Loading…
Loading…
Good afternoon, and welcome to F5 Inc.'s Third Quarter Fiscal 2025 Financial Results Conference Call. [Operator Instructions] Also, today's conference call is being recorded. If anyone has any objections, please disconnect at this time. I will now turn the call over to your host, Ms. Suzanne DuLong. Thank you. You may begin.
Hello, and welcome. I'm Suzanne DuLong, F5's Vice President of Investor Relations. We are here with you today to discuss our third-quarter fiscal year 2025 financial results. Francois Locoh-Donou, F5's President and CEO; and Cooper Werner, F5's Executive Vice President and CFO, will be making prepared remarks on today's call.
Other members of the F5 executive team are also here to answer questions during the Q&A session. A copy of today's press release is available on our website at f5.com, where an archived version of today's audio will be available through October 27, 2025. We will post the slide deck accompanying today's webcast to our IR site at the conclusion of today's call. To access the replay of today's webcast by phone, dial (877) 660-6853 or (201) 612-7415 and use meeting ID 13754228. The telephonic replay will be available through midnight Pacific Time, July 31, 2025. For additional information or follow-up questions, please reach out to me directly at s.dulong@f5.com.
Our discussion today will contain forward-looking statements, which include words such as believe, anticipate, expect and target. These forward-looking statements involve uncertainties and risks that may cause our actual results to differ materially from those expressed or implied by these statements. We've summarized factors that may affect our results in the press release announcing our financial results and in detail in our SEC filings. In addition, we will be referencing non-GAAP metrics during today's discussion.
Please see our full GAAP to non-GAAP reconciliation in today's press release and in the appendix of our earnings slide deck. Please note that F5 has no duty to update any information presented in this call. During today's call, Francois will speak to our Q3 highlights and our strategy and growth opportunities. Cooper will then review the details of our Q3 results and our outlook.
I'll now turn the call over to Francois.
François Locoh-Donou: Thank you, Suzanne, and hello, everyone. Our exceptional Q3 results highlight the strength of our business and F5's strong alignment with important secular trends. Customers are modernizing their data centers, adopting hybrid multi-cloud architectures and scaling to meet growing application capacity and performance needs. Our Q3 results demonstrate F5's position at the forefront of these transformative shifts. We delivered 12% total revenue growth, including 26% growth in product revenue, our strongest in 14 years. This performance is a testament to our team's execution, our continued innovation and the enormous trust the largest enterprises and service providers across the globe place in F5. F5's unique ability to deliver and secure every app, every API anywhere, on premises, in the cloud, at the edge and across hybrid multi-cloud environments is a significant competitive advantage. It powered strong demand across both hardware and software deployment models in Q3. Our systems revenue grew 39%, fueled by data center modernization, increased capacity requirements and adoption of our latest generation of hardware. Our software revenue grew 16%, driven by hybrid multi-cloud architecture adoption, which is contributing to continued strong software subscription renewals and expansions. We continue to operate the business with discipline. Combined with our strong revenue performance and continued operating margin leverage, we delivered exceptional EPS results in the quarter, reflecting 24% growth year-over-year. Looking ahead, our Q4 pipeline reflects continued strong demand in support of future forward architecture, data center modernization and an increasing range of application delivery and security services. As a result, we expect Q4 revenue in a range of $780 million to $800 million, implying approximately 9% revenue growth for FY '25. Cooper will provide a detailed overview of our Q3 performance and our outlook shortly. Before that, I will highlight early traction with our F5 application delivery and security platform. I will also recap recent-centric innovations and speak to some of our latest wins supporting AI workloads and infrastructure. We expect the long-term structural shift driving data center modernization will persist, continuing to reshape how companies invest in their IT infrastructure. Managing a skyrocketing number of apps and APIs across increasingly distributed environments is creating enormous challenges for IT teams. And preparing for and implementing AI is only making it worse. At F5, we have made it our mission to dramatically simplify this complex array of operational and cybersecurity challenges with the F5 application delivery and security platform. While there are platforms for endpoints, for network access and for cloud workloads, before the F5 ADSP, there was no platform that fully converges high-performance traffic management with advanced application and API security capabilities across hybrid and multi-cloud environments at scale. The F5 ADSP is unique in that it enables customers to, a, consolidate multiple delivery and security point solutions for applications and APIs on F5, leveraging a single platform with best-in-class capabilities, b, consistently deploy delivery and security across their hybrid multi-cloud environments via hardware, software and SaaS form factors; and c, dramatically simplify operations and reduce manual efforts by leveraging AI-powered analytics, insights and policy management. With these powerful benefits, it's no surprise that the initial customer response is very positive. In Q3, we won a number of deals that demonstrate the power and the benefit of F5's platform approach. For example, a North American health insurer chose F5 distributed cloud services and BIG-IP, consolidating multiple security