Cloudflare Profit Forecast Rises as AI Expands Network Demand

Cloudflare raised its annual earnings outlook again on November 5, strengthening its argument that artificial intelligence creates demand beyond the companies building models. The Cloudflare profit forecast now reflects faster business growth, although the company still reported a net loss under standard accounting rules.

The central tension concerns how that demand becomes profit. AI applications need connections, protection, and places to run software. Selling those services creates an opportunity, but serving additional workloads also carries costs.

Cloudflare’s results put that balance in focus for developers and enterprise buyers. Competitors including Akamai are pursuing the same surrounding infrastructure opportunity. The question is which providers can turn AI activity into lasting customer relationships and attractive economics.

The Cloudflare Profit Forecast Moves Higher Again

Cloudflare increased its expectations for annual adjusted earnings, extending a sequence of upward revisions during 2026.

The company’s quarterly financial results put expected annual adjusted earnings per share at $1.35. Earnings per share measures income allocated to each share. Adjusted figures exclude certain expenses.

Third-quarter revenue reached $782.5 million, increasing 39% from the previous year. The quarter ended September 30, making the announcement a report on completed business activity alongside updated expectations.

For comparison, the previous annual outlook projected adjusted earnings per share between $1.25 and $1.26. That August release reported second-quarter revenue growth of 36%.

The progression extends further back. Cloudflare’s first-quarter outlook projected annual adjusted earnings per share between $1.19 and $1.20. First-quarter revenue increased 34%.

Those comparisons establish the direction of the revision without depending on a stock-market reaction. Management has become more optimistic about earnings as reported revenue growth has accelerated.

However, an earnings forecast describes management’s expectations. It does not guarantee the final result or identify the contribution of every product.

The announcement also requires an accounting distinction. GAAP, or generally accepted accounting principles, provides the standard framework for reported financial statements. Cloudflare recorded a quarterly GAAP net loss of $34.1 million.

Calling this an improved profit outlook is therefore accurate when referring to adjusted earnings. Calling it an announcement of company-wide GAAP profitability would misrepresent the results.

That distinction matters beyond financial terminology. A provider can expand its business and improve selected measures while continuing to incur expenses that exceed reported income.

For customers, the revision provides evidence of business momentum. It offers less direct information about whether a particular application runs better, costs less to operate, or meets security requirements.

For competitors, the announcement raises the pressure to demonstrate commercial results behind their own AI strategies. Product announcements alone offer a weaker comparison than revenue and customer commitments.

The useful starting point is therefore narrow and concrete: Cloudflare expects more adjusted earnings from its business than it previously forecast. Understanding the AI connection requires examining what customers actually need from that business.

AI Demand Reaches Beyond the Model Provider

Cloudflare’s opportunity rests on the work surrounding an AI response, including connecting services, controlling requests, and operating applications.

A model generates an answer, but a working product must do considerably more. It needs to receive a request and deliver a response.

An AI agent, software that uses models and tools to carry out tasks, adds further interactions. It might retrieve information before deciding its next action.

That creates a plausible route from AI adoption to infrastructure demand. Applications need dependable connections even when their underlying models come from another company.

The important commercial distinction concerns where customers encounter operational problems. A demonstration can emphasize answer quality. A production application must also handle failures and unpredictable usage.

Cloudflare documents this surrounding role through AI application controls. Its AI Gateway sits between applications and model providers, offering request visibility and management features.

Those documented features include logging, request limits, retries, and fallback between models. Caching, which stores responses for reuse, offers another way to manage repeated work.

These capabilities explain why a network provider can participate without developing the model that produces the answer. Its role can sit between the application and external services.

The commercial argument is strongest when that position solves an expensive operational problem. A customer needs a reason to keep the intermediary after an experiment ends.

For example, a team might value a consistent view of failures across different models. Another might prioritize preventing an application from sending uncontrolled requests.

Those are illustrative purchasing considerations, rather than evidence of particular Cloudflare contracts. Their relevance comes from connecting documented capabilities to decisions buyers must make.

The distinction also clarifies what “AI demand” means. It can refer to direct model execution, supporting application services, or protection for systems interacting with AI.

Those categories should not be treated as interchangeable. Their technical requirements differ, and their contribution to company earnings needs separate evidence.

Cloudflare’s broader business also serves customers whose needs extend beyond AI. A security purchase does not automatically become an AI purchase because management discusses artificial intelligence.

The defensible interpretation is that AI expands the reasons customers might use infrastructure services. The earnings announcement does not quantify every part of that expansion.

This leaves a practical challenge for Cloudflare. It must connect growing interest in AI applications with services customers continue using after initial development.

Enterprise buyers face the corresponding challenge. They must identify which infrastructure additions improve their applications and which merely add another dependency.

A favorable earnings announcement makes that evaluation more timely. It does not replace workload-specific evidence or establish that every AI project needs the same architecture.

