Cloudflare Raises Long-Term Targets as Agentic Internet Traffic Explodes, Predicts Operating Leverage Inflection from AI Workloads
Analyst and Investor Day, June 9, 2026
Cloudflare announced a meaningful increase to its long-term financial targets at its investor day, raising its profitability North Star from a Rule of 40 to a Rule of 50 by fiscal 2027, with a clear path to exceed 50% thereafter. The company also increased its operating margin target range from over 20% to over 30%, while widening its gross margin range to 70% to 77% to reflect a multi-quarter journey back to historical levels. Perhaps most significantly, management committed to achieving GAAP profitability by 2028 at the latest, marking a notable acceleration in the company's profitability roadmap.
The catalyst driving this upgraded outlook is unmistakable. Bot and agent traffic has already exceeded human traffic on the internet, a milestone Cloudflare originally forecast for late 2027 before pulling the estimate forward to first half 2027 just months ago. The inflection came even faster than that revised timeline. CEO Matthew Prince noted that agentic traffic on Cloudflare's network surged 1,700% in the past year alone, and he would take the over on his own prediction that internet traffic will grow 10x to 100x over the next five years. "This is growing much faster than we ever anticipated," Prince acknowledged.
The Isolates Architecture Advantage in the Agentic Era
Rita Kozlov, Vice President of Product for the developer platform, delivered what may have been the most technically substantive presentation of the day. She made an emphatic case that Cloudflare's nine-year-old architectural bet on isolates rather than virtual machines or containers positions the company as uniquely capable of supporting agent-generated code at scale. The math is stark. Powering agents for the 100 million knowledge workers in the United States alone would require approximately 10 million CPUs, representing 20% to 30% of annual global server production. Scaling globally to the 1 billion knowledge workers worldwide, assuming conservative assumptions of just two agents per person, would demand 1 billion server CPUs, or 20 times current annual global production.
Kozlov argued that the cloud infrastructure that enabled the last era simply cannot scale to meet agentic demand. "Every application generally begins as a monolith," she explained. "For the past 20 years when we've been talking about scaling applications and microservices, this is exactly what we're talking about, running the same code over and over again. Agents just don't work like this. Agents don't need more copies of the same code. Agents need a distinct new piece of code written not just for every user, but for every single task and sometimes even subtask."
Cloudflare's isolates technology eliminates the need to spin up an entire operating system for VMs or a full language runtime for containers each time code needs to execute. The company can import agent-generated application code and execute it on demand with dramatically less overhead. According to Kozlov, this translates to 100x greater efficiency than traditional compute, and meaningful cost savings for customers. For applications requiring 10,000 agent runs, Cloudflare is approximately 63% more cost effective than alternatives. At 1 million agents, the cost advantage approaches 75%. For large enterprise customers operating at scale, these differences compound into substantial economic benefits.
Large Model Providers Choosing Cloudflare as Default Infrastructure
The theoretical advantages of the platform are now translating into commercial validation from the most sophisticated AI-native customers. OpenAI built its latest platform launched last week on Cloudflare. Anthropic is a partner. Prince noted that approximately 80% of leading AI companies use Cloudflare infrastructure. The company recently announced or commercialized major relationships with Anthropic, OpenAI, and Lovable specifically around its Workers developer platform as the preferred environment to build and scale AI agents.
Prince attributed this adoption to agents making fundamentally more rational purchasing decisions than humans. "When humans make decisions on what service they're going to buy or build, a lot of times it's what I'm used to, or it's who has the best advertising or what brand makes the most sense," he said. "What we're seeing is that as agents make those decisions, they actually make those decisions on much more foundational rational facts. Just because you have fancy airport advertising doesn't necessarily influence the agent. They choose based on what's the cheapest, what's the fastest, what's proven to be the best."
The company's cost discipline dating back to its founding is paying dividends in this environment. From the beginning, co-founders Prince and Michelle Zatlyn focused on driving bandwidth costs toward zero and building the most efficient routing engine possible. That obsession with unit economics now provides structural advantages as model companies seek to minimize inference and execution costs at massive scale.
