Market flips switch on Arista share value after 'cloud titan' turns on heel The share price for Arista Networks has crashed after the network switch maker warned of a sudden softening to its turnover due to a "cloud titan" customer cutting back on its spending plans. The company last night outlined financials for calendar Q3 ended 30 September with revenue up 7.6 per cent to $654.4m ($555m product and …
Looks like Facebook have realised its cheaper to make custom switches than use mainstream products.
While this fall is bad, at some point Microsoft/Azure are going to come to the same conclusion and Arista will lose even more annual revenue. The infrastructure and power costs of using off the shelf hardware is likely to be >50% higher than running your own hardware based on AWS and Google presentations. Those savings likely represent 2%-3% of a cloud providers running costs based on approximate costs (i.e. https://perspectives.mvdirona.com/2010/09/overall-data-center-costs/#comment-711044).
TL;DR: In 2018, Facebook/Azure spent >$400m/year on Arista. But not for long.
"If it's possible to make their equipment cheaper, why aren't Arista doing so?"
Because there is virtually no money in it for Arista. Aristas hardware magins are likely to be in the 50%-60% range
The information Google and AWS have released around their custom switches suggests higher specs than "typical" enterprise switches targeted at specific rack densities so it will likely be a design that only gets used at Facebook/Azure.
On the software side, most of the features get stripped out other than what is required.
Once you do all that, your long term market is likely to shrink (i.e. Facebook will be hosted on another provider)
That's an awful lot to deliver for revenue and no profit, particularly when you've been using increased revenue as your market guidance to indicate strong growth ahead.
>> "If it's possible to make their equipment cheaper, why aren't Arista doing so?"
> Because there is virtually no money in it for Arista.
> Aristas hardware magins are likely to be in the 50%-60% range
That answer makes no sense.
If margins are in the 50%-60% range,
then a small drop in cost could place margins in 60%-70% range
1. benefiting their share holders
2. benefiting their management, who are significant share holders
3. benefiting research & development, to create a better product to compete in the market
4. benefiting the health of the company, to create a cushion for a when the economy weakens
5. position the company for the future, to prepare for acquisition of other significant companies
An increase in good [keeping key people] profitability gives the latitude to drop price in order to:
1. better compete with lower end market in order to expand the market for their product
2. put downward pressure on competing vendors to take a larger size of market share
#Millennialism
I think you misunderstand cloud service provider requirements versus enterprise cost requirements.
AWS/Google have customised Broadcom chips for their own switches - it would cost Arista money to do something similar AND it's unlikely to have any value for enterprise customers who don't want the high speed, high density, high efficiency, minimal feature switches cloud providers want.
It's all additional cost for minimal revenue - margins aren't going to be increasing if Arista provide these products. A company with the engineers but no expensive corporate structure may make money from this but an established networking company would be hard pressed to make these items profitable.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
