Canva entered 2026 with 260 million monthly users and $3.5 billion in annual revenue, according to its review of 2025. It was used by 95 per cent of the Fortune 500 and had extended far beyond the templates that first made design accessible. Presentations, documents, whiteboards, websites, video, data tools and enterprise controls now sit inside the same product family. This breadth gives Melanie Perkins a credible claim on a much larger market than design software.
Canva AI 2.0 raises the ambition again. Introduced in 2026 as a research preview, it moves the platform towards conversational and agentic creation. A user can begin with an objective rather than a file type, generate layered editable work and ask the system to research, connect data, schedule activity, apply brand intelligence or build interactive experiences. Canva’s proprietary Design Model is intended to understand structure, hierarchy and brand logic rather than produce a flat image.
The strategic direction is understandable. Generative AI is compressing the cost of first drafts and eroding the boundaries between design, productivity and marketing software. If Canva remained a convenient template library, it would risk becoming a front end for models controlled by others. By building a creative operating system, Perkins is trying to own the workflow from idea to published output.
The danger is that Canva solves the competitive problem by sacrificing the quality that created the company: simplicity. An operating system can become a collection of adjacent features, overlapping plans and unfamiliar controls. Perkins must make a more capable Canva feel easier, not merely more powerful. That is a product architecture challenge, an enterprise sales challenge and a test of organisational restraint.
The interface is now the strategy
Canva’s early proposition reduced professional design friction for people without specialist training. The user selected a format, adapted a template and produced work quickly. Agentic software changes that interaction. The starting point becomes a conversation, and the system may choose tools, sources and layouts on the user’s behalf. Done well, this removes more friction. Done poorly, it replaces visible complexity with unpredictable behaviour.
Editable output is therefore critical. Many generative tools produce an attractive result that becomes difficult to revise. Canva’s layered approach allows teams to alter individual elements, enforce brand rules and continue collaborating. This preserves the user’s agency and connects AI generation with the company’s existing strength in visual editing. The value is not simply faster creation; it is faster creation that remains governable.
Perkins should insist on a consistent mental model across formats. A user moving from a presentation to a campaign, website or spreadsheet should recognise how content, permissions and brand assets behave. AI actions should expose their effects and make reversal easy. Living memory can save time, but customers need to know what is remembered, at which level and how it can be cleared. Simplicity requires clarity about hidden systems.
Feature launches must also be judged by completed outcomes rather than usage in isolation. A scheduling tool is valuable if a campaign reaches market faster. Research is valuable if it improves a brief. Code generation is valuable if a working interaction can be maintained after creation. If each feature becomes a destination competing for attention, the operating-system thesis will fragment.
Enterprise growth changes the standard of trust
Canva Enterprise has moved the company deeper into organisations that require central administration, approvals, identity management, auditability and security. It supports single sign-on, provisioning, brand controls and managed environments, with integrations across common workplace platforms. Thousands of companies adopted the enterprise offer in its first year, and customers have reported reductions in design time and brand-review delays.
This market offers larger and more durable contracts, but it does not buy on delight alone. Chief information officers need clarity on data residency, model training, intellectual-property risk, access controls and incident response. Marketing leaders want creative speed without unauthorised claims or brand drift. Designers need assurance that democratisation will not turn governance into a bottleneck they are asked to repair.
Canva Shield and administrative controls are necessary, yet trust depends on defaults as much as settings. An organisation should be able to decide which models and connectors are available, what data can leave a workspace and which outputs require approval. The platform should produce records that compliance teams can understand without specialised investigation. Enterprise customers will judge Canva by the least controlled path through the product, not the best-configured one.
The expansion also requires a different commercial machine. Product-led adoption can reveal where Canva already exists inside a company, but enterprise conversion needs sales, customer success, implementation and partners. Relationships with resellers, agencies and consultancies widen reach. Perkins must add that machinery without letting the roadmap become dominated by a small number of large buyers or turning onboarding into the complexity Canva was created to remove.