Act 4 and the Emerging Business Model for the Agentic Internet
Chief Strategy Officer Stephanie Cohen presented what the company calls Act 4, Cloudflare's framework for monetizing its position at the intersection of content supply and AI demand. The thesis rests on several observations. First, human traffic to many content sites has declined dramatically, in some cases approaching what customers internally refer to as "Google Zero," where essentially no traffic arrives from search. Retail, software, IT and finance sectors have seen human traffic declines approaching 40% in less than a year. Yet total traffic to these properties sitting behind Cloudflare's network has surged, driven entirely by bots and agents.
The advertising-based business model that funded content creation for two decades is breaking down because bots don't click on ads. Someone must pay for the infrastructure supporting exponentially growing agentic traffic, and someone must compensate content creators so high-quality information continues to be produced. Cloudflare believes it is uniquely positioned to build the rails enabling this new ecosystem because it has both supply and demand at scale. More than 20% of the internet runs on Cloudflare, including 36% of the most visited sites and over 40% of the Fortune 500. Nearly 80% of leading AI companies use the platform.
The company already has commercial proof points. People Inc implemented Cloudflare's bot management in July 2025, and by September had meaningfully improved its negotiating leverage with large language model providers by creating scarcity. The Financial Times and Condé Nast are among the large media companies that have moved to Cloudflare to ensure they are "agent enabled." But Cohen emphasized the opportunity extends far beyond splashy content licensing deals with frontier labs. Cloudflare processes more than 500 million transactions per second, dwarfing the largest payment networks which process fewer than 100,000 transactions per second. The company developed the x402 protocol with Coinbase and Stripe to enable machine-to-machine micropayments at internet scale. Today there are more than 2 billion "402 payment required" responses daily on Cloudflare's network, establishing the scaffolding for a micropayment infrastructure that could fundamentally alter how information access is monetized.
Cohen was explicit about the revenue opportunity. "We see about 500 million transactions through our network on any given month. We think that somewhere between 1% and 10% of those, you can have some sort of micropayment," Prince said in his fireside chat with Phil Winslow. "All the AI companies are like, yes, that makes sense. That's something that seems actually pretty fair that when we access these pages, we'll pay a fraction of a penny in order to get that content." For Cloudflare, moving from protecting websites and applications to monetizing trusted automated demand through transaction fees would represent an entirely new revenue stream layered on top of existing products.
Security as the Killer Application for Zero Trust Adoption
One of the more surprising themes throughout the day was how AI security concerns are driving enterprise adoption of Cloudflare's Zero Trust platform in ways the company had not initially anticipated. Prince acknowledged that when Cloudflare entered the Zero Trust market, it was late relative to established vendors and struggled to consistently get into enterprise consideration sets. That dynamic has reversed as buying conversations now begin with CIOs and CTOs asking how to enable AI capabilities securely, rather than with CISOs evaluating standalone security products.
Mark Anderson, President of Revenue, noted the shift in customer urgency. "If you go back roughly 9 months, most conversations centered on enablement. Organizations wanted to experiment with AI, but they were equally focused on governance and controls. The question was, how do we safely allow employees and developers to start using these technologies. Fast forward to today, the conversation has changed dramatically. The urgency level has gone from 0 to 100. Organizations are no longer asking whether AI will impact their business, they're asking how quickly they can operationalize it."
The company's own internal AI deployment, which CIO Sam Ray detailed in his presentation on Cloudflare OS, demonstrates the security architecture customers are seeking. Because Cloudflare built Gateway and other security primitives from the beginning, the company could provide employees with AI agents that have access to hundreds of internal systems and tools while maintaining strict controls and complete audit trails. Ray explained that agent interactions don't occur directly with systems of record but instead execute through isolates running in Cloudflare's network, allowing granular permissions without overwhelming context windows. All AI inference flows through the company's AI Gateway, enabling dynamic routing to cost-appropriate models and providing complete visibility into AI usage.
The Zero Trust portfolio saw annual recurring revenue growth of 43% in fiscal 2025. Prince suggested that AI may prove to be the category's killer application, with demand accelerating as enterprises move from pilots into production deployments. "I think we're seeing customers move from pilots and experimentation into production every single day," Anderson said. "And increasingly, those deployments involve AI agents interacting with applications, infrastructure, data and business processes."
Developer Platform Inflection Arriving After Nine-Year Build
When Cloudflare announced Workers in 2017, Prince told Winslow that developer platforms take 10 years. Nine years in, the hockey stick inflection has arrived. Developer platform annual recurring revenue surged 137% year-over-year in fiscal 2025. Wrangler downloads, the CLI tool for Workers, increased nearly 1,000% year-over-year. The company announced the acquisition of VoidZero, the company behind Vite, one of the fastest-growing developer tools of the past several years. Cloudflare's plugin for Vite now represents approximately 10% of total Vite downloads.