and delivery vendors across both its on-premises and SaaS environments. Leveraging our unique hybrid multi-cloud approach, F5 is significantly reducing operational complexity while also providing enhanced security and improved API visibility. In another example, an airline in our EMEA region selected F5 to modernize its application delivery and security posture. Already an F5 BIG-IP customer, they faced escalating application growth and challenges, securing a rapidly growing base of APIs across multi-cloud environments. After extensive competitive evaluation, they consolidated both their on-premises and SaaS environments on F5, replacing their incumbent SaaS provider and deploying F5 distributed cloud services with comprehensive WAAP and secure multi-cloud networking. F5 is enabling the airline to securely connect and manage more than 100 B2B applications across multi-cloud environments, ensuring scalability, security and operational efficiency. In a final example, a South American service provider consolidated all of its multi-cloud and on-premises security services on F5. Through discovery during the sales process of what was initially a WAF-only opportunity, our team learned the full scale of the customers' challenges. The customer embraced the benefits of a unified F5 approach, consolidating their WAF, bot protection and API security on F5 distributed cloud services. F5 is streamlining their operations, eliminating multi-vendor complexity and reducing costs by consolidating across all of its on-premises and multi-cloud environments. These examples underscore the significant value F5 delivers to customers through our ADSP. While there is still work ahead to unlock the full potential of the platform, we are confident that our ongoing innovation will enable us to deliver even greater benefits to our customers. Over recent quarters, we have been unveiling a suite of groundbreaking customer-centric innovations that showcase the power of the F5 ADSP while also highlighting how we are leveraging advanced AI technologies to enhance customer experiences and drive business growth. These innovations fall into 2 categories. The first is AI for ADC. These are innovations that leverage AI to reduce the operational complexity of delivering and securing applications in a hybrid multi-cloud world. These innovations empower customers to unlock the full potential of F5, simplifying deployment and accelerating their ability to scale F5 solutions seamlessly. The second is ADC for AI. These innovations highlight how we are applying our strength in delivery and security to enable AI-driven applications. During Q3, we continued to make advances in both of these categories. In AI for ADC, we introduced advancements to our F5 AI Assistant. Our goal is to make the F5 AI Assistant an indispensable member of every NetOps and SecOps team. We are systematically building it around a simple, powerful loop, understand what's happening, prioritize what matters and act to solve it. To dramatically improve our customers' ability to understand potential threats and streamline prioritization, we are integrating technology from our Fletch acquisition. This integration will enable the F5 AI Assistant to provide a real-time contextualized view of the threat landscape, along with proactive, actionable recommendations that help cut through the noise. But insight without action is incomplete. That's why we enabled the assistant to act, starting with its new ability to generate rules from a simple prompt. This translates customer intent into secure performance code, making it easier for more customers to harness and leverage F5's unmatched programmability. This complete loop from deep understanding to automated action is how we move our customers from reactive firefighting to proactive control. In ADC for AI, we announced expanded capabilities for our F5 AI Gateway to prevent data leaks and deliver cutting-edge AI data protection. Additionally, we introduced new functionality for F5 BIG-IP SSL Orchestrator to classify and defend encrypted data in motion and block unapproved AI use. Through industry-leading technology partnerships, F5 is also providing integrated, secure and streamlined solutions to support complex emerging AI ecosystems. We recently announced advances with 2 of our AI-focused partners. In June, building on our prior work with NVIDIA, we announced we are expanding F5's performance, multi-tenancy and security capabilities for large-scale AI infrastructures. We are leveraging BIG-IP Next for Kubernetes running natively on NVIDIA BlueField-3 DPUs. Feedback from customers who have tested the solution has been very promising. One customer reported an 18% increase in HTTP throughput, an 11x improvement in time to first bite and a 190x boost in network energy efficiency. Another customer reported a 20% improvement in GPU utilization. We believe this level of performance improvement could be meaningful to customers looking to scale and enhance the performance of their AI infrastructures, and we continue to work towards validating the solution with customers and NVIDIA. With customers leveraging F5 to move incredible amounts of data at high speed to and from data stores for AI modeling and inferencing, our work with storage partners is also critical. In July, we announced that we are deepening our alliance with MinIO, advancing our AI application and data delivery solution to manage training and inference-driven data growth. These innovations and partnerships highlight F5's role in enabling AI-driven applications. Today, dozens of leading enterprises are deploying F5 to ensure AI workflows operate flawlessly, scale seamlessly and remain resilient against evolving threats. These AI use cases represent net new insertion points for F5 and leverage technology and expertise built over decades. Today, customers are leveraging F5 across 3 primary AI-related use cases. The first is AI data delivery. In these use cases, customers are deploying F5 in front of data stores to ensure secure high-throughput data ingestion for AI model training and inferencing. F5 