Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Nvidia Stock Crash Prediction

entropicthoughts.com

Post reply on HN
Print view View original

341–350 of 385 posts

Re: Nvidia Stock Crash Prediction

#341

Post by seanhunter »

I would be wary of taking analysis like this website at face value unless you know enough about quant finance to check some of the working for yourself. Just a skim shows a few statements that are questionable at best. Eg

> the theory of unbiased random walks assumes constant volatility throughout the year

No. I’m pretty sure it doesn’t. If you assume a brownian motion with a constant volatility as your stochastic process for computing the walk then of course vol is constant by definition, but you can use a stochastic vol process (eg Heston[1]), one with jumps or even an SVJJ process to compute the walk[2] if you want to. As long as you don’t have a drift term and the jumps are symmetrical the process will still (I think) be unbiased.

There are technical reasons why it may or may not be important to use stochastic vol, but if I recall correctly, it only really matters if you care about “forward volatility” (eg the volatility of Nvidia one year from some future point in time) which you would if pricing something that uses forward-starting options. Then the term structure of the volatility surface at a future date is important so you need a stochastic vol model. If you care about the price evolution but not the future volatility then you can validly make the simplifying assumption that jumps will cancel each other out over time and that volatility is a locally deterministic function of time and price (if not constant, which it obviously is not) and use something like a Dupire model.[3]

More significantly, implied volatility is just the market price of a particular option expressed in terms of volatility. This is convenient for traders so they can compare option prices on a like for like basis between underlyers without constantly having to adjust for differences in the underlying price, strike and time. Implied volatility is not actually the overall expected volatility of the underlying instrument. For that, you would have to fit one of the models above to market prices and calculate the expectation over all strikes and times. And that still is just the market’s opinion of the volatility, not an actual probability even if you apply the BoE adjustment thing he does in the article.

[1] https://www.homepages.ucl.ac.uk/~ucahgon/Heston.pdf

[2] “SVJ” means stochastic vol with jumps (ie discontinities) in the underlying price evolution. SVJJ means stochastic vol with jumps both in the price of the underlying and in the volatility. An example of this is the Matytsin model, which everyone just calls “SVJJ” but it’s not the only possible svjj model https://www.maplesoft.com/support/help/maple/view.aspx?path=...

[3] https://www.math.kth.se/matstat/gru/5b1575/Projects2016/Vola...

Re: Nvidia Stock Crash Prediction

#342

Post by gregorygoc »

Der_Einzige wrote: post #210
lairv wrote: post #114

Earlier quoted context omitted.

NVIDIA stock tanked in 2025 when people learned that Google used TPUs to train Gemini, which everyone in the community knows since at least 2021. So I think it's very likely that NVIDIA stock could crash for non-rationale reasons edit: 2025* not 2024

Google did not use TPUs for literally every bit of compute that led to Gemini. GCP has millions of high end Nvidia GPUs and programming for them is an order of magnitude easier, even for googlers. Any claim from google that all of Gemini (including previous experiments) was trained entirely by TPUs is lies. What they are truthfully saying is that the final training run was done on all TPUs. The market shouldn’t react…

They are not lies.

Re: Nvidia Stock Crash Prediction

#343

Post by imtringued »

Der_Einzige wrote: post #210
lairv wrote: post #114

Earlier quoted context omitted.

NVIDIA stock tanked in 2025 when people learned that Google used TPUs to train Gemini, which everyone in the community knows since at least 2021. So I think it's very likely that NVIDIA stock could crash for non-rationale reasons edit: 2025* not 2024

Google did not use TPUs for literally every bit of compute that led to Gemini. GCP has millions of high end Nvidia GPUs and programming for them is an order of magnitude easier, even for googlers. Any claim from google that all of Gemini (including previous experiments) was trained entirely by TPUs is lies. What they are truthfully saying is that the final training run was done on all TPUs. The market shouldn’t react…

JAX is very easy to use. Give it a try.

Re: Nvidia Stock Crash Prediction

#344

Post by Balinares »

bigyabai wrote: post #306
tracker1 wrote: post #273

Earlier quoted context omitted.

Doesn't even necessarily need to be CUDA compatible... there's OpenCL and Vulkan as well, and likely China will throw enough resources at the problem to bring various libraries into closer alignment to ease of use/development. I do think China is still 3-5 years from being really competitive, but still even if they hit 40-50% of NVidia, depending on pricing and energy costs, it could still make significant inroads wi…

> there's OpenCL and Vulkan as well OpenCL is chronically undermaintained & undersupported, and Vulkan only covers a small subset of what CUDA does so far. Neither has the full support of the tech industry (though both are supported by Nvidia, ironically). It feels like nobody in the industry wants to beat Nvidia badly enough, yet. Apple and AMD are trying to supplement raster hardware with inference silicon; both of…

Ask Claude, HN tells me that it can implement the things that you ask.

