TL;DR

Microsoft unveiled Majorana 2, a quantum chip with qubits 1,000x more reliable than its predecessor, achieving a mean 20-second lifetime versus microseconds for competitors. Agentic AI via Microsoft Discovery accelerated the development, and Microsoft now targets a scalable quantum computer by 2029, halving its original timeline.

Microsoft has unveiled Majorana 2, a next-generation topological quantum chip whose qubits are 1,000 times more reliable than those in the first Majorana chip introduced last year. The improvement is so significant that Microsoft has cut its timeline for achieving a scalable quantum computer from 2033 to 2029, halving the original target. The company credits agentic AI, deployed through its Microsoft Discovery research platform, with accelerating the materials science, fabrication optimisation, and measurement automation that made the leap possible.

The numbers are striking. Majorana 2’s qubits maintain their quantum state for a mean lifetime of 20 seconds, with some instances lasting as long as one minute. Most competing quantum approaches measure qubit lifetimes in microseconds. Microsoft’s analogy: it is roughly comparable to a phone battery that lasts three years on a single charge instead of dying in a day. Combined with one-microsecond operations and a qubit size of 1/100th of a millimetre, the chip puts Microsoft on what it describes as a path to commercially valuable quantum computing by the end of the decade.

How agentic AI built a better chip

The key materials change was switching from aluminium to lead as the superconductor. Lead naturally shields qubits from cosmic disturbances that cause instability, but working with it introduced tradeoffs that took years to overcome. Quantum computing startups across Europe and the US are pursuing different approaches to the qubit stability problem, but Microsoft’s topological approach, which creates an entirely new state of matter, is architecturally distinct from the superconducting circuits used by IBM, Google, and most competitors.

Microsoft Discovery’s AI agents were deployed across the quantum team’s workflow in several ways. Agents automated the measurement process that previously took weeks when done manually, cutting cycle time by orders of magnitude. They analysed nearly two decades of experimental data across multiple formats and silos, finding correlations that no individual researcher could see across that volume. They optimised fabrication processes by running simulations to identify the most promising material compositions before physical experimentation. And they detected an uncalibrated temperature sensor that was introducing noise into the fabrication process, a flaw that had gone unnoticed by human review.

“Agentic AI has permeated almost everything we do,” said Chetan Nayak, Microsoft technical fellow. The application of AI to quantum hardware development represents a convergence that could accelerate the entire field: better AI helps build better quantum computers, which in turn could eventually run better AI.

Microsoft Discovery goes public

Alongside the Majorana 2 announcement, Microsoft made its Discovery platform generally available. The platform lets organisations deploy autonomous AI agent teams, guided by human expertise, to speed scientific research and development. It includes a Discovery Engine for research and reasoning workflows, enterprise-grade security and governance, and integration with Azure. Google, Anthropic, and OpenAI are all pursuing AI for science, but Microsoft is the first to ship a commercially available platform specifically designed for frontier R&D with built-in agent orchestration.

Microsoft also introduced a free Discovery app in early preview that individuals can download and run locally with a GitHub Copilot account. Customers including chemical company Syensqo are already using the platform to develop next-generation fluids for semiconductor manufacturing.

The competitive context

The quantum computing sector is experiencing a funding and IPO boom. Quantinuum’s massively oversubscribed IPO this week valued the Honeywell-backed company at $14.3 billion. The US government committed $2 billion to quantum firms in May, with IBM receiving $1 billion for its Anderon quantum chip foundry. Focused Energy raised $240 million for laser fusion. The market is pricing in the expectation that quantum will follow AI’s trajectory from laboratory curiosity to commercial capability within this decade.

Microsoft’s topological approach has been the most controversial in the field. The company’s 2018 claim to have observed Majorana zero modes was retracted after independent scrutiny. Majorana 1, introduced in 2025, re-established credibility with peer-reviewed results. Majorana 2’s 1,000x improvement and the accelerated 2029 timeline will face similar scrutiny, and the peer-reviewed paper accompanying the announcement will be the definitive test of whether the results hold up.

The energy and compute demands of AI make quantum computing’s potential more commercially relevant than at any point in its history. If Microsoft can deliver a scalable topological quantum computer by 2029, the applications in drug discovery, materials science, cryptography, and optimisation would be transformative. If it cannot, the 2029 target will join a long list of quantum computing timelines that proved optimistic. The difference this time is that AI is accelerating the research itself.

