Cloud spend is the only major expense that can grow fast without anyone making a clear decision.
It drifts. A new service ships. An environment never gets deleted. A “temporary” setting becomes permanent. Nobody feels the pain until the bill is already locked in.
We don’t know Atlassian’s internal playbook. But we can infer what companies at that scale have to do if they want cloud costs to stay controlled: commitment leverage, high-signal visibility, disciplined defaults, and a boring operating rhythm.
Why cloud spend becomes dangerous at scale
At small scale, you can “save” with heroics: a cleanup sprint, a round of rightsizing, a one-time purge.
At scale, heroics don’t work—because the system keeps producing waste.
Three reasons cloud becomes a quiet threat:
- Usage is abstract. Engineers don’t feel the price tag in the moment.
- Waste is distributed. A little per team becomes a lot in aggregate.
- The bill is a lagging indicator. By the time finance sees it, the architecture decision is already made.
What companies like Atlassian likely do to control it
The winning pattern isn’t “cut costs.” It’s run a feedback loop that makes spend visible, attributable, and negotiated.
Atlassian-scale operators usually build four layers:
- Ownership: someone can say yes/no and change defaults.
- Commitment leverage: buy predictable usage cheaply.
- Visibility: costs mapped to teams/services/products.
- Architecture discipline: defaults and guardrails that prevent waste.
The role of commitment pricing / volume leverage
For large cloud buyers, the “list price” is a starting point. Real economics come from commitments: reserved capacity, savings plans, or negotiated enterprise discounts.
Two practical truths:
- Commitments reward predictability. The more stable your baseline usage, the more you can lock in lower rates.
- Commitments punish chaos. If teams spin workloads up and down with no pattern, you can’t safely commit—and you pay the tax.
This is why “cloud cost control” is partly a finance negotiation and partly an engineering discipline problem. You need baseline stability to buy the baseline cheaply.
Why visibility matters more than panic cost cutting
Companies burn money when cloud cost becomes a surprise. Surprise turns into panic. Panic turns into blanket mandates that teams route around.
The better move is boring visibility with real accountability:
- Tagging that actually sticks (team, service, env, owner)
- Cost per unit where it matters (per customer, per workspace, per request—pick one)
- Alerts on anomalies (not on normal growth)
Visibility isn’t for blame. It’s for decision speed. When you can answer “what changed?” in minutes, you don’t need a cost-cutting theater campaign.
Architecture discipline and avoiding waste
Waste isn’t just “idle instances.” It’s choices that quietly multiply cost:
- Always-on environments that should be scheduled or ephemeral
- Overprovisioned defaults (big instances because it’s easier)
- Data retention sprawl (logs, snapshots, backups that never expire)
- Network and data transfer surprises that nobody owns
The operator fix is to make the cheap thing the easy thing. Defaults beat policies every time: standard sizes, standard retention, standard patterns, and guardrails that don’t require permission.
What smaller businesses can learn from this
You don’t need a FinOps org chart to get leverage. You need three moves:
- Pick an owner. One person owns the bill and the questions.
- Pick a unit. One metric that connects spend to the business (even if it’s imperfect).
- Run a rhythm. A weekly 30-minute sweep beats a quarterly emergency.
Then, once you can predict your baseline, you can start playing the commitment game—and stop paying list price for stability you already have.
Closing summary
Cloud costs don’t get controlled by willpower. They get controlled by a system: leverage on the baseline, visibility where decisions happen, and engineering defaults that prevent waste.
If you only remember one thing: don’t wait for the bill to tell you what happened. Build the loop that tells you while it’s happening.