Re: Nvidia Stock Crash Prediction

#345

Post by MichaelRo »

iLoveOncall wrote: post #160
agentcoops wrote: post #128

Earlier quoted context omitted.

I hear your argument, but short of major algorithmic breakthroughs I am not convinced the global demand for GPUs will drop any time soon. Of course I could easily be wrong, but regardless I think the most predictable cause for a drop in the NVIDIA price would be that the CHIPS act/recent decisions by the CCP leads a Chinese firm to bring to market a CUDA compatible and reliable GPU at a fraction of the cost. It shoul…

> short of major algorithmic breakthroughs I am not convinced the global demand for GPUs will drop any time soon Or, you know, when LLMs don't pay off.

> short of major algorithmic breakthroughs I am not convinced the global demand for GPUs will drop any time soon

>> Or, you know, when LLMs don't pay off.

Heh, exactly the observation that a fanatic religious believer cannot possibly foresee. "We need more churches! More priests! Until a breakthrough in praying technique will be achieved I don't foresee less demand for religious devotion!" Nobody foresaw Nietzsche and the decline in blind faith.

But then again, like an atheist back in the day, the furious zealots would burn me at the stake if they could, for saying this. Sadly no longer possible so let them downvotes pour instead!

Re: Nvidia Stock Crash Prediction

#347

Post by zozbot234 »

seanhunter wrote: post #341

I would be wary of taking analysis like this website at face value unless you know enough about quant finance to check some of the working for yourself. Just a skim shows a few statements that are questionable at best. Eg > the theory of unbiased random walks assumes constant volatility throughout the year No. I’m pretty sure it doesn’t. If you assume a brownian motion with a constant volatility as your stochastic pr…

AIUI, there's nothing wrong per se with treating the "market opinion" of the volatility as a subjective probability, since that's effectively what it becomes given sensible no-arbitrage constraints. Just keep in mind that "bad" states of the world will be heavily overweighted in the resulting subjective expectation, for the risk-adjustment reasons mentioned in the OP.

Re: Nvidia Stock Crash Prediction

#348

Post by zozbot234 »

_0ffh wrote: post #296
zozbot234 wrote: post #267

Earlier quoted context omitted.

> We will run out of additional material to train on This sounds a bit silly. More training will generally result in better modeling, even for a fixed amount of genuine original data. At current model sizes, it's essentially impossible to overfit to the training data so there's no reason why we should just "stop".

You'd be surprised how quickly improvement of autoregressive language models levels off with epoch count (though, admittedly, one epoch is a LOT). Diffusion language models otoh indeed keep profiting for much longer, fwiw.

Does this also apply to LLM training at scale? I would be a bit surprised if it does, fwiw.

Re: Nvidia Stock Crash Prediction

#349

Post by vintermann »

SequoiaHope wrote: post #326
pvab3 wrote: post #262

Earlier quoted context omitted.

inference requires a fraction of the power that training does. According to the Villalobos paper, the median date is 2028. At some point we won't be training bigger and bigger models every month. We will run out of additional material to train on, things will continue commodifying, and then the amount of training happening will significantly decrease unless new avenues open for new types of models. But our current LL…

Run out of training data? They’re going to put these things in humanoids (they are weirdly cheap now) and record high resolution video and other sensor data of real world tasks and train huge multimodal Vision Language Action models etc. The world is more than just text. We can never run out of pixels if we point cameras at the real world and move them around. I work in robotics and I don’t think people talking about…

Yeah, another source of "unlimited data" is genetics. The human reference genome is about 6.5 GB, but these days, they're moving to pangenomes, wanting to map out not just the genome of one reference individual, but all the genetic variation in a clade. Depending on how ambitious they are about that "all", they can be humongous. And unlike say video data, this is arguably a language. We're completely swimming in unmapped, uninterpreted language data.

Re: Nvidia Stock Crash Prediction

#350

Post by lazide »

pixl97 wrote: post #229
lazide wrote: post #216

Earlier quoted context omitted.

Huh? What does your statements have to do with what I’m saying? I’m just pointing out changing it out at 5 years is likely cheaper than at 3 years.

Depends at the rate of growth of the hardware. If your data center is full and fully booked, and hardware is doubling in speed every year it's cheaper to switch it out every couple of years.

So many goal posts being changed constantly?
Post reply on HN