Kozlov tied the developer platform momentum directly to agentic workloads. "Since November, December, we've seen a massive step function improvement in the productivity that models have been able to give to agents," she said, referencing models like Claude 4.6 and GPT 5.5. As models have become capable of writing increasingly sophisticated code autonomously, demand for execution environments optimized for dynamic, ephemeral workloads has accelerated. The developer platform provides the primitives, including Workers, Durable Objects, and the newly highlighted Workflows product, that agent systems require.
Importantly, Prince emphasized the platform's self-service nature creates adoption vectors unavailable to competitors. When OpenClaw and similar consumer-facing agent tools gained rapid traction, users immediately sought security solutions to safely provide these agents with broad system access. Cloudflare's combination of zero trust security and serverless compute emerged as a natural fit, with individual users adopting the platform and subsequently bringing it into their enterprise environments.
Unit Economics by Act Reveal Path to Margin Expansion
CFO Thomas Seifert provided the most granular view to date of how Cloudflare's four "Acts" differ in their unit economics, explaining why the company maintains confidence in margin expansion despite near-term gross margin pressure. Act 1, the original application security and performance business, resembles overall company averages with moderate cost to book, moderate cost to serve, and slightly higher attrition due to SMB exposure. Act 2, the Zero Trust portfolio, benefits from infrastructure already deployed for Act 1, resulting in very low cost to serve, very low attrition due to product stickiness, but higher cost to book given longer enterprise sales cycles and partner revenue sharing.
Act 3, the developer platform, currently carries lower gross margins than the rest of the portfolio, which Seifert acknowledged is the primary driver of recent gross margin compression alongside free-to-paid traffic mix shifts. However, he argued this is not indicative of inferior unit economics. "Platform businesses have this unique capacity that they can structurally shift," Seifert explained, drawing comparisons to AWS and Azure margin trajectories. Cloudflare has already demonstrated the ability to reduce developer platform infrastructure costs, as evidenced by Kozlov's data showing 60% to 75% cost advantages over alternatives for agent workloads. "At scale, we think that Act 3 could surpass Act 1 and Act 2 from a unit economic perspective," Seifert said.
Act 4, which Cohen defined as the combination of Act 1 and Act 3 plus transaction monetization, could prove transformative. Because traffic is already on the network, Act 4 will have by far the lowest cost to book and lowest cost to serve of any product category. Attrition should be minimal. "This unit economic curve for Act 4 could literally shift us off the chart from a performance and potential perspective," Seifert stated.
The company expects gross margins to stabilize in the second quarter and then begin a multi-quarter journey back toward the high end of the new 70% to 77% range, driven by engineering efficiency improvements and revenue mix shifting as Act 1 and Act 2 products grow alongside Act 3. More importantly, Seifert noted unit economic margins have continued expanding even as gross margins compressed, reaching approximately 44% exiting fiscal 2025.
Operating Expense Leverage from AI-Driven Productivity
The path to the upgraded operating margin targets incorporates meaningful assumptions about AI-driven productivity improvements within Cloudflare's own operations. Seifert lowered the sales and marketing expense target as a percentage of revenue, attributing this to two factors. First, Act 3 and Act 4 products have structurally lower customer acquisition costs. Second, Ray's presentation on Cloudflare OS demonstrated how the company is collapsing traditional go-to-market support ratios. "In the past, we have been building go-to-market organizations where we scale the support of quota-carrying employees by ratios," Seifert said. "These ratios are literally contracting and falling away." Rather than allowing those savings to drop to the bottom line, the company will redeploy capital into additional quota-carrying capacity, increasing sales productivity per dollar of spend.
Anderson reinforced this point, outlining how AI agents are automating pre-sales research, proposal generation, solution architecture customization, and post-sales adoption workflows. "Today, sales professionals probably spend less than half of their time actually selling. AI gives us an opportunity to change this big time," he said. The company is also leveraging AI for automated code review, dramatically reducing production incidents. Prince cited data from the production engineering team showing incident rates falling off a cliff after implementing an AI model trained on ten years of internal incident data to check every code release and configuration change.
