The promise of AI coding agents is driving enterprise adoption, but managing usage and controlling costs remain key hurdles. Google is seeking to address those concerns by integrating Antigravity into Gemini Enterprise and equipping it with new budgeting and consumption controls.
These tools, which were not available as part of Antigravity’s earlier enterprise deployment model through what was formerly Vertex AI and is now the Gemini Enterprise Agent Platform, address a key management gap around billing flexibility and license management at scale, according to Google.
While the earlier model allowed enterprises to deploy Antigravity through their cloud projects and benefit from Google Cloud’s security and governance infrastructure, it did not provide the same centralized mechanisms to pool unused capacity across developers, set usage-related spending limits, or manage overages. That left enterprises having to manage AI coding consumption through a more fragmented, project- and cloud-consumption-based model, rather than through a unified subscription and usage-management layer, resulting in less flexibility to allocate unused capacity, monitor consumption across developer teams, and control costs as usage scaled.
In contrast, the newer tools, such as granular spend thresholds and pooled quotas, will allow administrators to set monthly project-level budget caps and share token capacity across teams, respectively, Google executives wrote in a blog post.
The other two features, namely overage enablement and usage metrics, they added, are aimed at helping enterprises manage usage once those limits are reached and better understand how developers are consuming AI resources.
While overage enablement allows administrators to continue developer workflows after pooled quotas are exhausted by opting into additional usage, subject to monthly spending caps, centralized usage metrics provide visibility into token consumption, API calls, and developer activity, the executives explained. Analysts agree that the new cost control tools could help enterprises and their administrators rein in AI spending.
This will be possible primarily because under Antigravity’s earlier cloud-consumption model, spending was attributed to a project, giving finance teams visibility into the total cost but not necessarily who was driving that spending or which tasks were responsible, said Amit Kumar Jena, AI development head at IT consulting firm Kanerika.
Those insights can now be accessed by administrators via the usage metrics tool, which not only helps them understand where teams and developers are using the most AI, but also plan for AI spend optimization via the other tools, such as planning caps for developers and teams via the granular spend thresholds tool, echoed Pareekh Jain, principal analyst at Pareekh Consulting.
Similarly, pooled quotas and overage enablement, Jain added, can help enterprises further optimize spending by allowing unused AI capacity to be shared across developers and preventing unexpected bills, respectively.
However, not all the granular controls that Jain sees as key to optimizing AI spending will be available immediately. Google said per-user and team-level controls in the granular spend thresholds tool are expected to roll out later this year, meaning enterprises will initially have to rely on project-level budget caps.
Even when those controls become available, however, not every enterprise may be able to access them directly.
Antigravity’s new consumption and management tools are currently being offered through Gemini Enterprise Standard, Plus, and Standard Emerging Market licenses, meaning enterprises that have deployed the coding agent through the Gemini Enterprise Agent Platform but do not already have a qualifying plan may need to purchase an additional subscription.
That creates a potential cost-benefit question for CIOs and administrators: whether paying for a Gemini Enterprise subscription to gain better control over AI consumption and reduce wasted capacity will ultimately deliver enough savings to justify the additional expense.
The additional expense is only justified for enterprises that have teams running autonomous, multi-turn coding agents, where one unsupervised task chain can generate outsized token costs, said Manoj Chandra Jha, principal analyst at Nord-IQ Research.
“For these enterprises, pooled usage and hard spend caps function as a financial backstop rather than pure added expense,” Jha said.
“However, for enterprises doing largely basic, autocomplete-style assistance, upgrading mainly to unlock governance features is unlikely to be cost-effective,” Jha added.
In addition to the new tools, Google said it was also expanding the AI coding agent beyond its own desktop environment into popular IDEs, including VS Code, Visual Studio, JetBrains, and Zed in preview.
The move, Jha said, is Google’s attempt to lower the friction of adopting Antigravity by bringing the AI coding agent into the development environments where developers already work, rather than requiring them to switch to a separate application.