is enforcing policy-based controls and eliminating bottlenecks, enabling performance delivery of massive data sets while safeguarding sensitive information. The second is AI runtime security. In these use cases, F5 is protecting AI applications, APIs and models with F5 WAAP and AI Gateway. F5 prevents abuse, data leakage and attacks like prompt injection while ensuring full visibility and control over how AI models and applications interact across environments. The third is AI factory load balancing. In these cases, F5 is optimizing traffic both across and within AI factories. In AI, performance is measured in terms of token throughput. By intelligently distributing AI traffic to maximize GPU utilization, F5 AI factory load balancing directly increases token generation, reduces time to first token and lowers cost per token. In Q3, we secured several new AI wins. In an AI data delivery use case, a government institution in our EMEA region needed to dramatically accelerate data ingestion for its AI platform, but its existing infrastructure could not meet the throughput and traffic modification demands. The core challenge was processing data at 100 gigabits per second while performing intensive SSL decryption and inserting critical data for audit trails. To solve this, they placed high-performance F5 BIG-IP hardware in front of their S3 data stores. BIG-IP is able to perform SSL decryption at line rate while preserving critical data for auditing and policy enforcement. This combination of hardware-based encryption, intelligent traffic modification and robust traffic management allowed the institution to meet its demanding performance goals and build a scalable foundation for future AI workloads. In an AI factory load balancing use case, an e-commerce provider based in our APAC region strategically deployed F5 BIG-IP and NGINX in key layers of their GPU-as-a-Service cloud platform. They use high-performance BIG-IP hardware to manage and secure the main entry point to their infrastructure, while F5 NGINX is embedded directly in front of each AI cluster to enforce fine-grained authentication and authorization. This ensures that as their cloud scales, performance and security scale with it, enabling them to provide a foundational AI resource for their customers. Collectively, these successes underscore F5's expanding leadership in the hybrid multi-cloud landscape and the tangible value our platform approach delivers to customers, empowering them to simplify operations, enhance security and accelerate innovation across their environments. Now I will turn the call to Cooper to elaborate on our Q3 results and our Q4 outlook. Cooper?
Thank you, Francois, and hello, everyone. I will review our Q3 results before I elaborate on our Q4 and FY '25 outlook. As Francois noted, we delivered a strong Q3, growing revenue 12% to $780 million, reflecting a mix of 50% global services revenue and 50% product revenue. Global Services revenue of $392 million grew 1%, while product revenue of $389 million grew 26% year-over-year.
Our software revenue grew 16% year-over-year to $208 million as customers continue to expand consumption and adopt additional capabilities across our platform. Our subscription-based software revenue grew 19% year-over-year to $185 million, representing 89% of our total software revenue. Perpetual license software totaled $23 million, down slightly year-over-year. Systems revenue totaled $181 million, up 39% year-over-year, with strength driven by tech refresh and data center modernization, including customers preparing for AI as well as competitive takeouts.
Revenue from recurring sources contributed 73% of our Q3 revenue. Our recurring revenue consists of our subscription-based revenue and the maintenance portion of our global services revenue. Shifting to revenue distribution by region. Our teams drove growth across all theaters.
Revenue from the Americas grew 13% year-over-year, representing 55% of total revenue. EMEA delivered 6% growth, representing 26% of revenue, and APAC grew 21%, representing 19% of revenue. Looking at our major verticals. Enterprise customers represented 70% of Q3's product bookings.
Government customers represented 15% of product bookings, including 5% from the U.S. Federal. Finally, service providers represented 15% of Q3 product bookings. Our continued operating discipline contributed to our strong Q3 operating results.
GAAP gross margin was 81%. Non-GAAP gross margin was 83.1%. Our GAAP operating expenses were $435 million. Our non-GAAP operating expenses were $381 million.
Our GAAP operating margin was 25.2%. Our non-GAAP operating margin was 34.3%, an improvement of more than 80 basis points year-over-year. Our GAAP effective tax rate for the quarter was 10.8%. Our non-GAAP effective tax rate was 14.4%.
This includes a discrete benefit associated with the filing of our annual federal income tax return during the quarter. Our GAAP net income for the quarter was $190 million or $3.25 per share. Our non-GAAP net income was $243 million or $4.16 per share, reflecting 24% EPS growth from the year-ago period. I will now turn to cash flow and balance sheet metrics, all of which were very strong.
We generated a record $282 million in cash flow from operations in Q3. CapEx was $9 million. DSO for the quarter was 42 days. Cash and investments totaled approximately $1.44 billion at quarter's end. Deferred revenue was $1.96 billion, up 10% from the year-ago period.
In Q3, we repurchased $125 million worth of F5 shares at an average price of $256 per share. As of the end of Q3, we had $1 billion remaining on our authorized stock repurchase program. Year-to-date, we have repurchased shares equivalent to 52% of our annual free cash flow. Finally, we ended the quarter with approximately 6,540 employees.
I will now speak to our Q4 outlook. With the exception of revenue, my guidance comments reference non-GAAP metrics. As Francois noted, visibility into our Q4 pipeline and customer demand remains strong. We expect Q4 revenue in the range of $780 million to $800 million, implying 6% growth at the midpoint.