The Business Opportunity Is in Keeping Applications Running

The strongest infrastructure argument starts with a complete application and the operational work required to sustain it.

Cloudflare has already described concrete examples beyond text chat. Its partner model integration brought Leonardo.Ai image models and Deepgram speech models to Workers AI in August 2025.

Workers AI is Cloudflare’s service for running models on its infrastructure. The announcement outlined image-generation applications and voice agents as intended uses.

Cloudflare described combining image generation with application logic, storage, and image delivery. For voice applications, it described connecting speech processing with live communication services.

These examples provide a useful reference for the current earnings story. They show how model access can become part of a larger infrastructure relationship.

Consider the image-generation example as an application design. Producing an image completes only part of the user’s request. The result must remain available afterward.

A product may need to associate that image with an account, display it reliably, and support later access. Those surrounding requirements create additional infrastructure decisions.

The commercial opportunity comes from serving those requirements together. A provider that handles more of the application has more opportunities to become part of its ongoing operation.

That outcome remains conditional. A broad product catalog does not establish that customers adopt every component or that bundling always improves their experience.

Integration can reduce the coordination required between services. It can also make a migration more complicated when application behavior depends on provider-specific features.

A voice application highlights another consideration: responsiveness. A conversation can feel awkward when delays accumulate across speech processing, model generation, and delivery.

No single component determines the entire experience. Buyers need to evaluate the complete interaction rather than treating a model benchmark as an application benchmark.

Cloudflare’s partner announcement describes its intended architecture. It does not independently prove superior performance across customer workloads or establish a specific contribution to current earnings.

The inference for this earnings story is nevertheless straightforward. If developers build complete applications around a provider’s services, usage can extend beyond the initial model call.

That creates a different relationship from a short experiment with a model endpoint, the address an application contacts to request model output.

An experiment can end when a team changes models. An application relationship can persist because storage, access controls, and delivery remain necessary.

This is where Cloudflare’s annual outlook becomes relevant to developers. Business momentum supports the argument that its infrastructure strategy deserves consideration alongside individual AI features.

It still leaves the adoption question open. The evidence needed is sustained application usage and expanding customer commitments, rather than the number of capabilities listed on a product page.

For buyers, that shifts attention toward operational fit. The useful comparison concerns the full application’s behavior, ownership requirements, and ability to evolve.

Akamai Shows Why the Opportunity Is Competitive

Cloudflare faces competitors that also want to provide the infrastructure around AI applications, limiting any claim to an uncontested market.

Akamai’s managed inference service illustrates that overlap. Inference means running a trained model to produce an output.

Akamai describes managed model access alongside its broader cloud services. Its offering includes shared capacity and dedicated deployments, with availability conditions described on the product page.

That positioning gives buyers another route to combine model execution with application infrastructure. It also shows why Cloudflare’s financial momentum should not be confused with exclusive technical ownership.

Both companies describe ways to reduce infrastructure management for developers. Their marketing establishes competing propositions, rather than independent proof that either serves every workload better.

The central competition concerns ownership of the application relationship. A provider benefits when customers rely on it for recurring operational needs surrounding AI.

This competition can occur even when customers use the same underlying model. The difference may sit in deployment controls, integration requirements, or the surrounding services.

For a buyer, that makes model availability an incomplete comparison. Access to a recognizable model does not resolve questions about service behavior under production conditions.

The same applies to claims about global infrastructure. Geographic reach can matter, but the application’s actual dependencies determine where delays and constraints appear.

A widely distributed application can still depend on a distant database. A nearby model endpoint cannot eliminate every delay elsewhere in the system.

These are architectural considerations, rather than claims about a measured performance gap between Cloudflare and Akamai. They explain why competitive evaluation needs application-level evidence.

The pressure on Cloudflare is to make its broader platform useful enough that customers choose additional services. A forecast increase strengthens its commercial position.

The pressure on competitors is to demonstrate an equally convincing path from AI product availability to continued customer usage. A launch creates an option; adoption creates a business.

Buyers retain an important source of negotiating strength: the ability to distinguish services they need from services included in a broader pitch.

A team might prefer a single provider for simpler operations. Another might retain separate providers because its requirements differ across application components.

Neither choice follows automatically from the earnings announcement. The right comparison depends on the work the application performs and the obligations its operator accepts.

This is also why an AI infrastructure market can support multiple providers. Different applications place different weights on control, convenience, deployment flexibility, and operational responsibility.

Cloudflare’s results raise the stakes within that competition. They do not settle it or establish that competitors must follow exactly the same product strategy.

The next question is how effectively expanding usage translates into earnings after the costs of serving customers enter the calculation.

Faster Growth Leaves the Economics Open to Scrutiny

Growing activity supports an infrastructure business only when the provider can serve it on acceptable economic terms.

The Cloudflare profit forecast reflects management’s confidence in the business. It does not establish that every additional AI interaction produces the same financial return.

