Microsoft is simultaneously betting on the AI internet, quantum computing, and local AI assistants.
At its annual Build developer conference held on Tuesday, May 2nd (Eastern Time), Microsoft unveiled a series of major products and technologies: it launched the search infrastructure Web IQ for the AI agent era; released the new-generation quantum chip Majorana 2, significantly advancing the target date for a commercially viable quantum computer to 2029; and simultaneously announced the new AI assistant Scout and a toolchain enabling developers to deploy "Lobster" (OpenClaw)-style autonomous agent systems on Windows PCs.
Nvidia also played a central role on the Microsoft Build stage.
Nvidia CEO Jensen Huang joined Microsoft CEO Satya Nadella's keynote via livestream, where they showcased a unified agent AI technology stack spanning Windows devices, local computing, edge deployment, and Azure cloud, including a new generation of local AI development devices, Windows local model runtime capabilities, and new tools for developers to scale agent deployments directly from PCs to the cloud.
From search, models, operating systems, quantum hardware to AI computing platforms, the Build conference sent a clear signal: Microsoft is no longer content with being an AI application company but is building the full-stack ecosystem for the next generation of intelligent computing.
A review of this year's Build announcements reveals a significant shift: Microsoft's strategic landscape has clearly expanded beyond chatbots and Copilot.
Microsoft is now concurrently advancing: AI models (MAI series), the Agent internet (Web IQ), agent operating systems and local AI (Scout, Windows Agent toolchain), quantum computing hardware (Majorana 2 chip), and the Azure cloud and enterprise AI ecosystem.
In other words, Microsoft is attempting to simultaneously occupy several critical entry points for future intelligent computing: the search entry, the agent entry, the PC entry, the cloud entry, and the potential quantum computing entry.
What Build showcased was not merely more AI features.
It more closely resembled a comprehensive bet by Microsoft on the next computing platform.
Core Focus: Web IQ
Microsoft introduced Web IQ, a search engine designed for AI systems.
The Web IQ suite includes a series of AI-native Grounding APIs.
Microsoft defines Web IQ as a "network intelligence layer for AI agents," aiming not to serve traditional web search users but to directly provide AI agents, Copilots, and automated agents with real-time, trustworthy, structured internet information.
Simply put, what Microsoft wants to build is no longer just Bing for human user searches, but Bing for AI agent searches.
According to Microsoft's official blog, Web IQ is essentially a service layer that provides AI agents with real-time internet knowledge access capabilities.
Microsoft emphasizes that Web IQ is not a traditional web search but infrastructure to "help AI systems understand, reason, and act."
Microsoft stated that Web IQ enables AI agents to obtain the latest, trustworthy, structured information and execute complex tasks based on it.
This positioning means the primary users of Web IQ are no longer ordinary people but AI agents themselves.
In the era of traditional search, users input keywords and search engines return links; in the agent era, AI needs to directly read, understand, and invoke internet information to accomplish tasks.
For example: an AI travel agent automatically comparing hotels and flights; an AI procurement agent automatically screening suppliers; an AI office assistant automatically finding, summarizing, and executing workflows; an AI developer agent automatically reading documentation, calling APIs, and generating code.
What are Web IQ's capabilities?
Based on information disclosed, Web IQ possesses several key capabilities.
First, it can access real-time internet information.
Microsoft emphasizes that Web IQ, based on the Bing indexing system, can provide AI agents with real-time updated internet information, not just relying on training data.
This means agents can obtain: the latest news; real-time prices; dynamic inventory; real-time webpage content; the most recent API documentation; current company information.
Microsoft particularly emphasizes its grounding capability, helping AI reduce hallucination issues.
Microsoft has previously stated multiple times that Bing's real-time indexing capability is one of the important moats for its AI ecosystem.
Second, it allows agents to directly invoke structured web capabilities.
Microsoft states that Web IQ does not simply return webpages but will provide AI agents with "actionable information structures."
This means that in the future, AI will not only "know the answer" but also be able to: call website services; automatically complete transactions; understand page semantics; operate online tools; collaborate with external agents.
This aligns closely with Microsoft's previously promoted MCP (Model Context Protocol) strategy.
Microsoft is pushing the internet to evolve from "browsers reading webpages" to "AI agents reading services."
Third, it supports a multi-agent collaboration ecosystem.
Microsoft repeatedly emphasized "multi-agent systems" at Build.
One of the significant implications of Web IQ is enabling multiple agents to share internet context.
For example: one agent responsible for search; one for reasoning; one for execution; one for verification.
Microsoft believes such team collaboration among agents will become a core architecture for next-generation AI applications.
Quantum Computing Leap
Beyond the AI software ecosystem, another major announcement at this year's Build came from quantum computing.
Microsoft officially launched the new-generation quantum chip, Majorana 2.
This is the follow-up version to last year's controversial Majorana project and represents the latest achievement in Microsoft's push for the "topological qubits" roadmap.
Unlike the mainstream quantum approaches of Google and IBM, Microsoft has long bet on a more radical path: using Majorana quasiparticles to construct more stable, lower-error-rate qubits.
According to Microsoft's disclosed data, the new Majorana 2 features several key improvements: the number of qubits has increased to 12, up from 8 in the previous generation; the lifetime of a single qubit exceeds 20 seconds, compared to less than 12 milliseconds for the previous generation; and the chip material architecture has been adjusted, introducing lead-based superconducting materials to replace some of the old design.
