F5, Inc. (FFIV) 2026 Q3 法說會逐字稿

Good afternoon, and welcome to the F5 Inc. third quarter fiscal 2026 financial results conference call.(Operator Instructions) Also, today's conference is being recorded. If anyone has any objections, please disconnect at this time. I'll now turn the call over to Ms. Suzanne DuLong. Ma'am you may begin.

下午好，歡迎參加 F5 Inc. 2026 會計年度第三季財務結果電話會議。（接線員指示）另外，今天的會議將進行錄音。如有任何異議，請此刻中斷連線。現在我將把電話交給 Suzanne DuLong 女士。女士，您可以開始了。

Hello, and welcome. I'm Suzanne DuLong, F5's Vice President of Investor Relations. We are here to discuss our third quarter fiscal year 2026 financial results. Francois Locoh-Donou, F5's Chairman, 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.

各位好，歡迎參加。我是 Suzanne DuLong，F5 投資人關係副總裁。我們今天在此討論 2026 會計年度第三季財務結果。F5 董事長、總裁暨執行長 Francois Locoh-Donou，以及 F5 執行副總裁暨財務長 Cooper Werner，將在今天的電話會議中發表事先準備的談話。F5 管理團隊的其他成員也在現場，將於問答環節回答問題。

Today's press release is available on our website at f5.com, where an archived version of today's audio will be available through October 26, 2026. We will post the slide deck accompanying today's webcast to our IR site following this call. To access the replay of today's webcast by phone, dial (800) 770-2030 or 609-800-9909 and use meeting ID 6076834. The telephonic replay will be available through midnight Pacific Time, July 28, 2026. For additional information or follow-up questions, please reach out to me directly at s.dulong@f5.com.

今天的新聞稿可於我們的網站 f5.com 取得；今天音訊的存檔版本將提供至 2026 年 10 月 26 日。本次電話會議結束後，我們將把配合今日網路直播的簡報投影片上傳至我們的投資人關係網站。若要透過電話收聽今日網路直播的重播，請撥打 (800) 770-2030 或 609-800-9909，並輸入會議 ID 6076834。電話重播將提供至 2026 年 7 月 28 日太平洋時間午夜。如需更多資訊或有後續問題，請直接透過 s.dulong@f5.com 與我聯繫。

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 have 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 reference 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.

我們今天的討論將包含前瞻性陳述，其中包括 believe、anticipate、expect 與 target 等字眼。這些前瞻性陳述涉及不確定性與風險，可能導致我們的實際結果與這些陳述所表達或暗示者有重大差異。我們已在宣布財務結果的新聞稿中摘要可能影響我們結果的因素，並在我們向 SEC 提交的文件中加以詳述。此外，我們在今天的討論中將引用非 GAAP 指標。請參閱今日新聞稿以及財報簡報投影片附錄中的完整 GAAP 與非 GAAP 調節表。

Please note that F5 has no duty to update any information presented in this call. I will now turn the call over to Francois.

請注意，F5 無義務更新本次電話會議中所呈現的任何資訊。現在我將把電話交給 Francois。

Thank you, Suzanne, and hello, everyone. Q3 was another outstanding quarter. We delivered 19% product revenue growth, driving 11% total growth. We have now delivered eight consecutive quarters of double-digit product growth, a signal that F5 is precisely where the market needs us to be. The reason is straightforward.

謝謝你，Suzanne，各位好。第三季又是一個表現非常出色的季度。我們實現了產品營收 19% 的成長，帶動總營收成長 11%。我們現在已連續八個季度達成產品雙位數成長，這顯示 F5 正好處在市場最需要我們的位置。原因很簡單。

The world runs on applications, including a new generation of AI-powered applications, and the world is changing fast, driving higher demand for application security and delivery. AI is accelerating everything, more traffic, more APIs, more distributed architectures and a dramatically larger attack surface. The complexity our customers are navigating has never been greater.

世界由應用程式驅動，包括新一代由 AI 驅動的應用程式；而世界正快速變化，推升對應用程式安全與交付的需求。AI 正在加速一切：更多流量、更多 API、更分散式的架構，以及大幅擴大的攻擊面。客戶所面對的複雜度從未如此之高。

F5 targets this complexity, simplifying and unifying application delivery and security across hybrid multi-cloud environments. We sit at the front door of the world's most critical applications, delivering and securing every app and every API. It's a position we've earned over 3 decades and 1 we are strengthening everyday. F5 is at the intersection of three secular megatrends. Hybrid multi-cloud adoption, the expanding threat landscape and the AI inference inflection.

F5 針對這種複雜性，致力於在混合多雲環境中簡化並統一應用程式交付與安全。我們位於全球最關鍵應用程式的「前門」，交付並保護每一個應用程式與每一個 API。這是我們在超過 30 年間所建立的地位，而且我們每天都在持續強化。F5 位於三大長期結構性巨趨勢的交會點。混合多雲採用、威脅版圖擴大，以及 AI 推論拐點。

Our strong Q3 results reflect this trend and F5's alignment with where customers are investing. They also reflect the actions we are taking to drive faster growth. We are converting hybrid multi-cloud adoption into expansion opportunities, competitive displacements, digital sovereignty wins and platform consolidation to ensure we are capitalizing on heightened demand for best-in-class application and API security. And we are building AI momentum.

我們第三季的強勁表現反映了這些趨勢，也反映 F5 與客戶投資方向的一致性。同時也反映我們為推動更快成長所採取的行動。我們正把混合多雲採用轉化為擴張機會、競品替換、數位主權勝出，以及平台整併，以確保我們能把握對一流應用程式與 API 安全日益升高的需求。同時我們也在建立 AI 動能。

We had another strong quarter of wins across AI Data Delivery, AI Runtime Security and AI Factory Load Balancing. Our top line results tell that story clearly. We also continued to execute with discipline, delivering 14% non-GAAP earnings growth while continuing to invest in innovation and go-to-market capacity to sustain and extend our momentum.

在 AI 資料交付、AI 執行期安全，以及 AI 工廠負載平衡方面，我們又取得了多項強勁的勝利。我們的營收表現清楚地說明了這一點。我們也持續以紀律執行，在持續投資創新與市場推進能力、以維持並延伸動能的同時，實現了非 GAAP 盈利成長 14%。

Looking ahead, we see strong demand driven by durable tailwinds. As a result, we are raising our fiscal year 2026 outlook to approximately 9% to 10% revenue growth, up from 7% to 8%. And Cooper will cover more details about our outlook in his remarks. But first, let me bring our growth dynamics to life with customer examples from the quarter. Each one illustrates why the forces reshaping IT are driving growing demand for F5 and reinforcing our critical role at the application delivery and security layer.

展望未來，我們看到在持久利多因素帶動下的強勁需求。因此，我們將 2026 會計年度展望上調至約 9% 至 10% 的營收成長，先前為 7% 至 8%。Cooper 將在他的談話中提供更多展望細節。但首先，讓我用本季的客戶案例，具體呈現我們的成長動能。每個案例都說明了為何重塑 IT 的力量正推動對 F5 的需求成長，並強化我們在應用程式交付與安全層的關鍵角色。

Accelerating hybrid multi-cloud adoption is fueling multiple growth drivers for F5, across data center build-out and sovereignty, competitive displacement and platform consolidation. During Q3, a large energy and utilities provider with more than 1,000 mission-critical applications running on F5 expanded its footprint significantly in a major overhaul of its cloud infrastructure across three regions and multiple data centers. When problems with its cloud environment began putting critical services at risk, the customer repatriated those workloads on premises and onto F5 hardware. This shift reduced its dependency on third-party platforms and restored confidence in the reliability of its delivery infrastructure.

混合多雲採用加速，正為 F5 帶來多重成長動能，涵蓋資料中心建置與主權、競品替換，以及平台整併。在第三季，一家大型能源與公用事業供應商在 F5 上運行超過 1,000 個關鍵任務應用程式，並在其跨三個區域與多個資料中心的雲端基礎架構重大改造中，大幅擴展了其部署規模。當其雲端環境問題開始使關鍵服務面臨風險時，該客戶將這些工作負載回遷至自有機房，並部署到 F5 硬體上。此一轉變降低了其對第三方平台的依賴，並恢復了對其交付基礎架構可靠性的信心。

A European government agency wanted to standardize delivery and security across a growing multi-cloud environment while keeping select applications on premises for privacy and compliance. F5 delivered a sovereign-by-design architecture with consistent delivery, stronger security and simpler operations across their multi-cloud environment. The customer refreshed and expanded their BIG-IP footprint. They also added Distributed Cloud Services, including AI-powered WAF to automate policy and advanced threat protection.

一家歐洲政府機構希望在日益成長的多雲環境中，將交付與安全標準化，同時為了隱私與合規而將部分應用程式保留在自有機房。F5 提供了「以主權為設計核心」的架構，在其多雲環境中實現一致的交付、更強的安全性與更簡化的營運。該客戶更新並擴大了其 BIG-IP 部署規模。他們也新增了 Distributed Cloud Services，包括 AI 驅動的 WAF，用於自動化政策與進階威脅防護。

In an example that highlights both a competitive displacement with an AI data delivery win, we displaced an incumbent at a Fortune 100 global enterprise technology provider. The customer was looking to strengthen the delivery and security layer in front of its S3 compatible cloud storage service across 45 data centers worldwide. They selected BIG-IP for its ability to deliver the availability, resilience and security required to support their AI and data-intensive workloads at global scale. The expanding threat landscape also continues to create significant tailwinds for F5.

在一個同時凸顯競品替換與 AI 資料交付勝利的案例中，我們在一家《財富》100 強的全球企業科技供應商處取代了既有供應商。該客戶希望在其遍布全球 45 個資料中心、位於其 S3 相容雲端儲存服務之前的交付與安全層加以強化。他們選擇 BIG-IP，因其能提供在全球規模下支援其 AI 與資料密集型工作負載所需的可用性、韌性與安全性。不斷擴大的威脅版圖也持續為 F5 帶來顯著的利多動能。

In the last several years, the front lines of cyber defense have shifted to the application layer. Attacks on IT assets grew more than 60% over the last 3 years. But in that same time frame, the volume of application layer attacks surged 140%. This is where F5 plays and as attacks at this layer continue to grow, our relevance and opportunity grow too. During Q3, a multi-brand telecommunications provider replaced a fragmented collection of security tools with Distributed Cloud Services.

在過去幾年中，網路防禦的前線已轉移到應用程式層。過去 3 年間，對 IT 資產的攻擊成長超過 60%。但在同一期間，應用程式層攻擊量激增 140%。這正是 F5 的主戰場；隨著此層攻擊持續增加，我們的相關性與機會也隨之成長。在第三季，一家多品牌電信業者以 Distributed Cloud Services 取代了分散零碎的一系列安全工具。

The customer consolidated web application and API protection, bot defense and DDoS mitigation on F5. We displaced both a SaaS-only competitor and a cloud-native tool set. F5's automated API Discovery aligned with the customers' shift to micro services, reducing latency and keeping sensitive traffic off the public Internet. Our unified management layer lowered operational overhead supporting the customers' cost reduction goals and establishing F5 as an always-on security layer across their digital operations.

客戶在 F5 上整合了 Web 應用程式與 API 防護、機器人防禦以及 DDoS 緩解。我們取代了一家僅提供 SaaS 的競爭對手以及一套雲原生工具組合。F5 的自動化 API 探索（API Discovery）與客戶轉向微服務的方向一致，降低延遲並讓敏感流量不經由公用網際網路。我們的統一管理層降低了營運負擔，支援客戶的成本降低目標，並使 F5 成為其數位營運中隨時在線的安全層。

In a similar consolidation win, a large multinational health care company pursuing a hybrid cloud strategy, identified critical security gaps with their incumbent provider. Following a competitive evaluation, Distributed Cloud Services displaced the incumbent, providing web application and API protection, DDoS mitigation and bot defense on a unified platform. Since deployment, threats that previously evaded detection are now being identified and blocked. F5 improved security outcomes while reducing operational costs by consolidating multiple functions.