We expect growth driven by tech refresh demand, data center modernization and adoption across our ADSP platform in support of customers' hybrid multi-cloud architectures. We expect non-GAAP gross margin in a range of 84% to 84.5%. We estimate Q4 non-GAAP operating expenses of $376 million to $388 million. We expect Q4 share-based compensation expense of approximately $57 million to $59 million.
We anticipate Q4 non-GAAP EPS in a range of $3.87 to $3.99 per share. As we look ahead to Q4 and beyond, we see customers increasingly leveraging both the breadth of our solutions across our portfolio and the flexibility only F5 offers to deploy in any form factor. As evident in our strong systems performance this year, customers' holistic hybrid multi-cloud approach spans hardware, software and SaaS deployments and they are increasingly investing in high-performance hardware to address factors, including resiliency, data sovereignty and AI readiness. With that backdrop, I will recap the implications of our Q4 guidance on our fiscal year 2025 outlook.
Our Q4 revenue guidance implies revenue growth of approximately 9% for FY '25. This is up from our prior guidance for growth of 6.5% to 7.5% and reflects robust systems revenue growth and software growth at or around 10% for the year. As implied by our Q4 guidance, we continue to expect FY '25 non-GAAP gross margin in a range of 83% to 84% and non-GAAP operating margin at or around 35%. We are adjusting our expected FY '25 non-GAAP effective tax rate to a range of 18.5% to 19.5% from the prior range of 20% to 22%. This change reflects the onetime benefit recorded in Q3. Our Q4 guidance implies FY '25 EPS in a range of 14% to 15% growth, up from our prior guidance of 8% to 10% growth.
The increase is driven by the improvement in our revenue outlook and the benefit from the tax rate change. Finally, we expect our share repurchases for the full year to be at or above 50% of our free cash flow. With that, I'll turn the call over to Francois.
François Locoh-Donou: Thank you, Cooper. F5 delivers and secures every app and API anywhere, on-prem, cloud, edge or hybrid environments, we lead in solving today's toughest business challenges. Our F5 application delivery and security platform is the first in the industry, offering consistent policies, full visibility and AI-driven insights. We are enabling customers to modernize data centers, embrace hybrid multi-cloud and scale for rising performance and security demands in an AI-driven world. Operator, please open the call to questions.
[Operator Instructions] Our first question comes from Tim Long with Barclays.
Two, if I could. First, on the hardware side. Could you talk a little bit about, obviously, a lot of strength here, a lot of things going on with maybe tariffs and pull-ins and end of life and good competitive environment. So can you kind of break down, you talked about the applications and the reasons there, but some of the other factors that maybe contributed and how you think the sustainability of that will be going into next year?
And then second, on the software side, could you just touch on kind of pipeline and how you're thinking about renewals and new deal activity heading into the end of the fiscal year?
François Locoh-Donou: Tim, why don't I start with your question on hardware and then Cooper will take the second question. On the hardware team, as you saw, the results were very strong, 39% growth year-on-year on hardware. And if I separate it, we have 2 dynamics. One, of course, is tech refresh, which is a dynamic in the business. And we expect at this point, giving some platforms that are going end of life in the next 12 to 24 months, and that's having an effect. So we're seeing a strong tech refresh this year. But there are other dynamics at play. And when we look at -- when we kind of separate our hardware revenue between what's coming from tech refresh and what is not associated with tech refresh, the part of the revenue that's non-tech refresh is actually growing faster than any other part of the hardware sales. So what are the dynamics that are driving that? It's really secular trends that we're seeing and specifically around hybrid cloud and data centers and around AI. So in hybrid and multi-cloud, what we're seeing is the -- basically, we're seeing a lot of customers that are embracing hybrid multi-cloud architectures. They are modernizing their data centers, investing in data center capacity to get ready for future growth in applications and future architectures. That's a trend that's providing tailwind to the hardware business. We're also seeing customers respond to -- especially in financial services, more regulation requiring customers to have better resiliency, better resiliency drives them to be able to fail over from public clouds to an on-premises environment and ideally vice versa. So that drives customers to have stronger infrastructure on-premises in hardware that they can control. That is the second secular trend. And of course, the third one is AI. And in AI, we're seeing customers invest both in AI use cases, which we are seeing already today. And this revolves around moving large amounts of data between data stores and AI applications or AI models. And we are -- we have won several dozens of customers now that are deploying AI models today and using F5 to be able to move their data to those models. But we also see a number of customers investing in what we would call AI readiness, meaning they're not yet in the full swing of having deployed AI models, but they want to have their data in a good place. They want to have enough capacity for what they expect are going to be large flows of data that need to be moved with AI, and they're investing in that capacity to be ready to deploy AI. So those are some of the trends that we're seeing in hardware that I think are pretty durable. Now you asked about how does that affect next year. We've seen very strong growth next year. Our current expectation would be that hardware next year would probably be up year-on-year. But obviously, we would expect more modest growth than what we have seen this year. To your last point, we have not seen any evidence of pull-in to date in hardware, and we have not seen any effect of any of the tariff discussions on our business, whether it be hardware or software. And with that, Cooper?
