Canva told staff on 4 November that it is cutting 22% of its workforce — 1,271 people — in what co-founder and CEO Melanie Perkins called "without doubt the hardest moment in our thirteen-year history" (BBC). For teams building software in Nairobi, Lagos or Kigali, the interesting part is not the headcount. It is the reason behind it, because the same arithmetic is already running in our market.
What Canva actually announced
The cut is not a simple redundancy round. The 1,271 people will either lose their jobs, be asked to relocate to a centralised hub, or move into different roles (BBC). Canva is merging its product, growth and platform teams, reducing "subspecialisation" so people carry broader responsibilities, making Sydney the primary home for more product teams, and asking hybrid staff to be in the office two to three days a week from next year.
The financial context is the part worth reading twice. Canva cut its internal revenue growth forecast to 20% from 30% to absorb rising AI costs (TechCrunch), and its backers reportedly marked the business down from about A$60bn to roughly A$49bn (BBC). Growth slowed for the first time in a decade, and the company says it is shipping its own models to reduce reliance on third-party AI.
The number that matters is not the headcount
Twenty-two percent is dramatic, but it is a symptom. The mechanism is that generative features convert a software business from a high-gross-margin product into one with a variable cost per use, and that cost lands before the revenue does. Every prompt, generation and retry is a bill. When a company has spent a decade being valued on growth, a cost line that scales with usage is the one thing its structure cannot absorb quietly.
That is the sentence to take home: AI did not remove the jobs; the unit economics of AI did. Canva's answer — consolidate teams, cut subspecialisation, own more of the model layer — is what a company does when the marginal cost curve turns against it.
Three consequences for teams here
1. Your AI features are a variable cost with a fixed price. Most Kenyan software is sold on a per-seat or per-branch price, agreed annually. If a customer's usage doubles, your cost doubles and your price does not. Instrument cost per request before you ship the second AI feature, not after.
2. Distributed talent gets re-priced. Hub consolidation and office mandates are a deliberate trade: cheaper coordination for more expensive real estate. African teams have competed on the opposite trade — talent that is excellent and remote. That advantage holds only if the coordination cost stays low, which is a process problem, not a salary problem.
3. The funding bar moves from growth to margin. A valuation cut from A$60bn to A$49bn (BBC) is a message to every founder raising on a growth story: the multiple you are asking for now depends on the margin you can defend. Expect diligence questions about infrastructure spend, model costs and retention — the same numbers Canva just reorganised around.
What to do this quarter
- Price AI work per outcome, not per seat. A fixed fee for a fixed volume of documents, reconciliations or filings survives a cost shock; an unlimited seat does not. Our breakdown of what eTIMS integration actually costs is the same exercise for a compliance build.
- Make the variable cost visible. Log tokens and requests per tenant, set a hard ceiling per account, and cache aggressively. If you cannot answer "what does our heaviest customer cost us per month?", you are one pricing decision away from Canva's problem.
- Own the parts of the stack that carry margin. Managed platforms are a fine start and a poor destination once volume grows. The arithmetic behind a cheaper alternative to Railway is the same arithmetic that pushed Canva toward its own models.
- Keep compliance as a fixed line you can see. Registration and breach obligations under the ODPC framework do not scale down when revenue does. Budget them annually, not per project.
- Keep the hiring plan reversible. Canva is centralising because it can; a 20-person team in Nairobi cannot absorb a forced relocation. Contract capacity you can stop, and keep nil-return filing and statutory hygiene boring and current — the cheapest risk reduction available.
What this does not mean
It does not mean AI features are a mistake, and it is not evidence that design software is dying. Canva is still a large, profitable-by-design business making a deliberate structural bet: fewer, broader roles, more centralised, more of the stack owned in-house. What it does mean is that the era of shipping generative features without knowing their unit cost is closing — in Sydney and in Nairobi at the same time.
If your team is pricing AI work now, the question to answer this week is simple: what does one generated document cost you, and what do you charge for it?
Download the full deep dive (PDF): canva-layoffs-african-software-teams.pdf