Seifert projected the most dramatic operating leverage will come from general and administrative functions. "AI is transforming the G&A workflows that we have been operating with," he said. "I see this in my own finance organization now every day, how roles are collapsing, how functions are moving from manual and monolithic ERPs to agent-driven execution, whether that is accounts receivable, revenue recognition, even investor relations." The company believes it will lead a broader industry shift in G&A cost benchmarks, targeting G&A as a percentage of revenue well below historical software norms.
Revenue Recognition Complexity from Usage Model Evolution
Seifert spent considerable time explaining revenue recognition dynamics that will influence quarterly results, particularly as large customers increasingly adopt pool of funds contracts and as usage growth outpaces committed caps. The majority of Cloudflare's business now operates on cap models where customers purchase a ceiling of utilization, such as 10 million requests per month for a web application firewall. Revenue recognition is smooth and predictable as long as usage remains below the cap. However, when customers exceed their cap and purchase a larger commitment, revenue recognition lags the inflection in actual usage because customers must estimate future consumption rather than simply paying for current usage.
Pool of funds contracts, which represented 20% of new annual contract value in the fourth quarter, create even more pronounced revenue recognition patterns. Unlike ratable recognition, pool of funds revenue is recognized as services are consumed, creating a ramping profile. Early in the contract period, revenue lags what it would have been under ratable recognition, before catching up and eventually exceeding it. When large customers significantly increase their pool of funds commitments, as several AI-native customers have done recently, the step-up in commitment can temporarily depress revenue run rates even as consumption accelerates. Seifert shared an example of a large AI customer that increased its pool of funds commitment by more than 215% in the fourth quarter, negotiating better rates in the process. This created a revenue dip in the first and second quarters before the run rate begins recovering and ultimately exceeding prior levels.
Critically, Seifert emphasized that consumption and network utilization do not decline during these revenue dips. The company continues incurring infrastructure costs to support the increased usage, creating a temporary headwind to gross margins until revenue catch-up occurs. Management is explicitly managing investor expectations for quarterly variability as these dynamics play out, while maintaining conviction in the underlying business trajectory.
Go-to-Market Transformation Driving Enterprise Traction
Anderson, who announced he will retire at the end of fiscal 2026 after a seven-year tenure, provided a retrospective on the sales transformation he orchestrated. When he joined in late 2023, a large portion of his time was spent simply introducing Cloudflare to enterprise and government decision makers. Today, customers increasingly come to Cloudflare because the challenges they face align with what the company was built to solve. Brand awareness and enterprise engagement have improved materially.
The company has made substantial progress adding stage-appropriate leadership and frontline management, refining market segmentation to address the distinct needs of SMB, mid-market, digital native, large enterprise and public sector customers, and building a partner ecosystem that now contributes a steadily growing percentage of revenue. Sales productivity has reached new highs, with Anderson noting the momentum equation of ACV growth equals sales capacity times productivity is working in Cloudflare's favor. "When you have a differentiated suite of products at a time when customers are realizing they need this differentiation, and you can continue to build and ramp high-performing teams to articulate a compelling value proposition, you can roll up market share in a very meaningful way," he said.
The company's growth in million-dollar-plus customers shows clear inflection beginning in the second half of 2024, with momentum building through 2025 and the first quarter of 2026. Cloudflare counts 42% of the Fortune 500 as customers, though Anderson stated he will not be satisfied until all 500 are customers. Partner contribution is accelerating as the largest opportunities increasingly involve complex transformation projects where partners provide strategic guidance, implementation support and long-term operational expertise.
Notably, Anderson believes Cloudflare is still in early innings with large enterprise adoption despite the progress to date. The combination of improving sales execution, expanding product portfolio, and the market tailwind from AI-driven infrastructure modernization provides confidence in sustained capacity expansion. "My plan is to continue to invest in sales capacity, and I expect that sales productivity will continue to grow for a very long time," Anderson said.
Customers Facing Google Zero and Seeking Agent Enablement
Cohen's presentation on Act 4 included stark data on traffic pattern shifts affecting Cloudflare's customer base. For every hour spent searching online today, only 15 minutes is spent on the open web. Many news and content sites are preparing for "Google Zero," a moment where none of their traffic comes from search engines. The Faustian bargain between search and content has dissolved. But Cohen emphasized the challenge extends far beyond media. "Retail, software, IT and finance are all seeing meaningful declines in human traffic on their sites to the tune of almost 40% in less than 1 year," she said.