在另一個類似的整合勝利中，一家推動混合雲策略的大型跨國醫療保健公司，發現其既有供應商存在關鍵安全缺口。在競爭性評估後，Distributed Cloud Services 取代了原供應商，在統一平台上提供 Web 應用程式與 API 防護、DDoS 緩解與機器人防禦。自部署以來，過去能躲過偵測的威脅如今已可被識別並加以封鎖。F5 透過整合多項功能，在降低營運成本的同時改善了安全成效。

Frontier AI models have escalated the threat level for organizations globally, creating a new layer of urgency on top of existing security demand. Customers know the threat equation has changed, and early movers are already strengthening their defenses. For instance, a major European retail banking institution and existing F5 customer, accelerated plans to upgrade its application security infrastructure as the potential for AI-driven attacks escalated. The customer had been evaluating AI-based WAF solutions for nearly a year, and F5's AI-powered approach, including machine learning-based threat detection, granular per application policy control and behavioral bot defense differentiated us from their incumbent provider. In the first phase of their upgrade, they are deploying Distributed Cloud Services as the front door to their application environment, creating a consistent, scalable on-premises bot defense architecture.

前沿 AI 模型已在全球範圍內提高了組織面臨的威脅等級，並在既有的安全需求之上增添了一層新的急迫性。客戶知道威脅方程式已改變，而先行者已在強化其防禦。例如，一家歐洲主要零售銀行機構且為既有 F5 客戶，隨著 AI 驅動攻擊的可能性升高，加速了升級其應用程式安全基礎架構的計畫。該客戶評估以 AI 為基礎的 WAF 解決方案已近一年，而 F5 的 AI 驅動方法（包含以機器學習為基礎的威脅偵測、細緻到每個應用程式的政策控制，以及行為式機器人防禦）使我們有別於其既有供應商。在升級的第一階段，他們正部署 Distributed Cloud Services 作為其應用環境的前門，建立一致、可擴展的內部部署機器人防禦架構。

This foundation can be extended to every new environment they deploy, mitigating both agentic and traditional bot attacks. We saw the same urgency play out in Asia where one of the region's largest financial institutions and a long-standing F5 customer, accelerated its application security strategy amidst rising AI-driven attacks and zero-day vulnerabilities. The customer doubled down on F5 as a strategic security vendor deploying our AI-powered WAF to automate threat detection and accelerate virtual patching at scale.

此一基礎可延伸至其部署的每個新環境，同時緩解代理型（agentic）與傳統機器人攻擊。我們也在亞洲看到同樣的急迫性：該區域最大的金融機構之一、且為長期 F5 客戶，在 AI 驅動攻擊與零日漏洞上升之際，加速其應用程式安全策略。該客戶加碼將 F5 作為策略性安全供應商，部署我們的 AI 驅動 WAF，以自動化威脅偵測並在大規模下加速虛擬修補（virtual patching）。

Finally, the AI inference inflection is driving demand for F5, both indirectly as customers expand hybrid multi-cloud deployments and directly through our three AI use cases, AI data delivery AI runtime security and AI factory load balancing. Momentum is building across all three. In fact, our cumulative total customer count for direct AI use cases grew 50% just in Q3, underscoring the accelerating pace of demand. A few notable wins from the quarter illustrate our traction.

最後，AI 推論的拐點正推動對 F5 的需求：一方面是間接地因客戶擴大混合多雲部署；另一方面則是直接來自我們的三個 AI 使用案例——AI 資料傳遞、AI 執行期安全，以及 AI 工廠負載平衡。三者的動能皆在增強。事實上，我們用於直接 AI 使用案例的累計客戶總數僅在第三季就成長了 50%，凸顯需求加速的步伐。本季幾個值得注意的勝利展現了我們的進展。

In AI data delivery, we are accelerating secure AI data pipelines, ensuring storage, networking and data delivery and keeping AI clusters fully utilized across hybrid and multi-cloud deployments. Earlier, I highlighted an AI data delivery win and competitive takeout. In another sizable win during the quarter, a leading autonomous systems manufacturer selected F5 over a competitor to handle escalating data volumes and throughput demand. The customer previously deployed F5 to repatriate AI training data from public cloud to on-premises data centers deploying BIG-IP in front of their AI storage environment to improve latency.

在 AI 資料傳遞方面，我們正加速建置安全的 AI 資料管線，確保存儲、網路與資料傳遞，並在混合與多雲部署中讓 AI 叢集維持滿載運作。先前我已強調一個 AI 資料傳遞的勝利與競品替換案例。在本季另一個規模可觀的勝利中，一家領先的自主系統製造商選擇 F5 而非競爭對手，以因應不斷攀升的資料量與吞吐量需求。該客戶先前已部署 F5，將 AI 訓練資料從公有雲回遷至內部資料中心，並在其 AI 儲存環境前部署 BIG-IP 以改善延遲。

With exploding data volumes and throughput demand, the customer needed additional capacity. F5 outperformed the competitor, simplifying connectivity and storage sharing across their data centers and expanding our role as the trusted traffic layer between storage and compute. In AI runtime security, we are safeguarding AI applications, APIs and models from abuse, data leaks and attacks like prompt injection. We are also delivering real-time threat defense, red teaming models and robust guardrails.

隨著資料量與吞吐量需求爆炸性成長，客戶需要額外的容量。F5 的表現優於競爭對手，簡化其資料中心之間的連線與儲存共享，並擴大我們作為儲存與運算之間可信流量層的角色。在 AI 執行期安全方面，我們保護 AI 應用程式、API 與模型免於濫用、資料外洩，以及如提示注入（prompt injection）等攻擊。我們也提供即時威脅防禦、模型紅隊演練（red teaming）與強健的防護欄（guardrails）。

In Q3, a US-based professional sports league was rolling out AI-powered applications to analyze highly sensitive data, including player health information and team strategy insights. With competitive integrity and league trust on the line, they needed confidence that their applications could be proactively tested for vulnerabilities and protected continuously in production. The customer deployed AI Guardrails and AI Red Team on premises to validate applications before launch and defend them at run time. Our breadth of functional coverage beat out a competitor and enabled the league to standardize on F5 as their single trusted vendor.

在第三季，一個美國的職業運動聯盟正推出 AI 驅動的應用程式，用於分析高度敏感的資料，包括球員健康資訊與球隊策略洞察。在競賽公平性與聯盟信任攸關之際，他們需要確信其應用程式能被主動測試漏洞，並在生產環境中持續受到保護。該客戶在內部部署 AI Guardrails 與 AI Red Team，以在上線前驗證應用程式並在執行期間進行防護。我們在功能覆蓋面的廣度勝過競爭對手，並使該聯盟得以將 F5 標準化為其唯一可信供應商。

In AI Factory Load Balancing, we are optimizing traffic and GPU utilization both across and within AI factories, increasing token throughput, reducing time to first token and lowering per token costs. In the Q3 AI Factory Load Balancing win, a service provider in our EMEA region selected F5 to power its sovereign AI factory purpose-built to deliver GPU-as-a-service and AI-model-as-a-Service. To support these offerings at scale, the customer deployed BIG-IP Next for Kubernetes, Distributed Cloud Services and NGINX enabling secure multi-tenancy, traffic segmentation, token governance and large language model API endpoint protection.

在 AI 工廠負載平衡方面，我們在 AI 工廠之間與工廠內部最佳化流量與 GPU 使用率，提高 token 吞吐量、縮短首個 token 的時間（time to first token），並降低每個 token 的成本。在第三季的 AI 工廠負載平衡勝利中，我們 EMEA 區域的一家服務供應商選擇 F5，為其主權 AI 工廠提供動力；該工廠專為提供 GPU-as-a-service 與 AI-model-as-a-Service 而打造。為了在規模化下支援這些服務，客戶部署了 BIG-IP Next for Kubernetes、Distributed Cloud Services 與 NGINX，以實現安全的多租戶、流量分段、token 治理，以及大型語言模型 API 端點防護。

Before I pass the call to Cooper, I will speak to innovation at F5. The customer examples I shared highlight what we are hearing consistently from customers. They are modernizing their architectures and at the same time, the requirements for application delivery and security are rising. That's exactly why we are innovating to lead the next generation of capabilities and get them deployed with customers faster so they are ready as requirements intensify.

在我把電話交給 Cooper 之前，我想談談 F5 的創新。我分享的客戶案例凸顯了我們持續從客戶那裡聽到的訊息。他們正在現代化其架構，同時，對應用程式傳遞與安全的要求也在提高。這正是我們持續創新、引領下一代能力，並更快將其部署到客戶端的原因，讓他們在需求加劇時做好準備。

Frontier AI has fundamentally altered both sides of the security equation. In the post-mythos era, attackers can use AI to discover vulnerabilities, develop exploits and vary attacks at a speed and scale that was previously impossible. This is compressing the time between vulnerability discovery and exploitation and forcing every enterprise to rethink how it protects its applications.

前沿 AI 從根本上改變了安全方程式的兩端。在後神話（post-mythos）時代，攻擊者可使用 AI 以過去不可能的速度與規模來發現漏洞、開發攻擊利用（exploits）並變化攻擊手法。這正在壓縮從漏洞發現到被利用之間的時間，並迫使每一家企業重新思考如何保護其應用程式。

In this environment, security cannot depend only on identifying a vulnerability and issuing a patch after the fact. Customers increasingly need run time protection, ability to detect malicious behavior and stop attacks while their applications, APIs, models and agents are running. This is precisely where F5 is accelerating innovation.

在此環境下，安全不能只依賴於事後識別漏洞並發布修補程式。客戶日益需要執行期防護、偵測惡意行為的能力，並在其應用程式、API、模型與代理（agents）運行時就能阻止攻擊。這正是 F5 正在加速創新的領域。

Our AI-powered capabilities in Distributed Cloud WAF use a layered detection engine to assess the intent and risk of each request, moving customers beyond static signatures and manual policy tuning. Since launching in Q2, of 15% our Distributed Cloud WAF customers have adopted the capability. And of those, 75% are running in blocking mode versus monitor mode, that shift from monitor to block reflect growing customer urgency around zero-day threats, and it's warranted. In real-world use, our AI-powered WAF is already stopping zero-day attacks without requiring new signatures. We are applying the same urgency to the security of our own products.

我們在 Distributed Cloud WAF 中的 AI 驅動能力使用分層偵測引擎來評估每個請求的意圖與風險，讓客戶超越靜態特徵碼與手動政策調校。自第二季推出以來，我們的 Distributed Cloud WAF 客戶中有 15% 已採用此能力。而在這些採用者中，75% 以封鎖模式（blocking mode）運行，而非監控模式（monitor mode）；從監控轉向封鎖反映客戶對零日威脅日益增加的急迫性，而且這種急迫性是合理的。在真實世界的使用中，我們的 AI 驅動 WAF 已能在不需要新特徵碼的情況下阻擋零日攻擊。我們也以同樣的急迫性來強化自家產品的安全。

This quarter, we shipped our first hardened software release as part of a new monthly rather than quarterly cadence. We are hardening our software releases by applying advanced and preview frontier models to vulnerability discovery and remediation. As a result, we are finding and fixing issues faster.

