( ESNUG 588 Item 16 ) --------------------------------------------- [03/27/20]
Subject: Anirudh & Sawicki vs. Naveed & Costello on cloud, SaaS, pricing
DAC'19 Troublemakers Panel in Las Vegas, NV
Anirudh: Maybe you forgot that we (Cadence) announced our cloud initiative
more than a year ago. (See ESNUG 585 #6.)
Cooley: Yeah, you did it last DAC.
Anirudh: Yes. So, we do have a cloud offering. At first order, because
there are always improvements we can make in our offering -- but
the first order is not a technical issue.
We offer all our tools on the cloud with multiple cloud providers
and it's an acceptance issue of how customers are comfortable
with moving to the cloud. How much risk, perceived risk, they
want to take.
But in terms of running Xcelium (Rocketick) on the cloud or
running Virtuoso on the cloud, or the new Spectre-X on the
cloud, you can do that, through Cadence.
We have a regular flow and then also a couple of months ago at
CDNLive, we announced CloudBurst which is a hybrid environment.
So, I think the cloud is a delivery mechanism at the end of the
day. It's a little more than that because you can choose
multiple license, and you can adjust the number of CPUs to run
on it.
But this is something that is going to happen, is happening, and
by itself is not a vector to completely disrupt the toolchain
because all the providers will offer cloud offerings if they
have not offered already. So yes, you can get all the tools
from Cadence if you want on a cloud.
Naveed: I think delivering a tool through a cloud -- and having a cloud
model are two different things.
Cloud Model means that we deliver capability, as the customer
needed, in the amount he needed.
If you are going to say that you are buying 10 licenses from me;
or you can have it in the enterprise model; or you can have it
on the cloud. That's not a cloud model in my view. It's using
cloud to deliver your 10 licenses.
What Joe (Costello) is talking about and I would be interested
in, is how can we go to a true SaaS model where EDA is used as
per needed. You can have 1000 licenses; you can have 10,000
licenses when you need it. And when you don't need it, then
you don't need it.
Anirudh: That's a great point. So, I covered this in this panel last
year. (See ESNUG 585 #6.) Okay. There are three parts to this
cloud thing.
So, first is having the tools available on the cloud, okay.
Second thing is the business model.
We have an EDA card business model, in which you can go down to
one-week licenses. The last time I checked, any chip design
takes several months to do. Okay, maybe if you are really,
really good, it still takes several weeks to do.
What we think at this moment is that having granularity down to
one week is a good model. So, if you want it to get a Cadence
license on the cloud for one month, we have business model, to
allow you to do that.
Now the question maybe "well, is it cost effective or not?".
That's a pricing issue. We have a mechanism to allow
granularity of one week, two weeks, one month, three months,
six months pricing on the cloud.
The third part -- which is most important -- is the ecosystem
and the security issues and the foundry support.
This is where we (Cadence) have spent a lot of time in the last
two years, but having tools run on the cloud & having an EDA card
business model has been there forever.
And to say that other industries do it -- and EDA doesn't do it
-- okay I don't know about you, but I use Microsoft Office and
now I'm using it on the cloud and that is a yearly license for
$99. So, it went from a $99 perpetual license to a yearly
license of $99.
This is where we have spent a lot of time in the last two years.
But having tools run on the cloud and having a business model --
the EDA card has been there forever. And to say that other
industries do it and EDA doesn't do it? I don't know about you,
but I use Microsoft Office and now I'm using it on the cloud and
that is a yearly license for $99. So, it went from a perpetual
license to a yearly license of $99.
I don't think I can get Microsoft Office for you know one hour.
So, EDA sometimes gets a bad rap. The EDA business model's
transition to a yearly license happened several years ago with
both Cadence and Synopsys, and I don't know where Mentor is
with this.
And with Cadence, you have a lot of flexibility on that. Now
maybe you may say, well that's too expensive or not too
expensive... that's a pricing issue, and you know, we love our
customers...
And with Cadence, you have a lot of flexibility on that. Now
maybe you may say, well that's too expensive -- or not too
expensive. That's a pricing issue, and you know, we love our
customers...
Cooley: OK, you're working it out...
Sawicki, what's the Mentor take on this? Are you doing the
same strategy?
Sawicki: Well, one of the things I would look at, every time I hear this
argument about the cloud model versus EDA model. It talks about
a sales methodology that I haven't seen in probably 25 years.
It almost sounds like we have shoes on the shelf, and you come in,
you pay this much for shoes and we're making you buy 20 shoes.