AI economics must work at mass-market scale
Serving generative and agentic features to hundreds of millions of people creates a cost structure unlike conventional software. Model inference, storage, search and connectors have real marginal expense. More sophisticated tasks may require several model calls and external data retrieval before producing a result. A generous free product can accelerate adoption while quietly increasing cost faster than revenue.
Canva needs an economic architecture that directs expensive capabilities to customers who value them without making the product feel punitive. Usage allowances, plan differentiation and efficient proprietary models can help. So can routing simple tasks to smaller systems and reserving costly reasoning for work that requires it. The company’s scale provides training signals and purchasing power, but scale is advantageous only when unit cost declines with learning.
Perkins must also decide where ownership matters. A proprietary Design Model can differentiate output and reduce reliance on general-purpose providers. External models may remain useful for language, image or research tasks. The correct approach is likely a governed mixture, with Canva controlling orchestration, brand context, editing and the customer relationship. Attempting to own every model would consume capital and attention; depending entirely on suppliers would weaken defensibility.
Revenue quality matters alongside growth. The company should track conversion from free to paid, retention across teams, enterprise expansion and the gross margin of AI-heavy cohorts. A feature that generates attention but lowers willingness to pay may be strategically useful for a time, but it should not be mistaken for an economic moat.
Creators are part of the supply chain
Canva’s library and template ecosystem were built with contributions from creators whose work helps users start quickly. Generative AI changes that relationship because models can imitate styles, produce substitutes and use content as training material. Canva previously committed $200 million over three years to content and AI royalties and stated that creator data would require permission for model training.
The principle remains strategically important. A creative platform needs contributors to believe that participation expands their opportunity rather than trains their replacement. Compensation must be understandable, attribution meaningful where appropriate and opt-in choices genuine. The company should publish enough information for creators to assess how programmes work without disclosing sensitive model details.
This is not only an ethical question. It affects supply quality and regulatory exposure. If talented creators withdraw or provide generic material, the library becomes less distinctive. If rights are unclear, enterprise customers inherit risk. Perkins can turn fair treatment into product advantage by making licensed provenance and usage rights easier for organisations to manage.
Acquisitions should close workflows, not collect tools
Canva has used acquisitions to add capabilities in professional design, data visualisation and creative production. Its intention to acquire MagicBrief, a creative-intelligence platform whose tools had analysed more than $6 billion in advertising spend, points towards performance marketing. The logic is to connect the making of an advertisement with evidence about what works.
That connection could move Canva closer to marketing budgets and measurable return. It could also pull the company into attribution, media data and optimisation, fields with different customers and technical demands. Perkins should integrate such capabilities around a clear workflow: insight, creation, approval, publication and learning. A separate product tab with little shared data would add surface area without strengthening the platform.
The same discipline applies to every adjacency. Canva does not need to replace every specialist tool. It needs to make common creative work flow smoothly and allow specialists to connect when depth is required. An operating system wins partly through what it enables others to do, not only through the functions it absorbs.
Private scale brings public-company expectations
Canva remains privately held, profitable and able to invest without quarterly market pressure. Its size nevertheless creates expectations associated with a public institution: durable controls, financial predictability, succession depth and transparent treatment of employees and secondary shareholders. Speculation about a future listing will continue, but timing should be subordinate to operational readiness.
Perkins and her co-founders need a company that can explain the economics of AI, the responsibilities of its models and the allocation of capital across a widening product portfolio. Leadership must extend beyond the founders, with accountable executives for enterprise, technology, trust and international markets. The mission can remain expansive while operating decisions become more formal.
The scorecard for the next phase should be direct. Does AI increase the number and quality of finished outcomes per customer? Do enterprise teams expand usage after governance is enabled? Does gross margin remain resilient as model use rises? Can users understand and control their data? Do creators continue to contribute? Are acquisitions becoming integrated workflows?
Canva’s scale proves that Perkins made difficult creative software approachable. Her new challenge is to make an intelligent, multi-format, enterprise-grade system approachable without pretending that its risks are simple. If Canva AI can preserve editability, trust and coherence while expanding what users accomplish, the operating-system ambition will be credible. If complexity accumulates faster than value, the company will have recreated the barrier it once removed.
