Now today, at the edge, I don’t — we support GPUs today, but that would be more in the core data centers. I think for the inference engines, we’re running that just fine on CPUs. And so as we look at the edge, where you’d be running the inference engines, I don’t, I think, run on CPUs and be much more cost-effective than trying to buy GPUs or custom hardware there. I think where you might see that as more if you’re working with large-scale model development. And then that would be in the core data centers. That wouldn’t be at the edge. The edge is where you want to do the inferencing. And that’s — it seems like that’s going to be working very well on our edge platform with CPUs.
Rishi Jaluria: Got it. That’s really helpful. And then just in terms of the StackPath and Lumen contracts, again, I appreciate the color in terms of your set of assumptions. Can you walk us through what you can do on your part to ensure those customers, whether they’re net new or existing Akamai customers that maybe we’re trying to use a multi-CDN approach, what you can do to get those customers to stay and to not turn off on to other competitors or even just kind of figure out how to reduce the peaty.
Tom Leighton: Yes. Good question. I think first is the approach that we took, right? We got an inbound request from these 2 companies, and some of them are customers we know and have had a long relationship with. And providing an orderly transition is step 1 in the relationship, right? So now you get a warm relationship and introduction to it’s a new customer, and certainly, for the customers that we have. They obviously know us — and you get — instead of having a jump ball in the open market, you have a chance to sit down with the customer and understand what their needs are, understand what’s going on with their contracts, length of contracts, pricing, et cetera. And you could just set up a relationship and just go through a normal sales cycle effectively.
I think we’ve got a pretty good track record, and I’m certainly understanding the customers. And if I think about sort of the weighted average of where the revenue is, we’ve obviously gone and talked to a lot of those customers. We obviously have had a long-term relationship with a lot of these folks. So I think we have a fairly high degree of confidence in the numbers that we’ve put out. Obviously, things can change. But based on our experience with a lot of these folks and with some of these new folks that we have not worked with us in the past, we have an awful lot of other services to offer that they didn’t have before. And especially with security being such a big topic these days. We’re hearing from some of these customers that they’re glad that they have a relationship with us now.
Operator: The next question is from Ruby Kessinger of D.A. Davidson.
Ruby Kessinger: Ed, I just want to quantify the TSA impact. It seems to be about 1.5 points impact cash gross margins in Q4. Is that accurate? And just to be clear, it sounds like that, that TSA ends at year-end? Or when exactly does that TSA add?
Tom Leighton: Yes. So there might be just a tiny bit that goes into Q1 just depending on how the migration goes. Hopefully, we can be done with it by the next couple of months. Yes, it’s just under 1.5 points. I rounded down to a point, but you’re right. If you just take the amount divided by the revenue, it’s about 1.3%, 1.4%.
Ruby Kessinger: Okay. Got it. And then if I just look at your delivery guidance, which, I guess, is implied by your compute security revenue guidance, and then I back out the expected $17 million to $20 million of revenue from Lumin and StackPath in Q4, it implies delivery revenue flat to down in Q4 versus Q3. So is any extra conservatism in the Q4 delivery guide just based on what you’re seeing from traffic trends or any pricing pressure to call out? Or why not — why aren’t we seeing a typical seasonal Q4 uplift in delivery revenue ex those acquisitions.
Tom Leighton: Yes. That’s a good question. First of all, there’s $4 million of delivery in the Q4 number so that you have to look at the net delta between the 2. Also, there’s about an $8 million headwind from FX. So there is some implied growth in the delivery number. As I talked about there’s a high degree of uncertainty in Q4 as it relates to traffic, right? You’ve got — typically, we see a big seasonality in terms of retail and the media side of the business. Retail as we’ve gone to 0 overage is less impactful. We do still see some bursting. I think we’re being probably a bit more cautious, especially on the commerce side of the equation for now. But there is some implied growth in there as you factor in those other 2 items I just mentioned.
Ruby Kessinger: Okay. That’s helpful.
Operator: The next question is from William Power of Baird.
Q – Unidentified Analyst: This is Yan Simoes on for Will. So first of all, just looking at delivery, how are you thinking about Q4 delivery trends this year versus normal seasonality given all the questions around the consumer and media activity in general? Any color there would be great.
Tom Leighton: Yes. As I just mentioned on the last question that we’re probably a bit more cautious in terms of our outlook certainly with retail. I mentioned earlier in an earlier question that we’ve seen an uptick in bankruptcies. We do have a lot of customers that are on the 0 overage for commerce. So we’re going into disease. We haven’t seen it yet. It usually starts right after Thanksgiving. We’re just being a bit more cautious. And in terms of the media cycle, — we did see a little bit of gaming activity. Gaming has been light for the last 1.5 years. A couple of years ago, we saw a big console refresh cycle. We don’t — not expected to see that. So I’d say we’re going into this quarter, probably a bit more cautious than we normally have in Q4 and certainly what we’ve seen in the past.
Operator: The next question is from Michael Elias of TD Cowen.
Michael Elias: First, earlier you talked about portability, particularly for cloud-agnostic workloads. I guess my question for you is, as you think about moving and garnering more share here, what are the steps that you could do to pretty much increase the ease of portability of workloads? And as part of that, through your data center deployments, do you have essentially direct cross connect into the cloud on ramps of the major cloud providers to essentially help facilitate the movement of workloads to your platform? And then I have a quick follow-up.
Tom Leighton: Yes. Great question. I think partners are really helpful there because they do a lot of the work in the first place to get into the third-party cloud provider. And today, some of them move between third-party cloud providers. And so there is a natural partner ecosystem that can be helpful porting that to Akamai. And I think they’re finding that the terms are even more favorable for them as well as for the partners. Also, as we grow the ecosystem of third-party capabilities, which with our qualified compute partner program we’re doing early focus on media there, which is our initial focus in terms of applications to move to Akamai. And yes, what it depends on the data center and the third-party cloud provider, but we do have direct connections in many cases.
And of course, we operate, as I mentioned, one of the world’s largest backbones and in a position to also make direct connections to major enterprises, which can help quite a bit. And then just as a quick follow-up. Just curious if you could talk a little bit about the M&A environment that you’re seeing. I know you have that your security growth guidance that you gave at your Analyst Day, which includes some M&A. Just curious, any thoughts around the M&A environment, particularly for security? And maybe as part of that, just comments on valuation. Yes, valuations in the companies that we’re most interested in are still extremely high. And you see some of the recent acquisitions that have been done at very high revenue multiples. So not a lot of change yet.