本季，我們作為新的每月（而非每季）節奏的一部分，發布了第一個強化（hardened）的軟體版本。我們透過將先進與預覽版的前沿模型應用於漏洞發現與修復，來強化我們的軟體發布。因此，我們能更快地發現並修正問題。

More importantly, we are getting those fixes into customers' hands faster. This is a rigorous approach that we believe sets F5 apart in world of security. But delivering fixes faster only matters if customers can deploy them. That is why we also introduced new Fleet Management capabilities in F5 Insight.

更重要的是，我們正更快地將這些修補交付到客戶手中。這是一種嚴謹的方法，我們相信這讓 F5 在安全領域中脫穎而出。但更快交付修補只有在客戶能夠部署時才有意義。因此，我們也在 F5 Insight 中推出了新的車隊管理（Fleet Management）功能。

F5 Insight gives customers visibility into software versions, security posture and update readiness across the BIG-IP estates with guided workflows that simplify upgrades and patching. F5 Insight was launched last quarter and the strong early uptake is evidence that customers want a faster, more operationally practical way to reduce risk.

F5 Insight 讓客戶能夠透視 BIG-IP 資產群的軟體版本、安全態勢與更新就緒度，並透過引導式工作流程簡化升級與修補。F5 Insight 於上季推出，早期採用表現強勁，證明客戶希望以更快、且在營運上更務實的方式來降低風險。

Together, these innovations create a continuous defense model for the AI era. We are using AI to identify risk earlier, protecting applications at run time, delivering hardened software releases faster and helping customers deploy those protections across their environments. This is how F5 is turning the disruption created by Frontier AI into stronger security and greater resilience for our customers.

這些創新共同為 AI 時代打造了持續防禦模型。我們運用 AI 更早識別風險、在執行階段保護應用程式、更快交付強化後的軟體版本，並協助客戶在其環境中部署這些防護。這就是 F5 如何將 Frontier AI 帶來的顛覆轉化為客戶更強的安全性與更高的韌性。

Now I will turn the call over to Cooper who will walk through our Q3 results and our outlook. Cooper?

現在我把電話會議交給 Cooper，由他帶大家回顧我們的第三季業績與展望。Cooper？

Thank you, Francois, and good afternoon, everyone. Q3 marked another quarter of solid execution with both revenue and earnings well above the top end of our guidance. We continue to see strong demand with customers' hybrid multi-cloud adoption, driving expanding opportunity for F5. I will first review our Q3 results, followed by our guidance for Q4.

謝謝你，Francois，各位下午好。第三季再度展現穩健的執行力，營收與獲利均明顯高於我們指引區間的上緣。隨著客戶採用混合多雲持續推進，我們仍看到強勁需求，並為 F5 帶來擴大的機會。我將先回顧第三季業績，接著說明第四季指引。

We delivered a strong Q3, growing revenue 11% to $865 million, with a mix of 54% Product revenue and 46% Services revenue. Product revenue totaled $463 million increasing 19% year-over-year, while Services revenue of $402 million grew 3% year-over-year. Systems revenue totaled $240 million, up 32% over Q3 FY25.

我們交出強勁的第三季成績，營收成長 11% 至 8.65 億美元，其中產品營收占 54%，服務營收占 46%。產品營收合計 4.63 億美元，年增 19%；服務營收為 4.02 億美元，年增 3%。系統營收合計 2.40 億美元，較 2025 會計年度第三季成長 32%。

Our Software revenue of $223 million grew 7% year-over-year. Subscription-based software revenue totaled $201 million, up 9% year-on-year, representing 90% of our Q3 Software revenue. Perpetual license software totaled $22 million, down 4% year-over-year. Revenue from recurring sources contributed 69% of our Q3 revenue.

我們的軟體營收為 2.23 億美元，年增 7%。訂閱制軟體營收合計 2.01 億美元，年增 9%，占第三季軟體營收的 90%。永久授權軟體合計 2,200 萬美元，年減 4%。來自經常性來源的營收占第三季總營收的 69%。

Shifting to revenue distribution by region. Revenue from the Americas grew 11% year-over-year, representing 55% of total revenue. EMEA delivered yet another robust quarter with 27% revenue growth, representing 30% of revenue. Finally, APAC was down 11% against the very strong year ago period, representing 15% of revenue. Looking at our major verticals, enterprise customers contributed 71% of Q3's Product bookings.

接著看依地區的營收分布。美洲地區營收年增 11%，占總營收的 55%。EMEA 再度交出強勁的一季，營收成長 27%，占營收的 30%。最後，APAC 在去年同期基期非常強勁的情況下年減 11%，占營收的 15%。就主要垂直產業而言，企業客戶貢獻了第三季產品訂單（Product bookings）的 71%。

Government customers represented 19% of Product bookings, including 7% from US Federal. Finally, service providers contributed 11% of Q3 Product bookings. Our continued financial discipline contributed to our strong Q3 operating results. GAAP gross margin was 82.2%. Non-GAAP gross margin was 84.2%. Our GAAP operating expenses were $498 million. Our non-GAAP operating expenses were $426 million. Our GAAP operating margin was 24.7%.

政府客戶占產品訂單的 19%，其中美國聯邦政府占 7%。最後，服務供應商貢獻了第三季產品訂單的 11%。我們持續的財務紀律促成第三季強勁的營運成果。GAAP 毛利率為 82.2%。非 GAAP 毛利率為 84.2%。GAAP 營業費用為 4.98 億美元。非 GAAP 營業費用為 4.26 億美元。GAAP 營業利益率為 24.7%。

Our non-GAAP operating margin was 35%. Our GAAP effective tax rate for the quarter was 8%. Our non-GAAP effective tax rate was 12.4%. This is below our previously guided annual range primarily driven by discrete benefits associated with the filing of our annual federal income tax return and changes in unrecognized tax benefits during the quarter.

非 GAAP 營業利益率為 35%。本季 GAAP 有效稅率為 8%。非 GAAP 有效稅率為 12.4%。此數字低於我們先前提供的全年區間，主要是因為申報年度聯邦所得稅申報表所帶來的一次性利益，以及本季未確認稅務利益（unrecognized tax benefits）的變動。

Our GAAP net income for the quarter was $208 million or $3.62 per share. Our non-GAAP net income was $272 million or $4.73 per share reflecting 14% EPS growth from the year ago period. I will now turn to cash flow and balance sheet metrics. We generated $316 million in cash flow from operations in Q3 and free cash flow of $281 million. CapEx was $36 million. DSO for the quarter was 45 days. Cash and investments totaled $1.63 billion at quarter end.

本季 GAAP 淨利為 2.08 億美元，或每股 3.62 美元。非 GAAP 淨利為 2.72 億美元，或每股 4.73 美元，反映較去年同期 EPS 成長 14%。接下來我將說明現金流與資產負債表指標。第三季我們的營運活動現金流為 3.16 億美元，自由現金流為 2.81 億美元。資本支出（CapEx）為 3,600 萬美元。 DSO本季為 45 天。季末現金與投資合計 16.3 億美元。

Deferred revenue was $2.19 billion, up 12% from the year ago period. In Q3, we repurchased $100 million worth of F5 shares at an average price of $299 per share. We had $422 million remaining on authorized share repurchase program as of the end of the quarter. Finally, we ended the quarter with approximately 6,600 employees.

遞延營收為 21.9 億美元，較去年同期成長 12%。第三季我們以每股平均 299 美元回購 1 億美元的 F5 股票。截至季末，我們在已核准的股票回購計畫下仍有 4.22 億美元的剩餘額度。最後，本季結束時我們約有 6,600 名員工。

I will now speak to our outlook for Q4. We expect that the market trends we've outlined, hybrid multi-cloud adoption, threat landscape expansion and AI inference inflection will drive continued strong demand in Q4 and into FY27. As a result, we expect Q4 revenue in a range of $870 million to $890 million, reflecting just under 9% year-over-year growth at the midpoint.

現在我來談談我們對第四季的展望。我們預期先前所概述的市場趨勢——混合多雲採用、威脅版圖擴大，以及 AI 推論拐點——將在第四季並延續至 2027 會計年度推動持續強勁的需求。因此，我們預期第四季營收將介於 8.70 億至 8.90 億美元，中位數對應的年增率略低於 9%。

We now expect Q4 non-GAAP gross margin in the range of 83% to 84%, an improvement from our prior view. This improved outlook reflects a more favorable product mix towards higher performance systems, that carry stronger gross margins, along with lower Q4 component cost increases than our early estimates. While we are encouraged by the improved gross margin outlook for Q4, we are not revising our FY27 gross margin guidance of 80% to 82%, given the dynamic pricing environment for memory and storage components.

我們目前預期第四季非 GAAP 毛利率將介於 83% 至 84%，較先前看法有所改善。此一改善的展望反映產品組合更有利，偏向毛利率更高的高效能系統，同時第四季零組件成本的增幅也低於我們先前的初步估計。儘管我們對第四季毛利率展望改善感到鼓舞，但鑑於記憶體與儲存零組件的動態定價環境，我們不調整 2027 會計年度毛利率 80% 至 82% 的指引。

We estimate Q4 non-GAAP operating expenses of $430 million to $442 million. We expect Q4 share-based compensation expense of approximately $68 million to $70 million. We anticipate Q4 non-GAAP EPS in a range of $4.14 to $4.26 per share. Based on our strong Q3 results and our Q4 outlook, we now expect to finish FY26 ahead of our prior guidance. With our Q4 revenue outlook, we now expect approximately 9% to 10% revenue growth for the full year, up from our prior outlook of 7% to 8%.

我們估計第四季非 GAAP 營業費用為 4.30 億至 4.42 億美元。我們預期第四季以股份為基礎的薪酬費用約為 6,800 萬至 7,000 萬美元。我們預期第四季非 GAAP 每股盈餘（EPS）將介於每股 4.14 至 4.26 美元。基於第三季強勁的業績與第四季展望，我們目前預期 2026 會計年度的結束表現將優於先前指引。依據第四季營收展望，我們現在預期全年營收成長約 9% 至 10%，高於先前 7% 至 8% 的展望。

We continue to expect mid-single-digit Software revenue growth double-digit Systems revenue growth and low single-digit Services revenue growth for the year. We expect FY26 non-GAAP gross margin in the range of 83.5% to 84%. We expect non-GAAP operating margin in the range of 34.5% to 35.5%.

我們仍預期全年軟體營收為中個位數成長、系統營收為雙位數成長、服務營收為低個位數成長。我們預期 2026 會計年度非 GAAP 毛利率將介於 83.5% 至 84%。我們預期非 GAAP 營業利益率將介於 34.5% 至 35.5%。

We expect our FY26 non-GAAP effective tax rate will be in the range of 18% to 19% reflecting the impact of the tax rate benefit realized in Q3. We expect FY26 non-GAAP EPS in a range of $17.21 to $17.33 up from the prior range of $16.25 to $16.55. Finally, we expect our full year share repurchase to be at least 50% of our free cash flow.

我們預期 2026 會計年度非 GAAP 有效稅率將介於 18% 至 19%，反映第三季已實現的稅率利益影響。我們預期 2026 會計年度非 GAAP 每股盈餘將介於 17.21 至 17.33 美元，高於先前 16.25 至 16.55 美元的區間。最後，我們預期全年股票回購金額至少將達到自由現金流的 50%。

With that, I will pass the call back to Francois.

那麼，我把電話交回給Francois。

Thank you, Cooper. To recap briefly before we go to Q&A, our wins this quarter and our financial performance tell a consistent story. The market is moving toward F5. Hybrid multi-cloud is accelerating, threats are growing more sophisticated, and AI inference is becoming core to how applications run. These are durable structural shifts that are benefiting F5.