Cloudflare reported a third-quarter adjusted gross margin of 72.4%, compared with 75.3% a year earlier. Gross margin measures revenue remaining after the direct costs of providing services.

Its adjusted operating margin rose to 17.5%, from 15.3%. Operating margin accounts for additional business expenses beyond direct service costs.

Those measures moved in different directions. That is a reason to examine the business carefully, rather than summarize profitability with a single adjective.

The figures do not independently identify AI as the cause of gross-margin pressure. Product mix, infrastructure spending, and other factors require evidence before assigning causation.

They also do not establish that higher demand lacks value. A company can accept different economics across products while pursuing a broader customer relationship.

The question is whether those choices produce durable returns. That requires examining costs and earnings together over time.

An additional request represents work the provider must handle. Whether it produces additional revenue depends on the customer agreement and the service involved.

Some activity can increase operating requirements without creating a proportionate increase in customer spending. Treating all traffic as equally valuable would obscure that distinction.

AI introduces another complication through efficiency. A customer may want a service precisely because it reduces repeated model calls or limits unnecessary work.

That can still support an infrastructure business. The provider’s value would come from helping the customer operate efficiently, rather than maximizing raw request volume.

This is the clearest skeptical response to an uncomplicated AI growth narrative. Customer value and infrastructure activity do not always rise together.

An effective service might help an application complete tasks with fewer calls. Its business case then depends on the value customers assign to that improvement.

The same reasoning applies to application failures. Repeated attempts can generate activity while frustrating users and increasing the cost of completing useful work.

For enterprise buyers, successful task completion is therefore more informative than request growth alone. A busy system does not necessarily represent a productive system.

For Cloudflare, the corresponding commercial question concerns the quality of adoption. Continuing production usage carries a different implication from temporary experimentation.

The available evidence supports an improved company outlook. It leaves room for further scrutiny of the AI contribution and the costs attached to different workloads.

That is a normal boundary for an earnings analysis. Financial reporting establishes business results, while product documentation establishes capabilities. Connecting them requires additional disclosure.

Readers should therefore resist attributing every improvement to AI or every cost pressure to AI infrastructure. Both conclusions would exceed the evidence presented here.

The stronger assessment keeps the claims separate. Cloudflare has raised its expectations, offers relevant application infrastructure, and operates in a competitive market with continuing economic tradeoffs.

Watch Adoption, Margins, and Buyer Choice Next

The next useful evidence will connect customer commitments, service economics, and competitive adoption.

First, watch whether subsequent disclosures describe AI applications moving into sustained production use. The strongest examples will explain what customers operate and which services they depend on.

A named customer story can provide useful detail, but its scope matters. A limited trial does not establish broad deployment across an organization.

More specific disclosures would strengthen the argument that AI supports recurring infrastructure demand. Continued reliance on general enthusiasm would leave the contribution harder to assess.

Buyers should also distinguish customer interest from customer obligation. Evaluating a service, deploying it, and committing to continued usage represent different stages.

Second, watch whether earnings progress is accompanied by evidence about service costs. Revenue growth alone cannot answer whether new workloads improve the business’s economics.

The relevant question concerns how operating improvements relate to the cost of providing services. Management explanations will matter when financial measures move differently.

If customer adoption expands while service economics remain healthy, the infrastructure thesis becomes more convincing. If growth requires persistent concessions, the interpretation becomes less favorable.

This does not require every financial measure to improve simultaneously. It requires a credible explanation of how current investment supports future customer value and returns.

Third, watch competitive adoption around complete applications. Akamai’s offerings make clear that customers have alternative ways to combine AI execution with supporting infrastructure.

New product availability will matter most when accompanied by customer usage and operational detail. Announcements alone provide limited evidence of a changed competitive position.

Cloudflare’s position would strengthen if customers continue expanding their dependence on its application services. Competitors would challenge that position through comparable adoption and clearer operational advantages.

Developers should translate these signals into specific questions about their own systems. Which provider handles failures, where does application data live, and what would a migration require?

Enterprise buyers should ask how commitments align with actual usage. A contract should reflect the application’s operational needs rather than assume that experimentation predicts future demand.

Knowledge workers have a related interest when their tools depend on connected AI services. Reliability affects whether an assistant can retrieve information and complete work consistently.

The earnings story therefore reaches beyond shareholders. It concerns which companies will operate the infrastructure connecting models with the applications people actually use.

The Cloudflare profit forecast is an encouraging business signal within that larger contest. Its significance will become clearer as adoption and operating evidence accumulate.

Before choosing infrastructure for an AI project, document the assumptions behind the decision and the evidence that would change them. Keep deployment notes, service limits, and observed failures accessible to everyone responsible for the application. A searchable knowledge base can help preserve that reasoning as products and requirements change. Then revisit the decision when providers disclose concrete adoption or operating results. Ask whether the evidence supports the application you need to run, including its reliability and ownership requirements. Which part of your current architecture has demonstrated lasting value, and which part still rests on an untested expectation about AI demand?