But in this case, it has come to light that a shovel might not be needed for the job and a spoon will probably be enough. And the entry barrier for new players to sell spoons is relatively low.
Unless my analogy is off and I’m missing something here?
Disc: Not invested, but this company has been on my watchlist since last year.
@Randoinvestor is right! Now even the market understands that it was a mistake to sell Nvidia and other semi conductor stocks because of DeepSeek. There is much evidence to suggest that DeepSeek has led to a increase in the consumption of GPUs. Look at the picture that I have attached, this is for H100 spot pricing, look at the prices going up the day after Deepseek was released. Every one wants to try it without sending their data to China, so they spin up a H100 instance.
To borrow your analogy, if it so becomes known that you can get gold easily, more people will flood to the valley and the sales of shovels go up.
This is not to say stock will recover anytime soon. It won’t. But it was and is still a great business.
Disclaimer - Invested. I intend to stay invested for the long term with the firm. If I was running this firm, I would not have done it any differently and unit economics add up for me, that’s enough for me to stay invested.
Hi
We have been using e2e for the last 2years .Currently have 2 windows and 5 linux servers . Only once faced an issue and it was resolved in an hour.
However i have not used anything fancy, only basic servers where my team does the rest.
Still easier to use than aws.
E2E added to MSCI Small Cap Index
The question here is about the relevance of E2E with changing tech. This mainly stems from training/ML operations. But the inference i.e the consumption part of the AI still requires hardware above a dekstop computer. I run a 64 GB RAM Mac Mini M4 Pro. It can comfortably run a 32B Deepseek R1 model. Above this is, it’s not practical. So, in production settings, you need substantial hardware, infra and support to run it reliably around the clock if running the cutting edge models (600B+). That will keep E2E relevant. Deepseek made training cheaper. Inference is still expensive and it is something every company will need if self hosting.
So the definition of shovel will keep changing. When transformers becomes obsolete, it will be something else.
… and I like E2E managements decision to lazy load deploying the funds, this will ensure E2E stays relevant even when the shovels change.
I do see the shovels change btw, inferencing might be moving from GPUs to ASICs.
In the large scheme of things, when you are in as competitive a market as AI hardware, really only thing that you can bet on is a competent management team. Smart and driven group I think is the only true key differentiator in such markets.
I agree. I’ve said it before Dua has a nose to whiff out changing winds. The pivot to AI focus happened a few years back. It was not a coincidence. I wouldn’t be surprised if this is the reason they held back.
@vasu_gupta @vnktshb @satishwe @fundoo
Please help me understand what type of revenues we can assume from the IndiaAI Mission contributable to E2E. As I want to calculate the forward PE 3-4 qtr down the line. As I feel currently all the beating has already happen and it should not go down further(10% down can anyway happen).
We are the users of E2E since a year and mainly using them to train our models(80%) and do inferencing(20%) as well and we use them as they are the cheapest in providing the A100/H100 machines.
Disclaimer - Invested. I intend to stay invested for the long term.
Oh you are using them for GPUs?! I hope you have also used/checked out offerings of AWS/Google cloud, if you can share with us, if you have to tomorrow move to AWS/Google cloud for whatever reason, what will be the things you will miss about E2E’s offerings?
Price is only thing which has made us use the E2E. If AWS/GCP/Azure give same price or even 10-20% higher and let me use only 1 GPU machine as these A100 and H100 comes in 8 GPU slot. I will move back to them.
Current pricing is more than 50% cheaper in comparison with AWS/GCP/Azure.
As you are a user of E2Es GPUs, would like to know if you use them for Training or for Inference. Also, if you are using them for Training then can you switch to AWS/Azure easily or its a difficult task.
Have you checked the runpod.io and novita.ai? I find their pricing lot cheaper than E2E, easy to start, stop/pause and terminate the instances.
@nityanandparab
It is not that difficult to switch in GPU for training. We work on good scale and even inference switching is also not that difficult given enough incentives are given.
Now a days lot of startups change without AI workload on cloud as well once they get very good deal from another CSP in terms of discount or credits or combination. Like migration from AWS to Azure, GCP etc and vice versa.
will check and let you know.
i asked deepseek about its own newly released paper today, mainly about any improvements in efficiency and cost, and this is what it said -
Difficult to predict revenue potential from India AI Mission, since it is surely at discount from currently e2e is charging and different chips have different rates.
However, I see the opportunity in AI same as the emergence of Internet was back in late 90s, whoever has an earlier head start will be able to capture the opportunity more better, which e2e has.
And my conviction in there capability has been cemented since l&t had taken stake and company will have synergies and clients leads benefits from them.
I think, we just have to give company time to perform and a longer leeway, they will surely come out on winning side.
Checked Runpods pricing but it still seems to be more expensive than E2E by a significant margin.
I am getting RTX 3090 16 vCPU 62 GB RAM 25GB VRAM at 0.21$/hr or 13k per month inclusive of taxes at novita.ai
I am tracking E2E company since 2018, using product since 2020. I am still using their CPU instance at the moment.
Last time, I checked their lowest GPU(L4) started from Rs 36k with 18% taxes extra. Also, besides Pricing there are so many other things which developers look for.
L4 vs 3090 Benchmark
Hi, have you faced any issues regarding uptime & stuff while using services from E2E. How is it comparable to other players (especially giants like aws, google). And also, how are they in terms of solving customer problems?