The battle for compact high-performance desktops is heating up, and Nvidia appears ready to enter territory long dominated by Apple’s Mac Studio. At Computex 2026, MSI unveiled a new AI-focused mini PC called the MSI EdgeMesa N AI, powered by Nvidia’s brand-new RTX Spark platform.

The launch signals Nvidia’s growing ambition to push AI computing beyond traditional gaming desktops and into compact creator and workstation machines. More importantly, it also shows PC brands moving aggressively toward Apple’s increasingly successful formula of powerful desktop performance inside small, minimalist systems.

A tiny AI workstation built around Nvidia’s new RTX Spark platform

MSI’s EdgeMesa N AI is one of the first mini PCs announced using Nvidia’s new RTX Spark chip architecture. The system is designed specifically for AI workloads, local generative AI applications, creative software acceleration, and edge computing tasks.

While MSI has not fully disclosed every hardware detail yet, the company confirmed the mini PC combines Nvidia RTX Spark graphics with Intel-based processing hardware inside a compact chassis aimed at creators, developers, and AI-focused users.

The system is being positioned less like a traditional gaming PC and more like a local AI workstation capable of handling generative AI models, accelerated creative tasks, and productivity workloads directly on-device. That positioning immediately invites comparisons to Apple’s Mac Studio, which has become increasingly popular among creators, video editors, and developers looking for desktop-class performance in smaller form factors.

MSI says the EdgeMesa N AI is designed for local AI inference, AI-assisted workflows, content creation, and advanced multitasking scenarios that traditionally required much larger desktop systems.

MSI is not alone either. Other PC manufacturers, including ASUS, are also expected to adopt Nvidia’s RTX Spark platform for their own compact AI-focused desktops. ASUS is also pushing the RTX Spark platform far beyond laptops with its new ProArt Mini PC, a compact workstation measuring just 150 × 150 × 51mm. Despite the small footprint, the system supports up to 128GB unified LPDDR5X memory, delivers up to 1 petaflop of AI performance, and uses Nvidia’s 20-core Grace CPU paired with a Blackwell RTX GPU featuring 6,144 CUDA cores.

ASUS says the mini PC can handle 90GB+ 3D scenes, 120B-parameter large language models with up to one million tokens of context, and AI-assisted creative workloads locally. It also includes 10GbE networking, PCIe Gen 5 x4 storage expansion, and a thermal solution rated for up to 140W sustained workloads.

Why this matters

For years, Apple largely dominated the premium compact workstation category with devices like the Mac Studio and Mac mini. Now, Nvidia, alongside major PC brands, appears ready to challenge that space directly. The RTX Spark platform represents Nvidia’s attempt to create a standardized AI-focused desktop ecosystem for Windows PCs, particularly as AI workloads become more important for creators, developers, researchers, and businesses.

The shift also highlights a much larger industry transition happening right now. AI acceleration is rapidly becoming just as important as traditional CPU and GPU performance in next-generation PCs.

What happens next

MSI has not yet confirmed pricing or final availability details for the EdgeMesa N AI. However, the company is expected to reveal more specifications and launch timelines later this year. ASUS is also preparing its own RTX Spark-powered ProArt Mini PC lineup, which takes the concept even further with up to 128GB unified LPDDR5X memory, Nvidia Blackwell-based graphics, 10GbE networking, PCIe Gen 5 storage support, and claimed AI performance reaching 1 petaflop in an ultra-compact chassis.

As more manufacturers adopt Nvidia’s RTX Spark platform, compact AI desktops could quickly become one of the biggest new hardware categories emerging after the generative AI boom. Instead of massive workstation towers, creators and developers may soon have access to AI-focused machines small enough to sit beside a monitor while still handling local LLMs, advanced rendering, and accelerated AI workflows.

The bigger question is whether Windows-based AI mini PCs from brands like MSI and ASUS can truly compete with Apple’s ecosystem advantage and silicon efficiency. But one thing is becoming increasingly clear: the fight for the future of desktop computing is no longer just about raw performance. It is increasingly about who can build the smartest AI workstation in the smallest possible box.