FinOps Summer Camp started off strong with a look at where cloud and AI cost management stands in 2026. Jean Atelsek, a senior analyst at 451 Research by S&P Global, and Ben Schaechter, CEO and co-founder of Vantage, walked through new data from IT decision-makers and FinOps practitioners. Jean's summary of the year so far is that the cost management discipline is in flux. Here are the four insights behind that statement.

Insight 1: First-party cost management tools still top the list, but vendor-agnostic/multi-vendor alternatives are gaining

First-party cost tools from AWS, Google Cloud, and Azure are still the most widely used. They sit inside the cloud environment and are often free. But between 2025 and 2026, every alternative gained ground: commercial platforms, open source, even spreadsheets.

Jean's analysis is that teams want one vendor-agnostic view across cloud usage, software spend on Databricks and Snowflake, model usage, and anything else that can be pulled in. That beats running a separate tool per vendor. This is a familiar pattern that she sees often: a team starts with a native tool, finds it too limited, tries to build its own, and eventually buys a more capable third party.

Ben added this take from the Vantage perspective. Native tool usage is falling even as bill scrutiny rises, which he attributes to the surge in AI tooling that native billing views were slow to cover.

Insight 2: FinOps integration with adjacent management tools (including observability) is growing across the board

FinOps is integrating with adjacent tools, including asset management, service management, and observability. The changes were small but consistent across every category.

Jean tied it to two forces. AI costs are now material enough to track everywhere, and people across every department use the AI tools driving those costs. The FinOps Foundation has seen the same thing. Teams that proved out cloud cost savings get asked to apply the same discipline to adjacent areas.

Ben pointed to pricing as evidence. Seat-based and contract-based models are rotating toward consumption and token-based pricing, and that spend comes from across the whole organization. He also flagged MCP, the model context protocol, as a lighter way to pull adjacent data into cost tooling, from revenue systems to HR org charts, so token consumption can be tied back to specific teams.

Insight 3: GenAI-related usage was the top cause of overspending in 2025

In the 2025 survey, GenAI usage was the top cause of companies overspending their budgets by more than 10%. Jean expects the 2026 answer to shift toward agent experimentation, which can burn tokens fast when it runs ungoverned.

One number stood out. At the FinOps X conference, practitioners with mature FinOps programs said they now spend about 80% of their time managing AI spend, after years focused on right-sizing and rate optimization. The upheaval even shows up in naming: the community launched a Tokenomics Foundation, and the conference is being renamed Tokonomicon.

Ben described where the pressure lands. AI gateways and routers have become a new source of black-box spend, governance features are only now reaching provider APIs, and token cost is shifting from COGS into R&D. On anomaly detection, prompted by an audience question, his point was speed: an engineer can run up token spend within an hour, so near real-time data matters more than the old daily billing lag.

Insight 4: AI is accelerating a need for Cost Management tooling

For the fourth insight, Ben shared Vantage's own numbers: annualized infrastructure cost tracked on the platform, plotted against each new provider integration.

That figure has roughly tripled since the start of the year, across about 30 providers. Ben called it a Cambrian explosion of AI providers, naming OpenAI and Anthropic alongside newer entrants like ElevenLabs, a voice model company already at a $600 million run rate. A class of AI-native companies now runs on stacks that barely existed two years ago, and for some the AI bill is the largest line in COGS.

The harder problem is scale. Where a cloud bill lists an EC2 instance, an AI bill can carry a line item for every inference request. Processing and attributing that volume is part of why large enterprises turn to third-party tooling, which loops back to the first insight. Ben said attribution today relies on customer invocation logs and provider request metadata, and pointed toward a system of record that syncs budgets and governance across every gateway.

The state of play, and what's next

Even though everything is changing faster than ever, we can all agree that cloud and AI cost management is being rewritten in real time, and visibility and governance for AI spend are the open frontier. Most teams are early.

FinOps Summer Camp runs through the summer. Session 2 features the Make.com platform team on how they instrument AI cost in production, Thursday, July 30 at 11am ET. Register for the rest of the series here.