Cloud cost optimization for Ottawa companies: AI is undoing five years of progress

Every cloud invoice contains a version of the same line item: capacity that nobody is using. It hums along at a few percent utilization, lands on the bill every month, and survives because no single person owns it. For years the industry had this problem under control — waste percentages were falling, steadily, for five straight years. That streak just ended. For leadership teams evaluating cloud cost optimization for Ottawa companies, the 2026 data carries an uncomfortable message: the leak was sealed, and AI reopened it wider.

The waste number is moving the wrong way

Flexera's 2026 State of the Cloud Report, the 15th annual edition and a survey of 753 cloud decision-makers, found that organizations now estimate 29% of their IaaS and PaaS spend is wasted — producing nothing useful. That is the first increase in five years, reversing a long downward trend. Managing cloud spend remains the report's top challenge, cited by 85% of respondents, ahead of security. The scale matters as much as the direction: 76% of large enterprises now spend more than $5 million a month on public cloud, so a single point of waste percentage is a material budget line before anyone notices it.

Put a dollar figure on it and the picture sharpens. A Harness "FinOps in Focus" analysis projected roughly $44.5 billion in cloud infrastructure waste for 2025 — about 21% of enterprise cloud infrastructure spend on a narrower definition than Flexera's. Two details from the reporting should concern any mid-market executive. First, idle is not cheap: a server doing nothing still draws around 60% of its peak power, so an unused machine costs real money every hour it stays on. Second, the spread under the average is wide — most organizations report 28 to 35% waste, while mature programs sustain roughly 15 to 20%. That gap is not structural. It is governance, and some teams have already closed it.

AI turned a manageable leak into a flood

Why did five years of progress reverse in one? The same report points at AI. Generative AI use jumped to 81% of organizations — from 72% last year and 47% in 2024 — and AI workloads are surging onto cloud bills faster than finance teams can tag or govern them. At Gartner's IT Symposium this month, analysts reported that IT executives are spending 18 times more on AI than they did two years ago, a period the firm's analysts called "AI inflation." The squeeze is precise: in Gartner's CIO survey, 58% of CIOs reported increased pressure to reduce costs by using AI, while 51% expect AI to increase technology's total costs. Cut with AI, spend because of AI — at the same time.

AI waste is also harder to see than the old kind. The classic culprit — a virtual machine sized for peak demand that never arrived — at least looked idle on a dashboard. An inference endpoint built for a pilot, then abandoned when a better model shipped, can appear active in a monitoring view while serving almost no traffic. GPU clusters and half-finished model experiments get launched faster than they get shut down, and the monthly invoice quietly absorbs all of it. The FinOps Foundation's 2026 State of FinOps survey — 1,192 practitioners representing more than $83 billion in annual cloud spend — found that 98% of FinOps teams now manage AI spend in some form, up from 31% just two years ago. In two years, AI went from a rounding error to a line item on nearly every cost-management agenda in the industry.

Why dashboards alone will not fix it

If the tools exist, why does waste persist? Because the gap is organizational, not technical. In the engineering-leader research cited by GCN, 52% of engineering leaders said the disconnect between FinOps and development teams is the primary driver of wasted spend. Fewer than half of developers have real-time visibility into idle cloud resources, and only 33% can see whether their workloads are over- or under-provisioned. The people creating the cost are the last to see it — usually weeks later, on an invoice nobody on the engineering side reads.

The FinOps Foundation's data adds a second warning: optimization has hit diminishing returns. Practitioners report they have cleared the "big rocks" of waste and now face a high volume of smaller opportunities that cost more effort to capture. The discipline is shifting left — moving cost awareness earlier, into architecture and design decisions — and the single most requested new tool capability in the 2026 survey was pre-deployment architecture costing: knowing what a design will cost before anyone deploys it. Waiting for the invoice is officially a losing strategy.

What cloud cost optimization for Ottawa companies looks like in practice

Ottawa's mix — federal suppliers with compliance obligations, a deep tech sector, professional services firms modernizing their platforms — makes this acute: hybrid environments (73% of organizations run them, per Flexera) multiply the places cost can hide. The playbook that works, whether your platform team sits in Ottawa, Toronto or Edmonton:

• Allocate before you cut. Tag every workload to a business owner and publish a monthly showback by team. Waste thrives on anonymity; nothing drops idle resources faster than the team that created them seeing the number.
• Treat AI spend as cloud spend. Tokens, model subscriptions and GPU capacity belong in the same dashboards and the same governance as compute and storage. If AI costs are invisible to finance today, that is the first project — not a rightsizing exercise.
• Hunt idle AI infrastructure specifically. Pilot-era inference endpoints, orphaned experiments and unattached storage volumes are the new ghost instances. An hour of decommissioning review usually pays for itself in the first month.
• Make cost a design input. Ask what an architecture will cost before it ships — per transaction, per customer, per document processed. Unit economics catch over-provisioning that percentage-of-bill views miss.
• Fund AI from the savings. The FinOps Foundation found organizations are increasingly being asked to self-fund AI investments through optimization savings. That only works if the savings are actually harvested and routed — otherwise AI spend just stacks on top of waste.

Enriched Code Inc. is a Montreal-based custom-software consultancy that helps mid-size companies across Canada — from Ottawa and Winnipeg to Vancouver — get cloud and AI spending under control: cost allocation and showback that engineering will actually use, data platforms that make unit economics measurable, and the custom tooling that connects cloud bills to business outcomes. We work remotely with clients coast to coast and on the ground in Quebec. If your cloud invoice grew faster than your business did this year, talk to us — the first conversation is an assessment, not a quote.