AI spend doesn't always show up on AI vendor invoices. A significant portion can appear on cloud bills, and the amounts can change quickly. Larridin Spark data shows both.
The Short Answer
A large share of AI spend can run through cloud platforms like GCP Vertex and AWS Bedrock, where it’s easy to miss on AI budgets. Larridin’s own data shows that 59% of $269,353 in AI spend ($158,600) came through Vertex and Bedrock. In one customer, Bedrock went from $0 to the largest weekly line in a matter of weeks.
Key Wins
What’s a Spark Insight?
A Spark Insight is a data-backed finding Larridin surfaces automatically from how your teams actually use AI, flagging the patterns worth a leadership conversation.
Who This Matters To
CFO · CIO · Eng leaders
| If you’re asking… | The answer | Go deeper |
|---|---|---|
| Does AI spend show up on cloud invoices? | Yes, often. In Larridin’s own data, 59% of AI spend ran through GCP Vertex and AWS Bedrock. | The data |
| Why is cloud AI spend hard to track? | It’s mixed in with other cloud costs and often owned by engineering, not the team tracking AI budgets. | Why it happens |
| How fast can AI vendor mix change? | Within weeks. In one org, Bedrock spend went from $0 for six weeks to $10–11K per week. | The data |
| How do I find all my AI spend? | Combine AI vendor invoices with AI-specific line items from cloud providers into one view. | How to measure |
Most AI budgets track the obvious vendors: chat tools, coding assistants, copilots. But models accessed through cloud platforms land on the cloud invoice, where they mix with compute and storage. That spend can be the biggest part of the AI bill.
| What you’re struggling with | What you actually need | How Larridin helps | Proof from Spark data |
|---|---|---|---|
| “I don’t know our total AI spend.” | One view across vendors and cloud platforms | Pulls AI spend from vendor invoices and cloud providers together | Larridin’s data: 59% of AI spend via Vertex + Bedrock |
| “Cloud AI costs appeared out of nowhere.” | Early warning when a new source ramps up | Tracks weekly spend by source | One org: Bedrock $0 for 6 weeks, then $10–11K/week |
| “Spend shifted between vendors and nobody noticed.” | Trend data by vendor | Shows vendor mix over time | One org: Vertex peak $37K/week in mid-June, under $6K by mid-August |
| “We have AI costs in too many places.” | A consolidated list of cost sources | Maps every AI cost source in one place | Larridin’s data: 9 separate AI cost sources |
The question you’re asked: “What is our total AI spend?”
What to do: Include cloud AI line items in the AI budget. Ask for a single view across vendors and platforms.
For a broader look at financial oversight, see AI Monitoring for CFOs: Turn Untracked Spend Into Measurable ROI.
The question you’re asked: “Who owns cloud AI spend?”
What to do: Assign ownership for AI usage on cloud platforms and review it alongside other AI tools.
For more on shared IT and finance accountability, see Why CIO and CFO Cannot Agree on AI Accountability.
The question you’re asked: “Do I need to report model usage on our cloud account?”
What to do: Yes. Tag AI workloads so they can be separated from other cloud costs and reported with AI spend.
For a broader look at measuring engineering AI investments, see AI Monitoring for Engineering Leaders.
Because teams increasingly access models through cloud platforms, and that usage is billed with other cloud services.
Platforms like GCP Vertex and AWS Bedrock let engineering teams use models inside the cloud environment they already run because it’s convenient. That means the cost lands on the cloud bill rather than an AI vendor invoice.
Myth: Your AI vendor invoices show your AI spend.
Reality: In Larridin’s own data, more than half of AI spend came through cloud platforms.
Within weeks.
In one org, GCP Vertex spend peaked at $37,000 a week in mid-June and fell below $6,000 by mid-August. Over the same 12-week period, AWS Bedrock went from $0 for six weeks to the largest weekly AI cost source, at $10,000-$11,000.
Spark callout: A cloud AI line item that didn’t exist in the first six weeks of the quarter became the largest weekly AI cost by the end of it.
Treat it as AI spend, track it weekly, and give it an owner.
Tagging AI workloads on cloud accounts makes the spend visible. Reviewing it weekly with other AI costs catches shifts early. Assigning ownership ensures someone is responsible for decisions that can move tens of thousands of dollars a week.
Google Cloud's Vertex AI (now Gemini Enterprise Agent Platform) and Amazon Bedrock let teams access AI models through cloud platforms. Usage charges appear on the cloud provider's invoice.
Cloud costs are often owned by engineering or infrastructure, while AI budgets track named tools. AI usage on cloud platforms falls between the two.
Weekly, given how quickly the mix can change. Monthly reviews can miss a ramp until it’s already the largest cost.
See AI spend across every vendor and cloud platform in one view, and bring it to your next finance review.