of money off of you being consistently wrong.
Markets Today: (as I post this)
S&P500: 6,084.86, +42.75 points (+.70%) 50 points from an all-time record high
Dow Jones 30: 44.956.75, +251.07 points (+.55%) New 3 month high & only 80 points from an all-time record high.
NASDAQ: 19.596.29, +237.42 points (+1.19%)
Posted on 1/28/25 at 10:29 am to LSURussian
quote:
I'm making a lot of money off of you being consistently wrong.
Posted on 1/28/25 at 10:30 am to LSURussian
quote:
bragged about happening yesterday.
What happened Yesterday
- $600 billion Nvidia value obliterated
- $1 trillion wiped from US stocks
- $220 billion in crypto assets erased
Posted on 1/28/25 at 10:31 am to John Barron
quote:
- $600 billion Nvidia value obliterated
- $1 trillion wiped from US stocks
- $220 billion in crypto assets erased
He's asking for your predictions on what happens now, so you're on record and don't rely on hindsight bias.
And he's not asking for you to post the x post of another.
Posted on 1/28/25 at 10:41 am to SlowFlowPro
quote:
and don't rely on hindsight bias.
You mean like NC Boomer and NAFO Troll "claiming" on Monday when the stock is crashing that they sold their Nvidia stock on the previous Friday
This post was edited on 1/28/25 at 10:46 am
Page 13 of 14
Popular
Back to top
Follow TigerDroppings for LSU Football News