MicrosoftがNVIDIA RTX SparkをローカルAI推進の突破口と位置づけ、Build 2025で詳細発表へ Our goal is to deliver unmetered intelligence to every home and every desk with Windows. NVIDIA RTX Spark marks a real breakthrough toward that vision. Looking forward to sharing more with Jensen, who will be joining us live from Taiwan, at Build this week!

Microsoft CEO Satya Nadella took to LinkedIn ahead of Microsoft Build to frame NVIDIA RTX Spark as a pivotal milestone in the company's broader ambition: delivering what he calls "unmetered intelligence" to every home and every desk running Windows. The phrasing is deliberate — it evokes a future where AI is not rationed by API tokens, cloud subscriptions, or network connectivity, but runs freely and continuously on local hardware.

NVIDIA RTX Spark is a compact, palm-sized AI desktop computer powered by hardware derived from the GeForce RTX 50 series. It is designed to run large language models and other AI workloads entirely on-device, without routing requests through remote data centers. For end users, that translates to lower latency, enhanced privacy, and AI availability even in offline or bandwidth-constrained environments. In the context of Microsoft's Copilot+ PC initiative — which already requires a Neural Processing Unit capable of 40+ TOPS — RTX Spark represents a significant step up in raw local inference capability.

Microsoft Build is the company's flagship developer conference, and this year's edition is expected to center heavily on AI agents, local inference, and the evolving Windows platform. Nadella's post teased that NVIDIA CEO Jensen Huang would join the event live from Taiwan, a detail that carries both symbolic and practical weight. Huang has increasingly spotlighted Taiwan as a hub for NVIDIA's manufacturing and research operations, and his remote appearance from there aligns with a broader narrative about the geopolitics of semiconductor production.

The concept of unmetered, local AI is not unique to Microsoft. Apple has pursued a similar path with Apple Intelligence, emphasizing on-device processing and Private Cloud Compute for tasks that exceed local capacity. Qualcomm has made Snapdragon X Elite's NPU performance a central selling point for Windows on Arm laptops. What distinguishes the Microsoft-NVIDIA angle is the sheer GPU horsepower being brought to bear: where most Copilot+ PCs rely on relatively modest NPUs, RTX Spark brings data center-class GPU throughput into a consumer form factor.

The practical implications for developers could be substantial. Running models like Phi-3, Llama, or Mistral locally without cloud round-trips opens new possibilities for latency-sensitive applications — real-time coding assistants, offline document analysis, or privacy-first enterprise tooling. If Microsoft integrates RTX Spark deeply into its developer ecosystem and Windows AI APIs, it could meaningfully lower the barrier to building capable AI-native applications.

That said, several open questions remain. Pricing, availability, and the breadth of software support will determine how widely RTX Spark actually penetrates the market beyond enthusiasts and enterprise early adopters. The promise of unmetered intelligence is compelling, but delivering it at scale — across diverse hardware configurations and user skill levels — is a far more complex engineering and distribution challenge than any single product launch can resolve. Build 2025 is likely to answer at least some of those questions, and the Nadella-Huang pairing on stage (or screen) signals that both companies are treating this as a major strategic moment.