And you only buy one shoe at a time. Our Cost of Sales is
probably, for each one of us [EDA vendors], sitting north of 20%.
Why is it north of 20 percent? Because we have to go out there
and do these contracts with people that have incredible amount of
complexity -- that models all these effects such as you don't
use 8,000 licenses all the time. We put together use models
that allow that to be effective.
So, the idea that this is all of a sudden going to be some massive
cost savings because we'll do it by the minute... [sighs] I just
don't see that as being how the world's going to work.
You know, you've also got this aspect of everyone wants to model
what everyone else's cloud looks like. Well, they're very
different things.
If you look at right now, if you're using Salesforce, Salesforce
upgrades to the server.
If I upgraded the server on Calibre and you're in the middle of
your run, I can guarantee you that probably is not certified for
that particular runset on TSMC. And you probably have NOT validated
working with that particular version of Innovus or anything else,
too.
It's a very different type of business model. So, this naive thing
of all we need to do is move everything to the cloud and have you
pay-by-the-minute and you'll be all happy and all the dinosaurs
will disappear. It's just not describing the path of the world.
There's going to be a changeover to cloud going out there.
There's going to be a changeover to people who are going to move
out a lot of their infrastructure up to there.
It's not going to look like Salesforce.
Anirudh: Exactly. Cloud is a very fundamental shift, and let me be clear,
we are wholeheartedly...
Naveed: Yes, and by the way, it's like saying Uber does not exist. What
you just described to me is basically telling me Uber does not
exist.
Sawicki: What's Uber? [laughter]
Naveed: I think you have answered your question. I think fundamentally
Uber happened. Uber is a reality, not just in the United States,
but world over, right?
And why Uber is a reality is that people figured out that owning
a car for the whole year around, when you're only you'd use it
for 10 percent of the time doesn't make any bloody sense.
Costello: This discussion about the cloud, it's like a birther discussion.
It's like you're thinking old school and being dragged, kicking
and screaming into it.
I'm telling you, you know when you're in a cloud -- yes, there's
different business models for different things. But when you're
in a high compute-intensive thing, we're using millions of CPU
minutes, by-the-minute is exactly the right way to sell it.
Cloud native applications will be done in a SaaS mode with
infinite flexibility. They'll be built containerized, so you
can build complete flows in the cloud.
You're NOT just using them as a compute server farm and
downloading the results and diddling with them on a local
machine. You're going to do everything in the cloud including
user interface.
It's a different world, a different model. All the tools will
work together in a flow like containerize things do very simply
and easily, well-orchestrated, and you'll get a SaaS model for
them.
And yes, it will save you a ton of money vs. the license models
that people have today and give you infinitely more flexibility.
Anirudh: But actually, I want to make sure one thing is clear. We will
enable all tools run on the cloud, and to run on premise. It
depends on the customer, what their IT environment is, how much
risk they want to take, how much public versus private cloud they
want to use, what kind of application.
We promote our cloud offering, we're very happy with it, we have
a lot of progress in a cloud offering, but we follow the customer.
A lot of customers want to run things on the cloud. Some
customers don't want to run on the cloud.
It's a matter of providing the solution and let the customer
decide how this is going to go. I think the debate over cloud
or not cloud is, "how will they accept it? "
Costello: Yeah, and I will tell you. I realize is it's a small sample set
because I've only visit, I've only visited maybe 20 some
customers. 100% of those 20-some customers on a spectrum of
small-to-large, when the head of design or the CFO or the CEO
or the general manager looks at this whole thing, they all say
"I want to move to the cloud" for all the reasons that we just
described.
Cooley: Because it's cheaper. Yeah.
Costello: No, it's cheaper, and it's more flexible, and it solves my
tapeout dilemma.
Anirudh: I have to correct this. To have a perception of Cadence and
Mentor not doing the cloud -- and just started with doing the
cloud is completely wrong. Okay.
So, this is Uber example. Okay, we have both Uber and we have
cars you can own. It depends on the customer.
[audience laughing, clapping]
Naveed: You have rental cars. You don't have Uber.
Cooley: Just an odd coincidence, because Naveed was just...
Naveed: Rental car is not Uber. Okay, let's go on.
Cooley: Alright.
---- ---- ---- ---- ---- ---- ----
Related Articles:
Anirudh & Sawicki vs. Naveed & Costello on cloud, SaaS, pricing
Raik & Sawicki on OneSpin, Questa FV, Jasper, and formal apps
Costello on his custom RISC-V Montana vs. CDNS Rocketick Intel
Join
Index
Next->Item
|
|