Cloudflare's positioning at the intersection of supply and demand is generating commercial momentum. Reddit and PeopleInc are new customers, while existing customers like LinkedIn are expanding as Cloudflare products become strategic imperatives rather than tactical purchases. In sectors most affected by the agentic internet, Cloudflare is experiencing revenue growth north of 37%. Cohen noted that conversations around application performance and security now involve the CEO in addition to the CTO and CSO. "Cloudflare has never been more important to the C-suite," she stated.
The company's bot management products provide the foundation for customers to create scarcity and negotiate from a position of strength with AI companies seeking to access their content. But Cloudflare is also building tools to help AI companies more efficiently access information. The same signals that make the company's bot management best-in-class help AI companies decide what content to scrape and how often. The company's Agents SDK makes it easier to build agents that respect site owner preferences, providing faster and better access to content. "The vast majority of AI companies and agents want to do the right thing," Cohen said. "The reality is it's just not easy to traverse the internet."
Cohen argued Cloudflare's ability to see where asymmetries exist in content access across different model providers creates an opportunity to help smaller labs compete with the largest incumbents. Currently one large LLM has access to nearly twice as much information as the next model. "We want to create an ecosystem where small model providers have this opportunity to actually compete with the world's largest companies by being able to access the same information," she said. The combination of primitives across security, developer platform, and emerging payment infrastructure positions Cloudflare to enable what Cohen calls "the business model for the agentic internet."
Internal Cloudflare OS Deployment as Product Incubator
Ray's presentation on Cloudflare OS generated substantial investor interest, with several attendees asking whether the company intends to commercialize the internal tooling. The platform provides every employee with a browser-based agent workspace running on Cloudflare's infrastructure, eliminating the need for local software configuration and keeping workloads running when laptops are closed. It includes a centrally managed library of skill files mapping jobs to be done across various organizational functions, safe and secure access to hundreds of internal systems of record without overwhelming context windows by executing tool calls through code running in isolates, dynamic routing of all AI inference through the AI Gateway to optimize for cost and performance, and the ability for employees to build and schedule automated agents and workflows.
Prince indicated the company did not initially intend to turn Cloudflare OS into a product, but customer demand after mentions on earnings calls and in private conversations has been overwhelming. "Every time we show this to people, they're like, that's what I want. I need that," he said. The company is currently allowing select customers to use the platform and is evaluating commercialization options. Ray noted that when Prince mentioned Cloudflare OS on the most recent earnings call, multiple trusted customers reached out asking Cloudflare to deploy the system for them.
The platform's security model appears to be a key differentiator. "The second question in every presentation is what's the security model," Prince said. "Because we had that infrastructure built in from the beginning because that was core to every primitive that we built, I think that's allowed us to develop the tooling a lot faster than anyone else." The system provides complete auditability and control, with all agent actions visible to security and IT teams, making it viable for enterprise deployment in ways consumer AI tools are not.
Implications for Hyperscalers and Legacy Clouds
While Cloudflare executives were measured in their commentary on competitive dynamics, the implications of Kozlov's presentation on infrastructure requirements for agentic workloads are clear. If agents require distinct code execution for every task and subtask, and if spinning up VMs or containers for each of these ephemeral workloads is economically and technically infeasible at scale, then hyperscalers face a potentially significant architectural disadvantage in capturing agent compute spend. The math Kozlov presented on CPU requirements to power agents globally suggests the current cloud infrastructure paradigm hits fundamental scaling constraints well before reaching mass adoption.
Prince was more explicit about security vendors. "I think for the next 2 years, every security company is going to be wildly busy," he said, citing an expected wave of AI-enabled vulnerabilities. "But I think that what's going to change is that going forward, software is going to get a lot more secure. I think there's a flip probably about 2 years from now where all of a sudden it's gotten much harder to find where those bugs are. I think security as a standalone category, if you're just a security company, will be really exciting for 2 years and then kind of maybe fades away after that."
The company's partnership with frontier model labs to receive early feeds on newly discovered vulnerabilities, combined with its position securing more than 20% of the internet, allows Cloudflare to quietly patch vulnerabilities across its customer base before they can be widely exploited. This role as a trusted security partner to both the AI companies discovering vulnerabilities and the enterprises that need protection reinforces Cloudflare's platform value as AI-driven threats accelerate.