謝謝你，Cooper。在進入問答之前先簡要回顧一下，本季我們的勝利與財務表現傳達出一致的訊息。市場正朝向F5移動。混合多雲正在加速，威脅變得更為精密，而AI推論正成為應用程式運行方式的核心。這些是持久的結構性轉變，正在讓F5受益。

But we are not just riding these trends, we are using them to fuel our growth. We are innovating at a pace that is opening new markets and new use cases for F5 while deepening our relationships with existing customers and expanding our footprint across their environments. Our Application Delivery and Security platform delivers and secures every app and every API across on-premises, cloud and edge environment, on a single unified architecture.

但我們不只是順勢而為，我們也在利用這些趨勢來驅動成長。我們以加速的步伐創新，為F5開啟新市場與新使用情境，同時深化與既有客戶的關係，並在其各種環境中擴大我們的覆蓋範圍。我們的應用交付與安全平台，在單一統一架構下，於內部部署、雲端與邊緣環境中交付並保護每一個應用與每一個API。

We eliminate the complexity of stitching together point products and give customers consistent performance, security and policy wherever their applications run. As traffic, data and AI workloads become more distributed, that control point becomes even more valuable. We are executing with discipline, investing with conviction and innovating with urgency, and we look forward to continuing that momentum into Q4.

我們消除拼接各種點狀產品的複雜性，讓客戶無論其應用程式在哪裡運行，都能獲得一致的效能、安全性與政策。隨著流量、資料與AI工作負載變得更分散，這個控制點的價值就更為凸顯。我們以紀律執行、以信念投資、以緊迫感創新，並期待在第四季延續這股動能。

Operator, please open the call to questions.

接線員，請開放提問。

(Operator Instructions) Tim Long, Barclays.

(接線員指示) Tim Long，Barclays。

Two questions, if I could. First, on the AI front, Francois, thanks for detailing the wins and the different solutions there. Just curious, I think last quarter, you gave a little benchmark number for revenues. Are you seeing and expecting the revenues for your AI-related direct-related business to grow in line with that customer base? Should we -- should we think that 50% growth is kind of a ballpark number there?

如果可以的話我有兩個問題。第一個，關於AI方面，Francois，謝謝你詳細說明那些勝利以及不同的解決方案。我只是好奇，我記得上季你給了一個營收的基準數字。你是否看到、並預期你們AI相關、直接相關業務的營收會隨著那個客戶基礎同步成長？我們是否——我們是否可以把50%的成長視為一個大致的參考數字？

And then the second one, maybe for Cooper on gross margin. It sounds like in the near term, the component costs are not as bad as expected, but yet you still seem to be building in a buffer for next year. So could you maybe talk about maybe near-term gross margin related easing and maybe potential for pressure? I guess, it's still a pretty dynamic environment.

第二個問題，可能給Cooper，關於毛利率。聽起來短期內零組件成本沒有原先預期那麼糟，但你們似乎仍在為明年預留緩衝。所以你能否談談短期毛利率可能的緩解，以及潛在的壓力？我想，環境仍然相當動態。

Thank you, Tim. I'll start on your -- on the first part of your question. On AI, wouldn't necessarily approximate the number -- the growth in the number of customers with the growth of revenue. We did give an indicator of that last quarter that we have passed the $50 million revenue mark. That particular number, we don't intend to update every quarter.

謝謝你，Tim。我先回答你問題的第一部分。在AI方面，不一定能把客戶數量的成長，直接近似為營收的成長。我們上季確實提供了一個指標：我們已經跨過5,000萬美元的營收門檻。這個特定數字，我們不打算每一季都更新。

We may come back to that at the end of the year.

我們可能會在年底再回到這個話題。

But I can tell you that the momentum in direct AI use cases is very strong. We've seen traffic generally increase with our customers. But specifically, AI data delivery continues to see strong momentum and a gain in customers. We also had a very strong quarter in AI Security. The number of AI Security customers actually grew 100% just in the quarter, driven by a number of customers wanting best-in-class security in front of their AI model and their AI applications.

但我可以告訴你，直接AI使用情境的動能非常強。我們看到客戶的整體流量普遍增加。但特別是，AI資料交付持續展現強勁動能，客戶數也在增加。我們在AI安全方面也有非常強勁的一季。AI安全客戶數在單一季度內實際上成長了100%，主要是因為許多客戶希望在其AI模型與AI應用之前部署同級最佳的安全防護。

So we're very excited about the momentum that we're seeing there. In addition, Tim, I would just remind you as well that our AI opportunity isn't just the direct AI use cases, there is an indirect opportunity that manifests itself in the amount of traffic that we're seeing with our customers on our ADC systems.

因此，我們對目前看到的動能感到非常振奮。此外，Tim，我也想提醒你，我們的AI機會不僅限於直接AI使用情境，還有一個間接機會，體現在我們於客戶ADC系統上所看到的流量規模。

And where you see that is the amount of expansion we're getting on ADC systems. The growth that we have seen this year in Hardware this quarter, I think Hardware growth was 32% year-on-year. But if you look at even the first 3 quarters of the year, the Hardware growth is above 30%. And some of that expansion is because our customers are now scaling inference for AI in their enterprise environment, and we are in front of that traffic. So two categories of opportunities, and we're seeing strong momentum with both today.

你可以從ADC系統的擴充量看出來。我們今年在硬體方面的成長——本季我記得硬體年增為32%。但如果你看今年前三季，硬體成長也超過30%。其中一部分擴充，是因為客戶現在在其企業環境中擴大AI推論規模，而我們正位於那股流量之前端。所以有兩類機會，而我們目前在兩者上都看到強勁動能。

Yes. And Tim, thanks on the GM question. So yes, we saw the favorable result in Q3 and a little bit better outlook in Q4, and it was really tied to two dynamics. The first is that we're seeing a stronger adoption of some of our more high-performance appliances that carry a better gross margin profile. And then we're sourcing some of these components around memory and SSDs from vendors as part of our strategy to get in front of any potential supply chain risk on these components.

是的。Tim，謝謝你關於毛利率的問題。沒錯，我們在第三季看到較有利的結果，第四季展望也稍微更好，主要與兩個動態因素相關。第一，我們看到部分更高效能的設備採用率更高，而這些產品的毛利率結構更佳。第二，作為我們因應這些零組件潛在供應鏈風險的策略之一，我們正從供應商端採購部分記憶體與SSD等零組件。

And we had just a better outcome in terms of delivery from some vendors that carry a little bit lower cost profile on those components, for Q3 and Q4, and that's really what drove the improved outlook. So we're not updating the guidance at this time for next year, just given how dynamic the environment is, but that's something we'll continue to monitor.

而在第三季與第四季，我們在供貨交付方面取得了更好的結果，部分供應商在這些零組件上也具備較低的成本結構，這就是改善展望的主要原因。因此，鑑於環境仍相當動態，我們目前不更新明年的指引，但這是我們會持續監控的事項。

Joseph Cardoso, JPMorgan.

Joseph Cardoso，JPMorgan。

Maybe for my first one, can you just touch on just maybe going back to the mix impact on the high-performing systems that you're seeing here. What are you seeing that's driving customers to kind of move mix up into these higher performing systems? And then the second aspect there is more how sustainable that mix up is into the high-performance systems? And then I have a follow-up.

我的第一個問題，能否談一下你們在這裡看到的高效能系統組合（mix）影響。是什麼在驅動客戶把產品組合往這些更高效能的系統移動？第二個面向是，這種往高效能系統的組合上移有多可持續？然後我還有一個追問。

Yes. I think it's a lot of -- the dynamics that we've been talking about over the last several quarters, and we talked about our Hardware business and how much of the growth is coming from expansion use cases and digital sovereignty and then the big one is really the traffic we're seeing from AI use cases. And I think that it's really driving a need for customers to quickly provision higher performance units in support of that expansion initiative. And those higher-performing units have higher gross margins, but that's a trend that we think will continue. I think it's still relatively early -- it's one of the dynamics that's really been driving the overall growth we've been seeing in our Systems business.

好的。我認為這與我們過去幾季一直在談的許多動態因素有關；我們談到硬體業務，以及成長有多少來自擴充使用情境與數位主權，而最大的一項其實是我們從AI使用情境看到的流量。我認為這確實驅動客戶需要快速部署更高效能的設備，以支援那項擴充計畫。而那些更高效能的設備具有更高的毛利率，但我們認為這個趨勢會持續。我認為目前仍相對早期——這是推動我們系統業務整體成長的其中一個關鍵動態。

Got it. And then maybe just quickly on an update on how you're thinking about the trends between Software and System revenues and like more specifically there, as we think about software accelerating, obviously, still doing high single digits versus the long-term guidance more into that double-digit type of framework. I guess how are you guys thinking about potentially seeing an inflection upcoming? Is this still more in fiscal '27, better visibility there? Any kind of comments that you can provide around the trajectory of Software here.

了解。接著，想請你們快速更新一下你們如何看待軟體與系統營收之間的趨勢；更具體地說，當我們思考軟體加速時，顯然目前仍是高個位數成長，相較於長期指引更偏向雙位數的框架。我想問你們如何看待可能出現的拐點？這是否仍然更可能在2027財年，屆時能見度更高？你們能否就軟體的成長軌跡提供一些評論。

Yes. I'll provide a couple of thoughts. So first off, just right now, our pipeline looks really good. So we feel good about how we're positioned for the remainder of this year in terms of Software. We're seeing strong trends on our Distributed Cloud Business as well.

是的。我提供幾點看法。首先，就目前而言，我們的銷售管線看起來非常好。因此，就軟體而言，我們對今年剩餘期間的布局感到很有信心。我們也看到分散式雲端業務呈現強勁趨勢。

And we've talked about in the past that our SaaS and Managed Service business has been a little bit of a headwind to growth in prior years because we were undertaking a transition on legacy offerings, and we're effectively through that transition now.

我們過去也談過，我們的 SaaS 與託管服務業務在前幾年對成長造成了一些逆風，因為我們正在對既有（legacy）產品進行轉型，而我們現在基本上已經完成了這項轉型。

And so we anticipate that, that is going to be a growth driver to our Software business next year, along with a much stronger renewal cohort, on these multiyear agreements that we've been booking over time. And so we've got a really good base as we head into FY27. We've got pretty good visibility as to the utilization that we're seeing from customers across those Software opportunities and that gives us a good window to the expansion that we would expect to see for the year. So it all adds up to what we think is a pretty compelling growth opportunity, getting back to a solid double-digit growth outlook for next year.

因此我們預期，明年這將成為我們軟體業務的成長驅動力之一，同時，隨著我們一路以來所簽訂的多年期合約進入續約，我們也將迎來更強勁的續約群組。因此，邁入 FY27 時我們有非常好的基礎。我們對客戶在這些軟體機會上的使用率有相當不錯的能見度，這也讓我們能更清楚看到我們預期今年將出現的擴張。綜合來看，我們認為這是一個相當具吸引力的成長機會，明年有望回到穩健的兩位數成長展望。

George Notter, Wolfe Research.

George Notter，Wolfe Research。

I guess I wanted to ask about progress with the rSeries refresh. I'm just curious like what your updated comments are and the pacing of the refresh. Obviously, the end of the support is the end of this year. I'm just curious about how much installed base is left, how you see that kind of playing out in terms of the migration of iSeries to iSeries. And then also the impact of Mythos and Frontier AI models, I assume that's going to help kind of pull forward that refresh, but I'm wondering if you guys are seeing it?