Sure. Yes. Thanks. So I'll speak a little bit to the software dynamics that we're seeing.
So we talked about that we had a large renewal base that was coming up in the second half of the year, and that gave us really good visibility. And what we saw again in this quarter was very healthy expansion against that renewal base. And so that's really coming in 2 forms. One, it's increased consumption that customers have been driving over the course of their prior subscription term.
And then at that time of renewal, that consumption then comes through at a higher contract value at the renewal as well as additional expansion across the portfolio. So we're seeing new use cases built in at that time of renewal. And so that's really where we're seeing the growth in the software. On the new, when we reference new, we're talking about either net new customers or net new software projects where the customer wasn't previously consuming in software.
That business is up year-to-date, but it was down a little bit in Q3. And some of that kind of goes back to what Francois was referencing around an appetite from customers to support their applications in an environment where regulatory concerns are more prevalent, resiliency is more kind of at the forefront and just generally a bias for more performance. And so at the margins, we -- there were a couple of opportunities that actually closed in hardware where previously the customer might have chosen to go with software. And so that's really kind of what's behind the business, coming from new projects coming through more in hardware than software.
And so we feel good about the year. We have, again, a strong base in Q4. And just based on the expansion trends that we've seen and continue to see, we feel pretty good about our view into Q4.
Our next question comes from Meta Marshall with Morgan Stanley.
This is Mary on for Meta. I had a question on gross margins. Are there any reasons for why gross margins came in at the lower end of the range despite the upside on revenue?
Yes. So it was pretty slightly at the low end of the range. That was really driven more by some of the high-performance use cases on the systems business. We had some deals that had FIPS compliance demands, which tend to have a slightly lower gross margin profile.
And then just broadly, the strength in the systems business, in general. Of course, hardware has still a very high gross margin profile, but it's not quite as high as software. And so that was the other dynamic that really was behind our gross margins. But as you see from our guidance for Q4, we expect gross margins to improve in the current quarter.
Our next question comes from Michael Ng with Goldman Sachs.
I just have 2. First, on the hardware piece. It's encouraging to hear that you expect systems revenue to be up next year. In the past, you've talked a little bit about how much of the installed base was on the legacy iSeries and VIPRION. I was just wondering if you could talk about where you are today and how much of that systems refresh we should expect over the next couple of years?
And then second, just on software, kind of relatedly, could you see software revenue growth in fiscal '26, just given the very strong term renewals that we've seen this year?
François Locoh-Donou: Michael, thank you for the question. We'll do the same thing. I'll start with hardware and ask Cooper to take the second part. On hardware, Michael, we have indeed some end-of-software-support dates that are coming up in '26 and early '27. We expect the refresh to continue to be strong, certainly throughout 2026 and beyond because customers continue to refresh even after these end-of-software-support dates. Well after that, they continue to refresh. So not the entire installed base is not refreshed at those dates. So we continue to expect to see that to be strong over the next 18 months. Now that said, Michael, there is -- we're seeing a part of our hardware business that is driven not by tech refresh motions, by other trends that I've just articulated around hybrid multi-cloud architectures, data center modernization and increasingly customers investing in AI readiness and AI use cases. We think these trends are less cyclical and more durable. And so we're excited to see these developments, and we'll see how this play out over time. But for the time being, we're really encouraged by what we're seeing outside of the tech refresh motion in our hardware.
And then on the software side, it's a little bit early. We wouldn't guide software for next year, but I'll kind of give you just a few dynamics for consideration is kind of what's a reasonable estimate from where we sit today. So I'll start with just a way to think about FY '26 in terms of growth rates and then I'll get into some of the dynamics that we're looking at. I think it's reasonable to assume software would grow in the mid-single digits for next year and then reaccelerate in the following year.
And just to kind of walk through some of the things to consider. So first, we're still seeing really strong consumption in that renewal motion. So as I said earlier, that's increased performance and consumption along with new use cases that are part of that renewal motion. And just as a point of emphasis, the growth comes from -- over time is coming from new use cases and that increased consumption that's embedded in the renew and expand motion.
So it's not simply repeat business. Then a second dynamic, and we talked about this a little bit on the last call in April, is that the subscription base that comes up for renewal in FY '26, that base largely comes from our software revenue from FY '23 because of that 3-year renewal cycle. And so our FY '23 software sales were roughly flat over FY '22. So that represents a bit of a math headwind on that subscription base where we do that renew and expand motion. And then that same dynamic becomes a tailwind into FY '27, and it's why we would expect the growth rate to reinflect from there.
And then the last dynamic, we touched on it a little bit is just some of the evolving customer preferences around deployment models. And this is really one of the big strengths around our ADSP platform is that we do give customers a choice in how they want to deploy. And so when we talk about hardware and software, something to keep in mind is that those are not products, they're delivery models. BIG-IP is a product.
And some customers are choosing to deploy BIG-IP in that hardware form factor just because of the evolving needs for more performance. And so those are all just things that we consider as we look ahead to next year, and then we'll see how that plays out.
