Earlier quoted context omitted.
Saying the financials are terrible is a bit of a stretch. Rapidly growing revenue, decreasing loss/share and a loss/share similar to other companies that IPO'ed this year. The more concerning thing is just not having diversity of revenue, since most of it comes from G42.
Has G42 shipped any working AI models?
AI chipmaker Cerebras files for IPO
121–130 of 138 posts
Re: AI chipmaker Cerebras files for IPO
#122Earlier quoted context omitted.
You do realize that brokerages earn commissions on selling shares, so why wouldn't they contact people who may be interested?
The point is, there are IPO shares available to sell, even to people who have never expressed any interest in the company. That never happens if there's genuine demand for an IPO.
1. Have you purchased other companies' stock from the aforementioned broker?
2. The price of the shares in private markets has been steadily inclining, so I think there is demand.
Re: AI chipmaker Cerebras files for IPO
#123The only way for Cereberas to actually succeed in the market is to raise funds. They need better software, better developer relations, and better hardware too. It's a gamble, but if they can raise enough money then there's a chance of success, whereas if they can't it's pretty hopeless.
Time (and the market ?) will tell whether all the people clamoring for NVIDIA alternatives actually put their money on it (understanding that NVIDIA's headstart is a long-term heavy investment on software too: compilers, libraires, and of course hardware/software co-design). I still can't fathom how Intel thought Arc and/or Ponte Vecchio would pay for themselves on day one.
Re: AI chipmaker Cerebras files for IPO
#124Earlier quoted context omitted.
https://finance.yahoo.com/news/cerebras-launches-world-faste...
This isn't a benchmark, it's a press release. MLPerf has an inference component so they could have released numbers, but they chose not to. At the end of the day it's all about performance per dollar/TCO, too, not just raw perf. A standardized benchmark helps to evaluate that. My guess is that they neglected the software component (hardware guys always disdain software) and have to bend over backwards to get their ha…
MLPerf brings in exactly zero revenue. If they have sold every chip they can make for the next 2+ years, why would they be diverting resources to MLPerf benchmarking?
Artificial analysis does good API provider inference benchmarking and has evaluated Cerebras, Groq, Sambanova, the many Nvidia-based solutions, etc. IMO it makes way more sense to benchmark actual usable end points rather than submit closed and modified implementations to mlcommons. Graphcore had the fastest BERT submission at one point (when BERT was relevant lol) and it didn't really move the needle at all.
Re: AI chipmaker Cerebras files for IPO
#125Earlier quoted context omitted.
Saying the financials are terrible is a bit of a stretch. Rapidly growing revenue, decreasing loss/share and a loss/share similar to other companies that IPO'ed this year. The more concerning thing is just not having diversity of revenue, since most of it comes from G42.
Has G42 shipped any working AI models?
afaik they have the current SOTA language models for arabic
Re: AI chipmaker Cerebras files for IPO
#126So many things here smell funny... I have never heard of any models trained on this hardware. How does a company IPO on the basis of having the "best tech" in this industry, when all the top models are trained on other hardware. It just doesn't add up.
Seems like they support training on a bunch of industry standard models. I think most of the customers in the training space tend to be for fine tuning right? The P and T in GPT stand for pre-trained - then you tune for your actual specification. I don't think they will take over the insane computational effort of training Llama or GPT from scratch - those companies are using clusters that cost more than Cerebras' last evaluation.
Re: AI chipmaker Cerebras files for IPO
#127Earlier quoted context omitted.
Their tech is very impressive, look it up.
It's a deadend. SRAM doesn't scale on advanced nodes. Similar to Tenstorrent who chose GDDR instead of HBM, they throught production AI models won't get bigger than GPT3.5 due to cost.
I don't think they rely on SRAM very much for training. https://cerebras.ai/blog/the-complete-guide-to-scale-out-on-... outlines the memory architecture but it seems like they are able to keep most of the storage off wafer which is how they scale to 100s of GB of parameters with "only" 10s of GB of SRAM.
Re: AI chipmaker Cerebras files for IPO
#128Earlier quoted context omitted.
> That’s why tinycorp is betting on a simple ML framework (tinygrad, which they develop and make available open source) whose promise is, due to the few operations needed by the framework: it’ll be very easy to get this software to run on a (eg your) new chip and then you can run ML workloads. This sounds easy in theory, but in reality, based on current models, the implementations are often tuned to make them work fa…
Often enough, hardware-specific optimizations can be performed automatically by the compiler. On the flip side, depending on a small set of general-purpose primitives makes it easier to apply hardware-agnostic optimization passes to the model architecture. There are many efforts that are ultimately going in this direction, from Google's Tensorflow to the community project Aesara/PyTensor (née Theano) to the MLIR inte…
I'm a compiler engineer at a GPU company, and while tiny grad kernels might be made more performant by the JIT compiler underlying every GPU chips stack, oftentimes, a much bigger picture is needed to properly optimize all the chip's resources. The direction that companies like NVIDIA et al are going in involves whole model optimization, so I really don't see how tiny grad can be competitive here. I see it most useful in embedded, but Hotz is trying to make it a thing for training. Good luck.
> There are many efforts that are ultimately going in this direction, from Google's Tensorflow to the community project Aesara/PyTensor (née Theano) to the MLIR intermediate representation from the LLVM folks.
The various GPU companies (AMD, NVIDIA, Intel) are some of the largest contributors to MLIR, so saying that they're going in the direction of standardization is not wholly true. They're using MLIR as a way to share optimizations (really to stay at the cutting edge), but, unlike tiny grad, MLIR has a much higher level overview of the whole computation and the company's backends will thus be able to optimize over the whole model.
If tiny grad were focused on MLIR's ecosystem I'd say they had a fighting chance of getting NVIDIA-like performance, but they're off doing their own thing.
Re: AI chipmaker Cerebras files for IPO
#129Earlier quoted context omitted.
You seemed surprised that this company is having an IPO to actually raise funds for operations and expansion, vs as just an "exit" where VCs and other insiders can dump their shares onto the broader public. I might be a bit suspicious if a company in some low-capital-intensive industry was IPOing while unprofitable, but this is chip making. Even if they're not making their own fabs this is still an industry with high…
Does this mean that they couldn't find VCs to raise more cash?
I don’t think that’s been a problem. I’ve been following the pre IPO market on them for a while and pretty much any shares at a 7b valuation have been snapped up pretty much same day
Re: AI chipmaker Cerebras files for IPO
#130Earlier quoted context omitted.
This isn't a benchmark, it's a press release. MLPerf has an inference component so they could have released numbers, but they chose not to. At the end of the day it's all about performance per dollar/TCO, too, not just raw perf. A standardized benchmark helps to evaluate that. My guess is that they neglected the software component (hardware guys always disdain software) and have to bend over backwards to get their ha…
MLPerf brings in exactly zero revenue. If they have sold every chip they can make for the next 2+ years, why would they be diverting resources to MLPerf benchmarking? Artificial analysis does good API provider inference benchmarking and has evaluated Cerebras, Groq, Sambanova, the many Nvidia-based solutions, etc. IMO it makes way more sense to benchmark actual usable end points rather than submit closed and modified…
With Artificial Analysis I wonder if model tweaks are detectable. That’s the benefit of a standardized benchmark, you’re testing the hardware. If some inference vendor changes Llama under the hood, the changes are known. And of course if you don’t include precise repro. instructions in your standardized benchmark, nobody can tell how much money you’re losing (that is, how many chops are serving your requests).