我想問一下 rSeries 更新換代（refresh）的進展。我很好奇你們最新的看法以及更新的節奏。顯然，軟體支援在今年底就要結束。我想了解還剩下多少已安裝基礎（installed base），以及你們認為 iSeries 到 iSeries 的遷移會如何發展。另外還有 Mythos 與 Frontier AI 模型的影響；我猜這會有助於把更新換代的需求提前，但我想知道你們是否也觀察到這點？

Yes, I'll start, and then I'll let Francois follow up on the Mythos effect. So yes, it's going really well. When we talk about refresh, we've talked a lot about the dynamic we've been seeing with this refresh that we've likened it to a refresh plus because of the expansion that we've been seeing at the time of refresh.

是的，我先回答，然後再請 Francois 補充 Mythos 的影響。是的，進展非常順利。我們談到更新換代時，也多次提到我們看到的動態，我們把它形容為「refresh plus」，因為在更新換代的同時，我們也看到規模擴張。

But just in terms of refreshing the existing iSeries base, I would say it's been very orderly and in line with our expectations. It's been very consistent quarter-to-quarter. And so as you noted, there's an end of software support date in our Q2 of '27, which we think will be a bit of a driver for acceleration of that activity over the next few quarters.

但就更新既有 iSeries 基礎而言，我會說整體非常有秩序，且符合我們的預期。各季之間表現非常一致。因此如你所提到的，我們在 27 年的 Q2 有一個軟體支援終止日期，我們認為這將在未來幾季成為推動該活動加速的一個因素。

But the thing that's been more interesting to us is that rate of expansion we're seeing at that time of refresh has continued to improve. And then, of course, the business we're seeing outside of refresh has also been very strong. So we're seeing really good growth in terms of expansion both at the time of refresh and capacity expansion outside of the refresh. And then we're seeing new projects around data center modernization, digital sovereignty and then, again, performance associated with AI-driven workloads.

但對我們來說更有意思的是，我們在更新換代當下所看到的擴張速度持續改善。當然，我們在非更新換代情境下看到的業務也非常強勁。因此，無論是在更新換代當下的擴張，或是在更新換代之外的容量擴張，我們都看到非常好的成長。此外，我們也看到圍繞資料中心現代化、數位主權，以及再次提到的、與 AI 驅動工作負載相關的效能需求所帶來的新專案。

And then, George, on the Mythos effect, I think what we're seeing that with our customers in a couple of ways. The first is we're seeing a number of customers really focus on addressing technology debt and that includes avoiding to have aged estates. And that's actually a tailwind to the Refresh motion because some customers that may have had aged estates and may have been dragging their feet to refresh those estates. We're actually taking that very seriously.

然後 George，關於 Mythos 的影響，我認為我們從客戶端看到幾種面向。第一是，我們看到不少客戶非常聚焦於處理技術債，這也包括避免維持老舊的既有環境（aged estates）。這其實對更新換代動能是順風，因為有些客戶原本擁有老舊環境，可能一直拖延不願更新。現在他們確實非常認真看待這件事。

We're also seeing customers that are focusing a lot on remediating vulnerabilities that may exist in their environment and in some cases, stop development of new applications and new features to go and address vulnerabilities. But the other effect for F5 is what's happening in runtime security, and so what customers realize now is that if you don't have the time, if you don't have the same window that you used to have to be able to patch your applications, having the path to the applications be secure is incredibly important. And that's precisely where F5 is positioned in runtime security in front of application. And we're seeing that we had a very strong security quarter, just to give you a manifestation of that, we just launched a quarter ago, our AI-powered Web Application Firewall on Distributed Cloud. And the uptake of that has been extraordinary.

我們也看到客戶非常著重於修補其環境中可能存在的弱點，在某些情況下，甚至會停止開發新應用與新功能，轉而先去處理弱點。但對 F5 而言，另一個影響是執行期（runtime）安全的發展：客戶現在意識到，如果你沒有時間、也沒有像過去那樣的窗口期可以為應用程式打補丁，那麼讓通往應用程式的路徑保持安全就變得極其重要。而這正是 F5 在應用程式前端的執行期安全領域所處的位置。我們這一季的安全業務表現非常強勁；具體來說，我們在一季前剛推出分散式雲端上的 AI 驅動 Web 應用程式防火牆（WAF）。其採用情況非常驚人。

It's been the fastest-growing uptake of any product we've launched, really ever with hundreds of customers already running on this product in the first few months of the technology. And the vast majority of them actually using the product in blocking mode, to stop potential AI-powered attacks. And this product has already stopped potential zero-day attacks that would have come on to these customers. So we see a lot of energy with AI-powered Security, and that's why we're continuing to make these investments in WAF, API Security to drive that.

這是我們推出過的任何產品中成長最快的採用速度，真的可以說是史上最快；在技術推出的前幾個月，就已經有數百家客戶在使用這項產品。而其中絕大多數實際上是以阻擋模式（blocking mode）使用，用來阻止潛在的 AI 驅動攻擊。這項產品已經阻擋了可能針對這些客戶發動的零日（zero-day）攻擊。因此我們看到 AI 驅動安全有很高的動能，這也是為什麼我們持續在 WAF、API 安全上加大投資以推動成長。

Even beyond just AI-powered security, we're also seeing traction with securing AI itself. And I mentioned the number of AI security customers for F5 have grown 100% this quarter, largely because there is also a realization in the market that best-in-class security is also needed for AI models. You've probably followed the last week's saga of OpenAI and HuggingFace. And that just remind all of us of the importance of having best-in-class security for AI models because the security that's provided by the Frontier model providers is inadequate

甚至不僅是 AI 驅動的安全，我們也看到「保護 AI 本身」的進展。我提到過，本季 F5 的 AI 安全客戶數成長了 100%，主要是因為市場也意識到，AI 模型同樣需要同級最佳（best-in-class）的安全防護。你們可能也關注到上週 OpenAI 與 HuggingFace 的事件。那再次提醒我們，AI 模型需要同級最佳的安全防護，因為 Frontier 模型供應商所提供的安全性是不足的

And we have invested in best-in-class security. And we're seeing that traction now. I think I give the example in the script, but one of the major sports league in the US just recently put in production AI-enabled application that have sensitive data around player health or team strategies and needed to have best-in-class security for these AI model and chose F5 to provide that.

而我們已投資於同級最佳的安全能力。我們現在也看到這股動能。我想我在講稿中有舉例，但美國某個主要職業運動聯盟最近剛把一個 AI 賦能的應用上線到正式環境，該應用包含球員健康或球隊策略等敏感資料，需要為這些 AI 模型提供同級最佳的安全防護，因此選擇由 F5 來提供。

So the effects of Mythos are actually wide ranging. We're seeing that in refresh for customers that want to kind of remediate tech debt. We're seeing that in more energy in AI-powered security in front of applications. And we're also seeing it in terms of the momentum we're seeing in AI security now to secure AI models.

因此，Mythos 的影響其實相當廣泛。我們在更新換代上看到它的影響，因為客戶想要修補技術債。我們也在應用程式前端的 AI 驅動安全上看到更高的投入與動能。同時，我們也在 AI 安全（用於保護 AI 模型）方面看到目前的成長動能。

Matt Hedberg, RBC Capital Markets.

Matt Hedberg，RBC Capital Markets。

Congrats on the quarter. Great to see. My first question is kind of a follow-up to a couple of prior questions. Obviously, you're seeing strong Refresh activity and Cooper. I think in a prior question, you talked about in strong capacity expansion outside of the Refresh and obviously we'll have a new business opportunities here in AI.

恭喜本季表現。很高興看到這樣的成果。我的第一個問題算是延續前面幾個問題。顯然你們看到強勁的更新換代活動，Cooper。我記得在先前的提問中，你談到在更新換代之外也有強勁的容量擴張，而且顯然在 AI 方面也會有新的商機。

I guess, like philosophically speaking, growth from this cycle will peak at some point. But how should we think about growth post this cycle? Do you think it will be a shallower kind of trough than prior -- post prior cycles?

我想從更宏觀的角度來說，這一輪週期帶來的成長終究會在某個時間點見頂。但我們應該如何看待這個週期之後的成長？你認為相較於過去週期之後，這次的谷底會比較淺嗎？

Yes. I would say that is our belief, for a couple of reasons. One is just that in prior cycles, we didn't really have this kind of contribution to sales coming from new projects, whether that was Digital Sovereignty or AI-driven performance needs. The customer takeout opportunity -- or sorry, competitive takeout opportunity has been stronger this period as well compared to where we were at the last time we had a product cycle. And so those are all opportunities for us to continue to drive healthy growth.

是的。我會說這是我們的看法，原因有幾點。其一是在先前的週期中，我們並沒有看到來自新專案對銷售有這種程度的貢獻，不論是數位主權（Digital Sovereignty）或是由 AI 驅動的效能需求。本期客戶汰換競品的機會——抱歉，是競品替換（competitive takeout）機會——也比我們上一次產品週期時更強。因此，這些都是讓我們能持續推動健康成長的機會。

And then we think that it's kind of a bit of a wildcard, but post-quantum crypto is on the minds of a lot of our customers. And that potentially could drive an earlier adoption of the next product cycle. It's really too early for us to try to handicap that. But I think that those are all dynamics that give us confidence that we're going to continue to see healthy growth from this business over the next several years.

另外我們也認為這有點像是不確定因素，但後量子密碼（post-quantum crypto）正是許多客戶關注的議題。而這可能會推動下一個產品週期更早被採用。現在要我們去評估其影響還為時過早。但我認為，這些動態都讓我們有信心，未來幾年這項業務將持續呈現健康成長。

That's great to hear. And then, I've been really fascinated with sort of your AI factory success and being part of the NVIDIA reference architecture certainly seems to be helping you. Francois, in your prepared remarks, you talked about token cost. I guess -- I'm wondering could you talk a little bit about the ROI and maybe how it's added to growth when you're seeing customers see higher GPU utilization and optimize token spend?

很高興聽到。另外，我一直對你們的 AI 工廠（AI factory）成功案例很著迷，而成為 NVIDIA 參考架構的一部分顯然也在幫助你們。Francois，你在事先準備的發言中提到了 token 成本。我想問——你能否談談投資報酬率（ROI），以及當客戶看到更高的 GPU 使用率並最佳化 token 支出時，這如何帶動成長？

Yes, Matt. So we I think as I shared before, we -- the -- our Traffic Management Software when paired with NVIDIA GPU really on a number of tests has shown the ability to generate 40% -- anywhere between 30% to 50% more tokens for a given GPU infrastructure. we think that value proposition is compelling. And here's what we're seeing in the early market work that our teams have done with customers. Where we are actually interestingly, where we are seeing traction now is more with Sovereign AI customers.

是的，Matt。我想如同我之前分享的——我們的流量管理軟體（Traffic Management Software）在搭配 NVIDIA GPU 時，在多項測試中顯示，對於既定的 GPU 基礎設施，能產生多 40%——大約 30% 到 50% 更多的 tokens。我們認為這個價值主張很有吸引力。以下是我們團隊與客戶進行早期市場驗證時所看到的情況。有趣的是，我們目前看到的動能更多來自主權 AI（Sovereign AI）客戶。

This is typically telcos outside of the US that are implementing AI factories and are starting to put these AI factories in production, and they are quite cost conscious and really want to get the most amount of tokens for the infrastructure that they're building.

這通常是美國以外的電信業者（telcos），他們正在建置 AI 工廠並開始將這些 AI 工廠投入生產環境；他們相當重視成本，並且真的希望在所建置的基礎設施上取得最多的 tokens。

And so we think that will continue. That said, those sovereign AI factories are generally relative to the neo cloud and neo scalers, they're not the largest builds in infrastructure. With the neo scalers, what we're seeing today is most of them -- their business model is GPU-as-a-service and offering GPU as a service typically to hyperscalers. And so the neo scalers themselves are not concerned with how many tokens they're generating for GPU given that they're renting on a GPU unit. We think that is going to change.