Our next question comes from Samik Chatterjee with JPMorgan.
This is Priyanka Thapa on for Samik. Great job this quarter. I got a couple of questions. First of all, on the concept of software next year, what does the new business pipeline for software look like next year? And is that kind of incremental to your expectations of mid-single-digit software growth? And I have a follow-up.
Sure. Thank you, Priyanka. So yes, our pipeline right now, it's healthy. So it's early again.
I mean, our pipeline gives you really good visibility into the current quarter and then decent visibility into the next quarter. And so as you get beyond that, that's going to be business that we are surfacing now as we're engaging with our customers. But generally, I would say that as we do our planning for next year, we feel good about new opportunities for software. And again, these are either net new customers or customers that are consuming for the first time in software.
So that's kind of factored into some of our thinking around the growth rate for next year. But of course, the majority of our software revenue now comes through that renew and expand motion, which is great for visibility, and that's where we would expect the majority of the growth to come from next year.
All right. And on to my second question, you anticipate hardware to grow strongly in 2026. How much of that strength is this newfound shift where people are using systems instead of software that you would otherwise expect for them to use software like you saw in this particular quarter? Was there -- was this unexpected? And is this a trend that you think might continue?
Yes. So we would expect hardware to grow, albeit it will be at a more modest growth rate than what we're seeing this current year because clearly, we're well into the 20% growth rate year-to-date for the current year, but we expect continued growth next year. I would say that there is a portion of it that is coming from customer preferences to moving into a hardware model. I don't think that's the main driver.
It's really both tech refresh and some of the dynamics where customers really need more performance and they're trying to scale out their data center capacity to support those performance needs. And then at the margin, there are cases where customers may choose a hardware deployment model in lieu of what previously they may have been thinking software for the deployment model.
Our next question comes from George Notter with Wolfe Research.
I was just curious on kind of an update on some of the newer products. I was thinking about the NVIDIA DPU product. I think you mentioned it earlier in the monologue. Can you just remind me, is that GA now?
How have you guys priced that? How significant can that be in the context of your model? Just anything you can say on progress there? And then a similar question on the AI Gateway platform.
I know you're just getting that into the marketplace right now. I'm just curious on initial views, feedback on the product. What perspectives can you share?
François Locoh-Donou: Thank you, George. Let me start with the partnership with NVIDIA. Our solution is GA, and we are -- and as a reminder, it's really taking the BIG-IP software that we have refactored to work on ARM architectures and specifically integrate with NVIDIA's BlueField-3 DPUs. And the solution is GA, and we're now engaging customers in proof of concept to validate the benefit in their production and architectural environment. And the early feedback from customers around these proof of concepts and tests are generally very positive. I mentioned a couple in my prepared remarks earlier, but everything in AI, as you know, is about tokens and time to first token, the cost per token, the GPU utilization and the number of tokens one can generate in a given GPU infrastructure. And we're finding opportunities to basically increase all these metrics and increase the efficiencies of these AI factories. Now it's very early days because whilst we're doing these proof of concepts, we think the real strength of the value proposition comes when customers go into inferencing for their AI workloads. And most customers are not there yet. So the target customers for this solution are either the GPU-as-a-service providers who are selling or renting GPUs to customers and will want to make that infrastructure as efficient as possible or enterprises that will build self-hosted AI factories and will also want to make their own investments in AI factories as efficient as possible. We think -- so we think as enterprises start building more of these AI factories and really moving to inferencing these factories, then our opportunity to move more to revenue will be more concrete. But the technical validation has to proceed that with customers, and that's what we're doing, and it's really promising. As it relates to the AI Gateway, that solution is also early in the market. We have, I think, more than a dozen proof of concepts in place with customers. This really is about routing AI traffic and processing AI traffic at Layer 7 to really understand what is the cost per token that customers have, the cost of this model versus that model, providing security to this AI traffic. And we think this is going to be -- the whole area of AI delivery and security, we think is going to be a pretty important area in AI because if you step back from it, web infrastructure has really been about largely processing packets. And we think that -- which is largely a layer 3 activity. And we think that AI infrastructure is really going to be about delivering and securing token. And that is the Layer 7 activity. And we think F5 with our Layer 7 expertise built over decades is going to be really well-positioned for that. So more to come on the AI Gateway over time and what we're going to do in AI security, but it's an area that we are going to invest in and take a position.
Our next question comes from Matthew Hedberg with RBC Capital Markets.
This is Mike Richards on for Matt. Two quick ones for me. Just first, anything to call out on Fed? I know you guys were expecting there might be some pull-in in Q3 and anything to call out as we're 1 month into the Fed fiscal year-end? And then secondly, any guardrails you put around free cash flow looking into next year as we think about the R&D credits?
Yes. So I'd say Fed has been right at kind of at our plan for the year. Q3 was a little bit softer. I think there were a couple of projects that pushed out or were downsized kind of related to some of the government efficiency initiatives.