In quantum computing, qubit stability has always been the core bottleneck determining whether the technology can move towards commercialization.
Microsoft believes topological qubits are inherently more noise-resistant and less error-prone, making them more suitable for future scalable quantum systems.
Due to recent progress being faster than expected, Microsoft has halved its original timeline, now aiming to achieve a scalable, practical quantum computer before 2029.
Chetan Nayak, a Microsoft executive responsible for quantum hardware, stated that based on rapid advances in quantum chips, they are accelerating the roadmap to a scalable, practical quantum computer, having halved the timeline with the goal of achieving it by 2029.
However, the Majorana project remains controversial.
Following the release of the first-generation chip last year, some quantum researchers questioned whether Microsoft had sufficiently demonstrated its technological breakthrough, and some early related research had faced retraction controversies.
Microsoft emphasizes that its project is under continuous review by the U.S. DARPA team and that detailed data is open to partner evaluators.
New AI Assistant and the "Lobster" Connection
Another notable move at this year's Build was the introduction of the new personal AI assistant, Scout.
In terms of positioning, Scout clearly reflects Microsoft's bet on the "local agent" direction.
Unlike traditional chatbots, Scout is closer to a personal digital assistant capable of autonomous action.
Its capabilities include: browsing and understanding the local environment; invoking tools; executing multi-step tasks; long-chain planning; and running continuously on the device.
Notably, Scout is widely considered to be inspired by the recently popular open-source project—known domestically as "Lobster" (OpenClaw).
The core concept of the "Lobster" model is to allow AI agents to directly operate the operating system like a real computer user, rather than being confined to answering questions within a chat box.
The toolchain released by Microsoft further pushes this capability into the Windows ecosystem.
It is believed that the introduction of Scout shows Microsoft is rolling out new development capabilities to help developers deploy autonomous AI systems similar to "Lobster" on Windows PCs, enabling agents to run directly in the local computer environment.
This means that AI on future Windows devices may not just answer questions but also be able to: operate software; automatically complete workflows; use a browser; manage files; execute complex tasks.
To some extent, this is also an early layout by Microsoft for the "agent PC" era.
Nvidia's Showcase of a Unified AI Stack
Beyond Microsoft's own products, Nvidia's presence at Build was also extremely strong this year.
Jensen Huang, joining from Taipei via livestream in a joint demonstration with Microsoft CEO Satya Nadella, highlighted a unified AI deployment system the two companies are building together—allowing developers to seamlessly migrate Agent workloads between Windows computers, local devices, edge environments, and Azure cloud.
Nvidia stated that Huang and Nadella discussed the expanded partnership between the two companies, covering specific content including: Nvidia's PC chip RTX Spark and desktop AI supercomputer DGX Station for Windows, Nvidia open-source models on the Microsoft data analytics platform Microsoft Fabric and the Microsoft AI and agent development platform Microsoft Foundry accelerated by Nvidia GPUs, the Nvidia agent OpenShell secure runtime in GitHub Copilot, and the next-generation AI factory powered by Nvidia.
Nvidia's core logic is that future AI development should not be limited to the cloud.
Developers need a unified software and hardware stack that allows models to: run locally on Windows PCs; perform device-side inference during the development phase; deploy to edge devices; and then seamlessly scale to Azure cloud for training and production environments.
A key component showcased was the Surface RTX Spark Dev Box and RTX Spark platform, jointly advanced by Microsoft and Nvidia.
This platform, based on Nvidia's AI chip architecture, provides developers with high-performance local AI computing power, supporting the operation of very large models on personal devices.
Some devices can run models with up to 120 billion parameters, reducing developers' reliance on continuous cloud calls.
Simultaneously, Nvidia emphasized its work with Microsoft on a unified agent software stack, including: a Windows local AI runtime environment; Azure AI Foundry-compatible deployment; the Nvidia AI Enterprise ecosystem; NIM microservices and model toolchain; and a consistent local-to-cloud agent development workflow.
Additional Major Announcements
Beyond Web IQ, Microsoft's Build conference this year also announced several agent infrastructure items.
Project Solara is an agent device platform.
Microsoft introduced Project Solara, aiming to build a new type of device platform "centered on agents."
According to related reports, Microsoft is exploring a device interaction method "without traditional apps," where users in the future would complete operations more through AI agents rather than opening specific applications.
This direction aligns with Microsoft's recent strategy of deeply embedding Copilot into Windows.
Microsoft also released the MXC (Microsoft eXecution Container) sandbox system.
This is an operating system-level security container for AI agents, allowing agents to execute code, invoke tools, and access files within a controlled environment.
Microsoft emphasizes that as AI agents become increasingly capable of autonomous execution, security isolation mechanisms will become crucial.
It is reported that OpenAI and Nvidia are participating in the related ecosystem.
Microsoft also heavily promoted the Microsoft Agent Framework at Build.
This framework is designed to help developers build: multi-agent systems; agent workflows; agent orchestration; agent observability; and agent governance systems.
Microsoft clearly hopes to compete in the agent development framework space with ecosystems like LangChain, AutoGen, and Anthropic MCP.
Microsoft also announced further enhancements to the Windows AI development environment in Windows 11.
The enhancements are specifically reflected in: improved Linux development experience; AI inference optimization; local model runtime capabilities; and integration of agent development toolchains.
Microsoft is attempting to reforge Windows into the "preferred platform for AI development."
Comments