因此我們認為這種趨勢會持續。不過，這些主權 AI 工廠相較於新型雲（neo cloud）與新型擴展商（neo scalers），其基礎設施建置規模通常不是最大。至於新型擴展商，我們目前看到他們多數——其商業模式是 GPU-as-a-service，也就是以服務形式提供 GPU，通常提供給超大規模雲端業者（hyperscalers）。因此，新型擴展商本身並不在意在既定 GPU 下能產生多少 tokens，因為他們是以 GPU 單位出租。我們認為這將會改變。

We think the market is going to evolve and that the economics of tokens are going to become important for neo scalers as they go into the enterprise market, or even as they sell more and more to AI native companies. So we think that market is going to evolve over time and neo scalers will have -- will care more about the tokenomics, sorry, which is really the value add that F5 provides.

我們認為市場會演進，當新型擴展商進入企業市場，或甚至當他們愈來愈多地銷售給 AI 原生公司（AI native companies）時，token 的經濟性將變得重要。所以我們認為市場會隨時間演變，而新型擴展商將會——會更在意 tokenomics（抱歉），也就是 F5 所提供的加值所在。

Simon Leopold, Raymond James.

Simon Leopold，Raymond James。

I guess I'm trying to understand why Systems are continuing to outgrow software. And I guess a couple of theories I wanted to throw out there. One is just a total cost of ownership for Systems maybe being less costly than the TCO on the software solution or whether it's a possibility that your customers are having trouble getting the servers to run software on. We've heard that from others or if there are other reasons for this growth difference?

我想了解為什麼系統（Systems）仍持續成長得比軟體更快。我有幾個假設想提出來。其一是系統的總持有成本（TCO）可能比軟體解決方案的 TCO 更低；或是你們的客戶可能難以取得可用來跑軟體的伺服器。我們也聽其他人提過這點；或者還有其他原因造成這種成長差異？

No, it's a really good question. Look, in terms of software, the growth that we're seeing this year is actually very much in line with our expectations, the expectations we have set at the beginning of the year, I think we had shared the renewal cohort coming from the 2023 year was not a very strong renewal cohort.

不，這是個非常好的問題。你看，就軟體而言，我們今年看到的成長其實非常符合我們的預期，也就是我們在年初所設定的預期；我想我們曾分享過，來自 2023 年的續約族群（renewal cohort）並不是很強。

And so we expected growth this year that was in the mid-single digits. And I think we will do that maybe a little better. But that's been the expectation. And conversely, as Cooper shared earlier, we expect Software growth to reaccelerate next year in part because of a much stronger renewal cohort from 2024.

因此我們預期今年的成長會是中個位數（mid-single digits）。我想我們會達到，甚至可能略好一些。但這一直是我們的預期。相對地，如同 Cooper 先前分享的，我們預期明年軟體成長會再加速，部分原因是 2024 年的續約族群要強得多。

Now as it relates to hardware, there are a number of dynamics that are playing out that are really causing our customers to want to put more Hardware in. One, of course, is we are in a refresh cycle. And so of course, that provides strength to Hardware bookings. But there are other structural shifts we're seeing that are driving customers to want higher capacity hardware, higher throughput for higher performance. And those are -- those dynamics include some -- basically Sovereignty is a big part of it.

至於硬體方面，有多項動態正在發生，確實促使客戶希望部署更多硬體。當然，其中之一是我們正處於汰換更新週期（refresh cycle）。因此這自然會強化硬體訂單（Hardware bookings）。但我們也看到其他結構性轉變，驅動客戶想要更高容量的硬體、更高吞吐量以取得更高效能。而這些——這些動態包含一些——基本上主權（Sovereignty）是其中很大的一部分。

So we're seeing the customers -- we started to see this trend in Europe a few quarters ago, but now it's expanded to the Middle East. It's expanded to Asia, where customers are repatriating workloads from global hyperscalers into local alternatives.

因此我們看到客戶——我們在幾季前先在歐洲看到這個趨勢，但現在已擴展到中東。也擴展到亞洲，客戶正把工作負載從全球超大規模雲端業者（global hyperscalers）遷回到本地替代方案。

In fact, there is a forecast out by one of the analysts, I think, this quarter that there's an expectation now that 75% of enterprises internationally will repatriate workloads into local alternatives inside the next 3 to 5 years. So we're seeing that phenomenon. And sovereignty also implies often building on their own infrastructure stack in their private data centers.

事實上，有一位分析師在本季提出一份預測，我想是本季，現在預期國際企業中有 75% 將在未來 3 到 5 年內把工作負載遷回本地替代方案。所以我們正在看到這個現象。而主權也往往意味著在其私有資料中心內，建立自有的基礎設施堆疊（infrastructure stack）。

There's also a dynamic around competitive displacement. We continue to win share from our competitors in Hardware because we have placed bets on hardware innovation when they did not. And today, we find out also with hardware that has a better throughput, higher performance, higher capacity less power consumption, and those benefits are translating into customer wins.

另外還有一個與競品替換（competitive displacement）相關的動態。我們持續在硬體上從競爭對手手中贏得市占，因為當他們沒有下注時，我們在硬體創新上做了投入。而今天我們也發現，我們的硬體具備更好的吞吐量、更高效能、更高容量且更低耗電，這些優勢正在轉化為客戶的勝出。

And then there's another dynamic around Hardware, which is consolidation when customers have used multiple point solutions and want to consolidate on F5. So all of these dynamics are really driving the growth that we're seeing on Hardware. And many of these elements are structural in nature, and we think will continue, which is the reason our Investor Day a month ago, we changed our long-term guidance on Hardware, which had been in the mid-single-digit decline, and we confirm that we expect that Hardware to be growing for the next few years.

接著還有另一個與硬體相關的動態，就是整併：當客戶過去使用多個點狀解決方案（point solutions），現在希望整併到 F5 上。因此，所有這些動態都在推動我們在硬體上看到的成長。而其中許多因素本質上是結構性的，我們認為會持續下去；這也是為什麼在一個月前的投資人日（Investor Day），我們調整了硬體的長期指引：原先預期是中個位數下滑（mid-single-digit decline），現在我們確認，未來幾年硬體將會成長。

James Fish, Piper Sandler.

James Fish，Piper Sandler。

Nice quarter. Maybe just specifically, what are you seeing on the inbound interest in pipeline build, specifically on AI Guardrails as of late? And are you seeing any customers specifically budget for AI Agent Security as sort of this new line item? Or is it kind of being bundled into that broader application security platform. I just have a follow-up on the billings.

本季表現不錯。我想更具體問一下，最近在 AI Guardrails 方面，你們看到的主動詢問（inbound interest）與管線建立（pipeline build）情況如何？另外，你們是否看到有客戶針對 AI 代理安全（AI Agent Security）特別編列預算，把它當作一個新的預算科目？還是它多半被打包在更廣泛的應用程式安全平台裡？我還有一個關於帳單（billings）的追問。

It's interesting. It varies from customer to customer. What we're seeing right now is in financial services, financial services and then other organizations that are quite advanced or sophisticated in their use of AI or their care for safety. We are seeing that they are budgeting specifically for securing AI. And that includes securing AI models, securing agents, governance around AI, discovery of what their employees are using for AI.

這很有意思。因客戶而異。我們目前看到的是，在金融服務業，以及其他在 AI 使用上相當先進或成熟、或對安全性相當重視的組織。我們看到他們正在專門編列預算用於 AI 安全防護。這包括保護 AI 模型、保護代理（agents）、AI 治理，以及盤點員工正在使用哪些 AI。

It's a number of functions, and we have invested quite rapidly in this area in -- around testing and penetration of AI models securing AI models in production. We made an acquisition this quarter of a company called SurePath AI that complemented our capabilities with Discovery. And we launched our AI Security platform because we're seeing customers increasingly wanting more and more of these point solutions combined into a single platform.

這涵蓋多項功能，而我們也在這個領域投入得相當快——包括 AI 模型的測試與滲透測試，以及在生產環境中保護 AI 模型。我們本季收購了一家名為 SurePath AI 的公司，透過 Discovery（探索/盤點）能力補強了我們的能力。我們也推出了 AI Security 平台，因為我們看到客戶愈來愈希望把更多這類點狀解決方案整合到單一平台中。

And we think that's where the market is headed, just like the WAF market consolidated a few years ago with WAF, DDoS, API and bot defense in a single platform, we see the same consolidation of functionality happening with AI Security platforms. So that's where we're seeing very strong interest. And I shared earlier that the number of customers in that area doubled this quarter, and we expect that momentum to continue to grow with the investments we're making in this area.

我們認為市場正朝這個方向發展；就像幾年前 WAF 市場整併，把 WAF、DDoS、API 與機器人防護整合到單一平台一樣，我們也看到 AI Security 平台的功能正在發生同樣的整併。因此我們在這方面看到非常強烈的興趣。我先前也分享過，該領域的客戶數在本季翻倍，我們預期在我們持續投入之下，這股動能會繼續成長。

And then Cooper, just for you. Obviously, good top line quarter, but actually, if I look at billings here, another exceptional quarter again. Are you seeing Product backlog build and that explains some of this? Or is it more so the Services lags here just has yet to flow into kind of the revenue line?

接著 Cooper，特別問你。顯然本季營收表現不錯，但如果我看 billings（帳單/開立金額），又是另一個非常出色的季度。你們是否看到產品積壓（backlog）在增加，因而解釋了其中一部分？還是說主要是服務（Services）端的落後，尚未反映到營收線上？

Yes, thanks. So no, it doesn't have anything to do with backlog because we wouldn't have billings yet -- we don't bill until we ship product. It's more to do with, as you said, the lag factor on the Service revenue and then we had a strong bookings quarter with new FCPs, and so there's a portion of the FCP booking that gets deferred. So you get 2/3 or so gets recognized upfront and the remainder gets referred. And so that drove some strength in the deferred revenue on the deferred subscription revenue, which fed into that billings improvement.

是的，謝謝。所以不是，這與 backlog 無關，因為我們還不會有 billings——我們要到產品出貨才會開立帳單。這更多是你提到的服務營收認列的時間落差，另外本季新 FCP 的 bookings 很強，因此 FCP 訂單中有一部分會遞延。所以大約有 2/3 會先行認列，其餘部分會遞延。因此帶動了遞延收入、特別是遞延訂閱收入的增長，進而推升了 billings 的改善。

Meta Marshall, Morgan Stanley.

Meta Marshall，摩根士丹利。

Great. Maybe first for Francois. Just in the three AI use cases, data delivery, runtime security and load balancing, noted the commentary on the growth in customers on the AI security side. But just are there any developments in terms of kind of deal sizes, or just how you would size any of those opportunities versus a particularly as runtime security and load balancing mature?

好的。先問 Francois。在三個 AI 使用案例——資料傳遞（data delivery）、執行期安全（runtime security）與負載平衡（load balancing）——你提到 AI 安全端客戶成長的評論。但想請問在交易規模方面是否有任何進展？或者你會如何衡量這些機會的規模，特別是當執行期安全與負載平衡逐漸成熟之後？

And then maybe just a follow-up question for Cooper, just if there's any kind of contribution to the quarter, we should be mindful of from pricing? And if you could just outline that.

然後可能再追問 Cooper，本季是否有任何來自定價（pricing）的貢獻需要我們留意？如果可以的話，請你概述一下。

Thank you, Meta. Look, the deal sizes are -- it's interesting because there's a wide divergence. They could be as small as hundreds of case to be as large as we've done some 8-figure deals in this in this area of AI. I think in security, as we move more towards platform over time and we consolidate these functionalities, deal sizes will obviously grow. In AI Data Delivery, we are already seeing substantial deals because it's all about -- AI Data Delivery, it's all about getting data faster to models in training or for inference.