But broadly, the pipeline is still healthy for federal, so nothing else that we would really call out. And then can you remind me the question on free cash flow?
Yes. Just as we're thinking about the tax changes around the R&D credits, just anything around free cash flow as we look to next year?
No, there wouldn't be a big impact from what we see in our business right now. There are certain elections that you have the discretion to take. So that's something we're evaluating right now, but no material change from where we sit today.
Our next question comes from James Fish with Piper Sandler.
Nice print here. And Francois, you keep talking about the non-refresh piece on systems growing faster, but let's be frank, most of the systems is refresh. So really, the crux of my question though is can you just help us with how big essentially the non-refresh piece is and really like trying to understand how much of an AI benefit you're seeing today?
François Locoh-Donou: So Jim, let me just resize this for you. About 2/3 of the -- what we're seeing in hardware deployment models is what we would attribute to tech refresh and about 1/3 is what we would attribute to non-tech refresh. Just to give you a little bit of the weighting of what we are seeing this year. Now to your second question about AI, there really are a couple of answers to that. There is what we are seeing as direct AI use cases where our sales teams working with customers can actually attribute a project or a certain order specifically to a net new AI initiative. And today, I would say that this is still small in single-digit millions of dollars on a quarterly basis, several dozens of wins over the last several quarters. There is, however, another portion of AI, which is harder to estimate exactly, but it's kind of indirect AI spend coming to us. And this is where customers are investing in having higher capacity for data connections that may be related to an AI use cases, but the folks who are interacting with F5 may not be involved directly in this AI use case and don't know about the ultimate use of the technology. And we're calling that internally kind of shadow AI use cases. We think there are more and more of these growing, but it's difficult to give you an exact sizing of how much that is impacting the business.
Understood. And maybe if I could follow up there. Appreciate that added color. How are you thinking about maybe the magnitude of customers that on tech refresh could actually become more virtual or even DCS given nearly -- and given the nearly 40% growth here, and we've talked about this in a piece we did earlier this week, but what inning of refresh do you think we're actually in?
Yes. So Jim, we're pretty early in terms of the refresh opportunity for the iSeries and VIPRION product families. Well over half of the base is still on those families. And as Francois said that every customer is different in terms of their timelines, we would expect that base to get refreshed really over the next 2-plus years.
So we think that it's a good growth opportunity for FY '26 and into '27. And then you would expect that to start to tail off in FY '28. Where we -- we're also -- we're not seeing a lot of migration from those legacy systems into software form factors. I think the dynamic that we're actually seeing right now is where customers are really giving more consideration to their performance needs as they evaluate between hardware and software.
Our next question comes from Simon Leopold with Raymond James.
I guess one of the things I'm curious about is you've cautioned us in the past about your service provider or telco vertical having a tendency towards lumpiness. And this was a particularly strong quarter. It looks like it was up more than 40% sequentially and nice year-over-year growth. How should we be thinking about that?
Should we take this as more of a one-off? Or is there some new trending that we should think about? And then I've got a quick follow-up.
François Locoh-Donou: Thank you, Simon. The service provider business continues to be lumpy, and it is tied to specific projects at some of our large carrier customers. And so I wouldn't read too much into a single quarter trend in service provider, whether it's a big quarter with things going north very quickly or a down quarter. Generally, what we're seeing with service providers, Simon, is the promise of 5G hasn't fully materialized, I would say, for them or for equipment providers to a large extent, in part because there hasn't been the takeoff that we expected. We continue to see service providers invest in 4G to 5G infrastructure. Over the last couple of years, they have tended to sweat their assets significantly. We have seen that relaxed a little bit over the last couple of quarters. But I don't think the dynamics are fundamentally changing in that segment.
And then I want to see if you could talk a little bit more about how to think about the longer-term trends or fiscal '26 expectations for services in that I would have thought we'd see some correlation to the stronger hardware business, but maybe there's a lag effect. And I'm assuming there's very little correlation to your software stand-alone business. So how should we think about services trending?
Yes. So you're actually right. There is a lag effect on product. And so we would expect to see the growth rate kind of come up from where -- what we just reported with the 1%.
I think there's kind of an outlier dynamic that's behind the deceleration in the last couple of quarters, which is really around the kind of the last of the refresh from the prior product family, those laggard customers that have finally retired some of those end of technical support units. And so those came out of the maintenance base that drives our services revenue. That's now behind us. And so what we do expect is to see the services revenue start to grow now from that lag effect on the strength of the product revenue.
And you can see that in the deferred revenue, which is up 10% year-over-year. The short-term deferred revenue, which is really kind of a proxy for the maintenance revenue growth, that's up 5% year-over-year. And so we do expect to see the services growth rates to be a little bit better into FY '26.
Our next question comes from Tal Liani with Bank of America.
It's Tomer Zilberman on for Tal. Two questions for you. Maybe going back to hardware, but asking it a different way. I believe last quarter, you mentioned you expected hardware systems to moderate.
