Infrastructure Meets Intelligence: Google’s Bid to Become the Operating System for AI Agents
PLUS: Cursor acquires AI code review startup Graphite
Venture Radar
Infrastructure Meets Intelligence: Google’s Bid to Become the Operating System for AI Agents
Google’s launch of managed MCP servers marks a quiet but decisive shift in the economics of artificial intelligence. For the past few years, power in AI has oscillated between model builders who control reasoning capability and tool providers who control real-world execution. In an agent-driven future, that separation becomes friction. Google is collapsing the model–tool divide by offering managed MCP servers that allow AI agents to plug directly into its ecosystem with minimal setup.
This move gives Google structural control over how agents interact with the digital world: search, maps, email, cloud, documents, payments, and enterprise workflows. Instead of developers manually wiring APIs, permissions, and orchestration layers, agents can now connect natively through standardized MCP interfaces. The result is not just easier integration, but tighter coupling between intelligence and action. Google is positioning its tools not as optional plugins, but as the default execution layer for autonomous systems.
A managed MCP layer turns tooling into platform sovereignty. By abstracting away infrastructure complexity, Google shifts the locus of value upward—from individual integrations to systemic dependency. Every agent action becomes mediated through Google’s managed environment, allowing the company to optimize reliability, security, identity, and performance at scale. Just as cloud platforms standardized compute, MCP servers standardize agency. Developers stop building glue code and start building behaviors.
The implications are broad. As AI agents move from demos to production, reliability, permissions, and auditability matter more than raw model performance. Google’s approach sets expectations for how agents should be deployed in regulated enterprises, consumer applications, and multi-agent systems. Competitors offering “bring-your-own-tools” flexibility may find themselves at a disadvantage when enterprises prefer managed, compliant, and deeply integrated stacks.
There are risks. Centralizing agent infrastructure concentrates power and raises questions about lock-in, neutrality, and long-term ecosystem openness. If MCP standards evolve in ways that favor first-party tools, third-party innovation could be constrained. And as with any platform bet, failures at the infrastructure layer propagate widely when agents depend on it for execution.
Still, the strategic direction is clear. The next phase of AI will not be won solely by who has the smartest models, but by who controls how intelligence touches the real world. By managing the connective tissue between agents and tools, Google is signaling its ambition to become not just an AI provider, but the operating system through which autonomous software thinks, acts, and scales.
Geeks of the Week
Startup Name: devlo
Geography: US
One-liner: AI Engineering Productivity Platform, where you can track and improve your team’s productivity.
Founder(s) Background: Senior Software Engineer (AI) at Google.
Thoughts:
Devlo is positioning itself as an AI-native engineering teammate that streamlines the full software development lifecycle - moving beyond code completion to automate pull requests, testing, bug fixes, and deployment workflows, allowing lean engineering teams to ship production-ready software faster without adding headcount
Startup Name: July AI
Geography: US
One-liner: July AI is the first institutional platform to teach AI red teaming.
Founder(s) Background: Director (New Verticals) at DoorDash, BizOps Lead at Adept.
Thoughts:
July AI is positioning itself at the intersection of AI safety, workforce development, and institutional demand by formalizing AI red teaming as a teachable and scalable discipline. Unlike traditional security training or prompt engineering courses, July targets a nascent but increasingly critical function - stress-testing AI systems for failure modes, misuse, and alignment risks - and aims to build both human capital (trained red teamers) and infrastructure outputs (datasets, evaluation artifacts, campaign support). If executed well, this creates a defensible two-sided model where education feeds directly into enterprise and government demand for AI safety validation
Startup Name: Eraser
Geography: US
One-liner: AI for technical design and documentation.
Founder(s) Background: Chief of Staff for Elad Gill, Private Equity Associate at Oak Hill Capital.
Thoughts:
Eraser is targeting a structurally under-served but high-frequency workflow in software engineering: translating system intent into clear, consistent technical diagrams and documentation. While most developer tools optimize for writing code, Eraser focuses on the design layer - enabling engineers to generate architecture diagrams, infrastructure flows, and technical documentation using AI-assisted prompts and “diagram-as-code” primitives. This positions Eraser as a productivity and alignment tool rather than a coding assistant, reducing design drift between teams, speeding up design reviews, and improving knowledge transfer across organizations.
Founder(s) building in stealth
Notable Mergers and Acquisitions
Cursor acquires AI code review startup Graphite
Cursor, an AI-powered code editor platform, announced the acquisition of Graphite, a New York–based code review startup, on 19 December 2025. Graphite, previously valued at around $290 million, will continue operating as an independent product under Cursor while gaining access to expanded engineering and go-to-market resources.
Flipkart acquires majority stake in AI solutions startup Minivet
Indian e-commerce giant Flipkart agreed to acquire a majority stake in Minivet, an AI/ML solutions startup founded in 2024, in a deal announced in mid-December 2025. Minivet’s generative AI technologies are expected to enhance Flipkart’s platform capabilities, powering conversational, immersive, and visual shopping experiences
This edition is brought to you in partnership with Stella Capital.