謝謝你，Meta。交易規模——很有意思，因為差異非常大。小的可能只有數十萬美元，大的話，我們在 AI 這個領域也做過一些達到八位數金額的交易。我認為在安全領域，隨著時間推進、我們更朝平台化發展並整合這些功能，交易規模顯然會變大。在 AI 資料傳遞方面，我們已經看到相當可觀的交易，因為——AI 資料傳遞的核心就是讓資料更快送到模型，用於訓練或推論。

So customers are highly sensitive to latency, want very high throughput. And generally, that means they want high-capacity hardware to perform that, and that leads to large deal sizes.

因此客戶對延遲非常敏感，並且希望有非常高的吞吐量。而一般而言，這代表他們需要高容量硬體來達成，進而帶來較大的交易規模。

In AI Factory Load Balancing, the deal sizes are initially smaller when you have factory build-outs of just a few hundred GPUs. I think as more of the -- we get into more of the larger AI factories, we should see greater deal size there. Overall, look, I think our AI revenues are growing quite rapidly, and we may be able to touch on this more in -- at the end of the year once we have a year's worth of data to share with you.

在 AI 工廠負載平衡方面，當 AI 工廠的建置規模只有幾百張 GPU 時，初期交易規模會比較小。我認為隨著我們進入更大型的 AI 工廠，我們應該會看到那裡的交易規模變得更大。整體來看，我們的 AI 相關營收成長相當快；等到年底我們累積有一整年的數據可與各位分享時，或許能再更深入談這點。

And then in terms of price realization, so there's -- we saw a modest contribution from price realization that was tied more to our discount governance. We've been pretty effective this year just continuing to improve our discounting over time. The adjustments that we've made to pass through some of the cost on components will be more of a contributor to our FY27 revenue, but nothing that really hit our revenue in Q3.

至於價格實現（price realization），我們看到有一個溫和的貢獻，主要與我們的折扣治理（discount governance）更相關。今年我們在持續改善折扣策略方面做得相當有效。至於我們為轉嫁部分零組件成本所做的調整，對 FY27 的營收貢獻會更明顯，但在 Q3 並沒有對營收造成什麼實質影響。

Tal Liani, Bank of America.

Tal Liani，美國銀行。

It's Tomer Zilberman on for Tal tonight. Just wanted to ask on a geographic basis, if we look at your Americas revenue, it accelerated from the last few quarters of low to mid-single digits. I think you said to 11% on the call. Just wanted to ask what the trends you were -- you saw in the geo -- was it more of a catch-up because of the last few quarters of underinvestment, or what did you see there?

今晚由 Tomer Zilberman 代 Tal 提問。想從地理區域的角度問一下，如果看你們美洲（Americas）營收，從過去幾季的低到中個位數成長加速。我想你在電話會議上說是到 11%。想請問你們在區域趨勢上看到什麼——這比較像是前幾季投資不足後的追趕（catch-up），還是你們看到其他因素？

Yes. I'll take that one, Tom. So really, we've seen good strength across all theaters from a bookings perspective. So it is more to do with the timing of shipments. And Francois noted that last quarter on the call because the growth rate was down a bit from Americas.

是的。Tomer，這題我來回答。其實從 bookings 的角度來看，我們在所有區域（theaters）都看到不錯的強勁表現。所以這更多與出貨時點（timing of shipments）有關。Francois 上季在電話會議也提到，因為美洲的成長率當時有些下滑。

And we said that the underlying demand is actually quite strong and then we thought that, that would play out over time as we were able to ship more of those orders, in this environment, there is a bit of a lag on shipping time from when we book the orders. And so we saw that come through in the reported revenue in Q3.

我們當時說，底層需求其實相當強勁，而我們認為隨著時間推進、當我們能出更多這些訂單時，這會逐步反映出來；在目前的環境下，從接單到出貨之間確實存在一些時間落差。因此我們在 Q3 的已報告營收中看到這點反映出來。

You'll also notice this quarter, APAC is down 11% on a revenue reported basis, that is also not indicative of the underlying demand strength that we're seeing in APAC. So I think just in this environment where there's a little more of a lag on the shipping timing, you're going to get more of that variability in the growth rates. But over a longer period of time, multiple quarters you start to see that growth rate start to normalize. Broad answer to your question, the underlying demand growth has been healthy across all of our theaters.

你也會注意到本季亞太（APAC）在已報告營收基礎上下降 11%，這同樣不代表我們在 APAC 看到的底層需求強度。所以我認為在這種出貨時點落差稍大的環境下，你會看到成長率有更大的波動。但拉長到更長時間、多個季度來看，你會開始看到成長率趨於正常化。總結回答你的問題：我們在所有區域看到的底層需求成長都相當健康。

Got it. And maybe as a follow-up and a similar question to touch on the subscription line. Last quarter, it was, I think, flat on a dollar basis sequentially, and you talked about a weak renewal base, which I think you also mentioned in the prepared remarks today. So if I look at '23 as a week renewal base, I think in aggregate is what you mentioned, what drove the stronger performance this quarter?

了解。再追問一個類似問題，想談訂閱（subscription）這條線。上季我記得按美元金額來看是季比持平，你們提到續約基期偏弱，我想你們今天在事先準備的講稿中也有提到。所以如果我看 2023 年作為較弱的續約基期——我想你提到的是整體而言——那本季表現更強的驅動因素是什麼？

Yes. So when we talked about the weak renewal base, we're largely talking about the full year. So we had flattish Software revenue in FY22 -- sorry, FY23 against FY22. And so as a result, the base that we had to renew for this year, was more modest than it typically would be. And it's going to be much stronger in our FY27. But from a sequential basis, you still get that variability. And so we did see some strength in our Subscription business this quarter tied to a stronger cohort and also good expansion in the quarter.

是的。所以當我們談到續約基期偏弱時，我們主要是在談整個年度。因此，我們在 FY22——抱歉，是 FY23 相較 FY22 的軟體營收大致持平。因此，今年需要續約的基期就比通常情況更為溫和。而在我們的 FY27，這個基期會強得多。但從逐季的角度來看，你仍然會看到那種波動性。因此，本季我們確實看到訂閱業務有一些強勁表現，與較強的同批客群（cohort）以及本季良好的擴張有關。

But overall, the growth rate has been a little bit more orderly in terms of the quarter-to-quarter variability in the growth rate this year than what we've seen in past years. Now next year, we think that the growth rates will be stronger in the second half of the year. So it's just one of these things where we're always tracking what that base is, it's coming up for renewal, what the expansion opportunities are and then trying to give some visibility if there's unusual patterns that we see. But for this year, it's been more muted the variability in the growth rates.

但整體而言，今年的成長率在逐季波動方面，比我們過去幾年看到的更為有序一些。而明年，我們認為下半年成長率會更強。所以這只是我們一直在追蹤：那個基期是什麼、哪些即將到期要續約、有哪些擴張機會，然後如果我們看到不尋常的模式，就嘗試提供一些能見度。但就今年而言，成長率的波動性比較收斂。

Michael Ng, Goldman Sachs.

Michael Ng，高盛。

I just had a few follow-ups on kind of implications for 2027. Maybe first on the Systems side, just given all the momentum that you have there in terms of refresh and the expanding use cases. Do you see upside to the prior outlook of mid- to high single digits for 2027? Just would love any general thoughts there? And I certainly hear you on the reiterated gross margin outlook for 2027.

我有幾個關於 2027 年影響的追問。也許先從系統（Systems）端來看，考量你們在更新換代（refresh）以及使用案例擴張方面的動能都很強。你們是否看到 2027 年先前「中到高個位數」展望有上行空間？很想聽聽你們的整體看法？另外，我也確實聽到你們重申 2027 年的毛利率展望。

Just would love some commentary on whether you are seeing some better operating leverage now but are kind of reserving some margin in the case that there's some outsized commodity cost inflation. Just trying to understand like what's kind of embedded in the reiterated 82% gross margin guide next year?

也很想請你們評論一下：你們現在是否看到更好的營運槓桿，但為了以防大宗商品成本通膨超預期，而先保留一些利潤空間。我想理解的是：明年重申的 82% 毛利率指引中，究竟內含了哪些假設？

Michael, thank you. Let me start, and then Cooper will address the gross margin question. On the revenue expectations for next year, as you know, we've given guidance of upper single-digit growth for '27 and actually beyond at our Investor Day about a month ago. We're not updating that guidance today. We -- generally, we feel good about the -- what we're seeing in the marketplace, the demand we're seeing, both for Hardware and our Software.

Michael，謝謝你。我先回答，接著 Cooper 會回應毛利率的問題。關於明年的營收預期，如你所知，我們在大約一個月前的投資人日已給出 '27 以及其後的高個位數成長指引。我們今天不更新該指引。整體而言，我們對市場上看到的情況、看到的需求（硬體與軟體）都感到樂觀。

But beyond the immediate demand, what we are most pleased about is the structural demand drivers that we're seeing. Our -- we continue to drive competitive wins we continue to drive consolidation on to F5. We have a number of customers that were using point solutions for cloud providers or other players in the space, and our whole investment into a platform strategy with our application delivery and security platform. We're seeing real momentum with the ADSP platform and the real effect of that in consolidation.

但除了短期需求之外，我們最滿意的是我們看到的結構性需求驅動因素。我們——持續拿下競爭性勝利，也持續推動整合到 F5 平台上。我們有不少客戶原本使用雲端供應商或其他同業的點狀解決方案，而我們在應用交付與安全平台（application delivery and security platform）的平台策略上投入了大量資源。我們看到 ADSP 平台有實質動能，也看到它在整合上的實際效果。

Just launched F5 Insight also this quarter, which really glues elements of our platform together. We're seeing very strong adoption of F5 Insights among our customers, which tells us that they're seeing significant value in our platform. And of course, we're seeing growth in AI use cases across the data delivery and security. And so you take all of these drivers, I think we feel we feel good about the guidance we've given for the next 3 years. And we'll talk specifically about 2027 on our October call.

本季我們也剛推出 F5 Insight，這能把我們平台的各個要素真正黏合在一起。我們看到客戶對 F5 Insights 的採用非常強勁，這告訴我們他們在我們的平台上看到了顯著價值。當然，我們也看到 AI 使用案例在資料交付與安全領域持續成長。綜合這些驅動因素，我想我們對未來 3 年所給出的指引感到有信心。我們會在 10 月的電話會議上更具體談 2027 年。

Yes. And then in terms of the gross margin, so we didn't reiterate our gross margin for next year. We just are not updating at this time. I know it's a little bit of a subtle nuance, but really, I think the way to think of it is we're encouraged with what we saw in Q3 and in Q4. But we are cognizant that things are fluid.

是的。至於毛利率方面，我們並沒有重申明年的毛利率。我們只是目前不做更新。我知道這是個有點微妙的差別，但我認為可以這樣理解：我們對 Q3 和 Q4 看到的情況感到鼓舞。但我們也意識到情勢仍在變動。

And transparently, we are getting mixed signals. So we've seen a lot of signs of some stability in terms of memory pricing this quarter. But then we're -- we also hear other signals that it still could be more variable. And so -- and then we're looking at the rest of our BOM as well and other components that we're sourcing and just as early as it is right now, we felt like it was too early to make an adjustment to our gross margin forecast for next year.