So what changed in the last 90 days for that to accelerate? Was it a better tech refresh than you initially expected? I know you called out last quarter that you were seeing competitive displacements in hardware. Are you still seeing that?
François Locoh-Donou: Yes. So yes, to the last point, yes, we still continue to see competitive displacements. Our value proposition is very strong relative to competitors, both with our current offering and our road maps. And so that continues. I think what's changed over the last 90 days is we saw more of these secular trends gathering pace on customers investing in more capacity in their data centers, in part because they're embracing hybrid multi-cloud, in part because they are sometimes getting more resiliency in place, sometimes in response to regulation in certain sectors and in part because of getting ready for AI. We see customers repatriating some apps. We see customers repatriating data in readiness for AI and doing so because the cost of having this data in public clouds and accessing this data are quite prohibitive. So we see more customers also doing data repatriation and needed to securely connect these data stores that are on-prem with their AI applications. So some of these trends, when you aggregate them, created a positive surprise on our hardware deployment model. That said, we've talked a lot on this call about hardware versus software. I really want to make sure that what we convey to you is what we're seeing is the power of F5's platform, our application delivery and security platform, which really brings all of these deployment models together, hardware, software and Software as a Service in a single platform. Increasingly, we're seeing is that's what's differentiating us in the marketplace. So this quarter, we are happy about the numbers, of course, but we are even happier about where we are winning. We are winning in customers that may have had a single SaaS vendor and/or a single hardware vendor, and we are consolidating all that spend on F5 because we can deliver in hardware, in software and SaaS. We are seeing customers that have point security solutions, maybe in software or hardware, and they're consolidating on F5 because we bring this platform that brings all these form factors together, all the delivery and security capabilities together in a single pane of glass to make the operation of these technologies much easier. And increasingly, that's what you're seeing from large enterprise customers is they want their life to be simplified and the platform approach that we have taken speaks to that pain point, and we're getting significant traction as a result of that. So that's what is exciting for us in the business is really the early traction we're getting on our platform strategy and the wins across multiple form factors.
Got it. Maybe as a follow-up, Francois, on the AI piece, you mentioned that it's still early days as we're awaiting enterprises to really build out these AI factories. We generally view that phase maybe by the end of 2026, moving into 2027. My question for you is, do you see that adoption timeline actually accelerating as we move more into the world of Agentic? And then generally speaking, how big do you view the opportunity in AI delivery versus AI security?
What's going to be more beneficial to you?
François Locoh-Donou: Well, that is -- both are great questions, and I'll pick up my crystal ball to give you an answer. Of course, it's very difficult to extrapolate when you're very, very early stage in markets to make big predictions about the impact. But I would say AI data delivery is probably more immediate for us because it's things that F5 has always done, the ability to process traffic at Layer 7 and securely connect applications to applications or users to applications. Now we have to connect data stores to applications. It requires new protocols, but F5 is ideally positioned to do that. And the more customers need high performance and high scale, the more they come to F5 because of our ability to do that at high speed and to support protocols like MC3 and other protocols that are specific to AI in general and increasingly Agentic AI. AI security, we think, especially in the case of F5, AI runtime security. Protecting AI workloads in production, we think it's going to be a very substantial opportunity for F5. Again, that AI Layer 7 security is going to be extremely important where you have to look at every token and make sure that it's going in the right place and there's no malicious prompt injection and other threats that are dealt with. We think that is going to be a very large opportunity. I think it will develop a little -- it will take a little longer to develop than AI data delivery, which we're starting to see right now. But we're starting to see customers already be very sensitive to securing their AI workloads and starting to investigate what solutions they could have to do that. So we think that market is absolutely going to happen. The pace at which it happens is less clear for us.
Our next question comes from Ryan Koontz with Needham & Company.
First, just a clarification on when you talk about tech refresh, I assume you're talking about F5 to F5 legacy to modern product. Second question I have is regarding your modest price increase you've talked about phasing in. Kind of where are you in terms of that making its way through the model? And what's been the customer feedback on that relative to some of your competitors' moves?
Sure. Yes. When we talk about tech refresh, we're talking about refreshing F5. We also have had good success displacing competitors. So there is a kind of a separate refresh motion around competitive installed base, but that's not part of that dynamic when we're speaking specifically to tech refresh.
And then in terms of the price increases, so yes, we announced a price increase in January that -- so we're starting to see that come through in the numbers. I think that customer reception has been reasonable. We've seen other peers in the space that have had much more aggressive pricing practices that have frankly turned off a lot of customers, and that's driving some business our way. We want to make sure that we're delivering value commensurate with the price increases that we introduced in January.
So we feel pretty good about where we sit from that perspective.
We have reached the end of the question-and-answer session. I'd now like to turn the call back over to Francois Locoh-Donou for closing comments.
François Locoh-Donou: Thank you for joining us today. We look forward to seeing many of you during the quarter and to discussing F5's growing role in the broader hybrid multi-cloud landscape. Thank you.
This concludes today's conference. You may disconnect your lines at this time, and we thank you for your participation.