坦白說，我們收到的訊號是混合的。本季我們看到記憶體價格在穩定性方面有不少跡象。但同時——我們也聽到其他訊號顯示它仍可能更具波動。因此——我們也在看整體物料清單（BOM）的其他部分，以及我們採購的其他元件；以目前這麼早的時間點來看，我們覺得現在調整明年毛利率預測還太早。

But that said, I think our manufacturing team has done an outstanding job in terms of doing -- making advanced purchases on some of these more scarce components, diversifying the supply base. And so we feel like we've got a pretty good handle on the supply that we need for our revenue for next year. And we think we'll be in a position to give a better update on the gross margin guidance in October.

不過話說回來，我認為我們的製造團隊在一些較稀缺元件的提前採購、以及供應來源多元化方面做得非常出色。因此，我們覺得對於明年營收所需的供應量，我們掌握得相當不錯。我們也認為在 10 月時，我們會更有條件就毛利率指引提供更好的更新。

Jeffrey Hopson, Needham.

Jeffrey Hopson，Needham。

We're seeing with AI workloads, some of a recent push to moving things on-prem for security or sovereign reasons. Just curious in those more regulated industries that are on-prem kind of where are we in that adoption of AI and where are we with newer AI companies starting to consider on-prem deployments?

我們看到在 AI 工作負載方面，近期有一些推動把東西移回地端（on-prem）的趨勢，原因可能是安全或主權等考量。想請教在那些更受監管的產業中，AI 的採用目前進展到哪個階段？以及一些較新的 AI 公司開始考慮地端部署的情況又如何？

Well, I think -- well, in regulated industries, frankly, it varies from an industry to the next and from one customer to the next. I think in financial services, we're seeing rapid adoption of AI. I think we're also starting to see that in health care, and we're seeing our telco customers also regulated, but leveraging AI to significantly improve profitability in operations and the experience of their customers.

嗯，我想——在受監管產業中，坦白說，不同產業之間、以及不同客戶之間差異很大。我認為在金融服務業，我們看到 AI 正在快速採用。我認為我們也開始在醫療保健領域看到同樣情況；同時我們也看到電信客戶同樣受監管，但正在運用 AI 來大幅提升營運獲利能力與客戶體驗。

In terms of what stack are people using and what models people are using, what I think we're seeing already is a very strong signs that both enterprises, government agencies and telcos are all going to use a mix of models and they are going to use closed models, some frontier AI providers and open weight and/or open-source models. They are going to consume public AI from the public cloud, and they are going to build their own stack, their own infrastructure, on-premise to run their models. And they're going to want to basically find the best model for each task and ideally, the cheapest model for each task. And the multiplicity of models and types of infrastructure is exactly what plays to the strength of F5 because we have built the hybrid multi-cloud company precisely to serve and enable that choice for our customers.

至於大家使用什麼技術堆疊、使用什麼模型，我認為我們已經看到非常強烈的跡象：企業、政府機構與電信業者都會採用混合的模型組合；他們會使用封閉模型、一些前沿 AI 供應商的模型，以及開放權重和／或開源模型。他們會從公有雲消費公有 AI，同時也會自建自己的技術堆疊與基礎設施，在地端（on-premise）運行他們的模型。而他們基本上會想為每個任務找到最合適的模型，理想情況下，也要為每個任務找到成本最低的模型。模型與基礎設施型態的多樣性，正好發揮 F5 的強項，因為我們打造混合多雲公司（hybrid multi-cloud company）的初衷，就是為了服務並賦能客戶做出這樣的選擇。

Just in the same way that we predicted several years ago that our customers would end up with wanting to have the choice in infrastructure environments and deploy applications in multiple infrastructure environments. And we designed our company to meet that moment. We think the same thing is happening with AI where customers want the best model, the best environment for each task, and they're going to want to partner that can deliver these models and secure all these models across any environment and in any cloud. And that is the company that we've built. That is the company we're building for AI, and we are, I believe, meeting the moment.

就像我們幾年前預測的那樣：客戶最終會希望在基礎設施環境上擁有選擇權，並在多種基礎設施環境中部署應用程式。而我們也把公司設計成能迎接那個時刻。我們認為 AI 也正在發生同樣的事情：客戶希望針對每個任務選擇最佳模型、最佳環境；他們也會希望有合作夥伴能在任何環境、任何雲端中交付這些模型並保護所有這些模型。而這正是我們所打造的公司。這也是我們為 AI 正在打造的公司，而我相信我們正在把握這個時刻。

And we're very excited about the ways in which different customers are going to deploy AI and the way we're going to enable that choice for them.

我們對不同客戶將如何部署 AI，以及我們將如何讓他們能夠做出這種選擇，感到非常興奮。

Got it. And maybe just a follow-up on agentic workloads across apps. Is there any way to think about how they may be similar or different than traditional workloads and the resulting impacts on F5 then?

了解。另外也許追問一下，關於跨應用程式的代理式（agentic）工作負載。有沒有什麼方式可以思考它們與傳統工作負載有哪些相似或不同之處，以及因此對 F5 會帶來哪些影響？

So agents are fundamentally just regular applications with the exception that they are invoking models. They're invoking models for inference, and agents also have the ability to make subsequent calls inside of a data center or an environment. And so what we really see is agents are increasing data center traffic or hybrid multi-cloud traffic inside these organizational environments.

所以，代理本質上就是一般的應用程式，只是例外在於它們會呼叫模型。它們會呼叫模型進行推論，而代理也具備在資料中心或某個環境內進一步發出後續呼叫的能力。因此我們真正看到的是，代理正在增加這些組織環境內的資料中心流量或混合多雲流量。

And so for F5, what that fundamentally means is sitting in front of more of this traffic that's going to APIs, that's going through traditional applications, securing that traffic, but then also securing and delivering traffic to these AI models as well, whether they're frontier models or models that are fundamentally set up inside of a local data center or in a local environment for an organization.

因此對 F5 而言，這在根本上意味著我們會位於更多這類流量之前：這些流量會流向 API、會穿越傳統應用程式；我們要保護這些流量，同時也要保護並傳遞流量到這些 AI 模型，不論是前沿模型，或是基本上部署在組織本地資料中心或本地環境中的模型。

Thank you, Kunal. And Jeff, that was Kunal, our Chief Product Officer.

謝謝你，Kunal。Jeff，剛剛那位是我們的產品長 Kunal。

Amit Daryanani, Evercore.

Amit Daryanani，Evercore。

Francois, you folks saw obviously some very impressive growth on the system side and you've talked about multiple factors that are sort of helping folks out in that space right now. I think what investors are really going to struggle in trying to understand is, how many of these vectors are cyclical versus secular. And so just from your perspective, qualitatively or quantitatively, how do you think about this 32% growth and how much of this you think is cyclical versus secular going forward?

Francois，你們在系統端顯然看到非常亮眼的成長，也談到目前有多項因素正在支撐該領域的客戶。我認為投資人真正會苦於理解的是：這些動能中有多少是景氣循環性的（cyclical），又有多少是結構性的（secular）。因此從你的角度，不論定性或定量，你如何看待這 32% 的成長？往前看，你認為其中有多少是循環性、多少是結構性？

Well, clearly, I mean, we think a portion of it is cyclical. If we didn't think that we would be guiding 30% Hardware growth for the next 3 years. This is not what we -- as you know, that's not what we guided at Investor Day, we shared we expected Hardware to be in the single-digit growth to mid-single-digit growth going forward. So what's cyclical, obviously, is that we are in the middle of a refresh cycle with customers. And that, by definition, is cyclical.

嗯，很明顯，我們認為其中一部分是循環性的。如果我們不這麼認為，我們就會指引未來 3 年硬體成長 30%。但這不是我們——如你所知，這也不是我們在投資人日所給的指引；我們當時分享的是，預期硬體未來將是個位數成長到中個位數成長。所以循環性的部分，顯然是我們正處於客戶的汰換更新週期中。而這依定義就是循環性的。

But there are a number of more secular shifts that have layered on to that and that are driving hardware growth in where you're seeing it. And those secular driver as we've talked about. It's the move now outside Europe, Middle East and Asia for sovereignty that creates reinvestment in data centers. If there was a phase where customers were thinking about cloud first, anytime they were thinking about new applications.

但在此之上也疊加了多項更偏結構性的轉變，正推動你所看到的硬體成長。而這些結構性驅動因素如我們所談過的。也就是現在在歐洲、中東與亞洲以外地區，主權（sovereignty）的趨勢帶來資料中心再投資。如果過去有一段時間客戶在思考新應用時採取「雲端優先」。

Now we're in a phase where customers think about sovereignty first, and it means that they're reinvesting in data centers or building resilience between their cloud infrastructure and their data center infrastructure. That's a driver. A second driver, which is difficult to measure today, but I think we'll get more refined about it in the future and get -- we'll get more intelligent about it in the future is the amount of AI traffic that is now flowing on on-premise infrastructure and how much that is driving customers to either need more units from F5 or higher capacity units from F5. But that is clearly also a driver.

現在我們進入一個客戶「主權優先」的階段，這意味著他們正在對資料中心再投資，或是在雲端基礎設施與資料中心基礎設施之間建立韌性。這是一個驅動因素。第二個驅動因素，目前較難衡量，但我認為未來我們會更精準、也會——未來我們會更有洞見的是：現在有多少 AI 流量正在流經內部部署（on-premise）基礎設施，以及這在多大程度上驅動客戶需要向 F5 購買更多台設備或更高容量的設備。但這顯然也是一個驅動因素。

And then there are things that are specific to our competitive position, specifically what we're doing with platform adoption, what we're doing with competitive takeouts that are also exacerbating this trend. And so all of that is it's not trivial to completely separate and quantify for you what's cyclical, what's -- these other drivers because sometimes these other drivers also manifest themselves in a refreshed motion, meaning we -- our teams are speaking with customers about a refresh.

另外還有一些與我們競爭地位相關的因素，特別是我們在平台採用方面的作為、以及我們在競品替換（competitive takeouts）方面的作為，也在加劇這個趨勢。因此，這些因素要完全拆分並量化給你哪些是循環性、哪些是——其他驅動因素並不容易，因為有時候這些其他驅動因素也會以更新汰換的形式呈現；也就是說，我們——我們的團隊正在與客戶談更新汰換。

And rather than refreshing just the capacity they have, we're seeing very strong expansion at refresh that is driven by new use cases. So those -- all of these factors kind of come together and manifest themselves in the numbers, but I think you can fairly think about it as -- there's clearly a refresh portion that is cyclical, and there is three or four drivers that we think are durable and that is why we have changed our view on the trajectory of hardware, not just for this year but for the next few years.

而不是只更新他們既有的容量，我們看到在更新汰換時有非常強勁的擴充，這是由新的使用情境所驅動。所以——所有這些因素某種程度上匯聚在一起，並反映在數字上；但我認為你可以合理地這樣看：其中確實有一部分更新汰換是循環性的，同時我們也認為有三到四個驅動因素是具持久性的，這也是為什麼我們改變了對硬體成長軌跡的看法，不僅是今年，也包括未來幾年。

That concludes our question-and-answer session. I will now turn the call back to Suzanne DuLong for closing remarks.

我們的問答環節到此結束。接下來我將把電話交回給 Suzanne DuLong 作結語。

Thank you, Tiffany. Thanks, everybody, for joining us. We look forward to seeing many of you out and about at conferences during the quarter. Please reach out with any questions.

謝謝你，Tiffany。也謝謝各位的參與。我們期待在本季期間於各項會議上與許多各位見面交流。如有任何問題，請隨時與我們聯繫。

Ladies and gentlemen, this concludes today's call. Thank you all for joining. You may now disconnect.

各位女士先生，今天的電話會議到此結束。感謝各位參與。您現在可以掛線。