Nvidia RTX Spark Targets PC AI at Computex; Copilot Drops Flat Pricing

Nvidia's RTX Spark GPU debuts at Computex as GitHub Copilot replaces flat subscriptions with token billing on June 1.

🗄️ Nvidia Unveils RTX Spark GPU at Computex to Run AI Inference on Personal Computers

Decoded: Nvidia CEO Jensen Huang introduced the RTX Spark GPU during a Computex keynote in Taipei on June 1 — Nvidia's first discrete GPU designed explicitly around on-device AI inference for personal computers, according to Reuters and Bloomberg. The RTX Spark is aimed at running AI models locally rather than routing queries to cloud servers, extending Nvidia's AI hardware strategy from data centers into consumer PC hardware. Microsoft Surface and Dell laptops are among the first devices planned around Nvidia's PC silicon, per Techmeme-reported pre-show briefings. The announcement follows Nvidia's N1X ARM SoC reveal at the same Computex event and marks a two-pronged push into the Windows PC platform: one chip for the CPU layer (N1X), one for GPU-accelerated AI inference (RTX Spark). (Reuters, Bloomberg, June 1, 2026)

Why it matters: Local AI inference is the next major battleground in the hardware cycle — and the RTX Spark puts Nvidia in direct competition with Apple's (AAPL) M-series Neural Engine and Qualcomm's Snapdragon X NPU for the AI PC workload that was previously owned by cloud APIs. For investors, local inference reduces per-query cloud costs for AI-intensive applications and positions Nvidia to capture consumer PC hardware revenue as AI agents shift from server-side to on-device execution. If model inference migrates substantially to edge hardware over the next two to three years, it structurally shifts revenue from cloud AI platforms — Microsoft Azure, Google Cloud, AWS — toward silicon vendors. The Computex timing, alongside Surface and Dell device commitments, signals that Nvidia's PC AI push has Microsoft's institutional backing at launch.

🛠 GitHub Copilot Drops Flat-Rate Pricing; Token-Based Billing Begins Today

Decoded: Microsoft's GitHub Copilot shifted from flat monthly subscriptions to token-based billing effective June 1, 2026, replacing the fixed-price model that made it the world's most widely adopted AI coding assistant. Under the new system, core code completion features and Next Edit Suggestions remain unlimited and unmetered. Agentic capabilities — Copilot Agent Mode, multi-file edits, and AI-generated code reviews — are now billed by token consumption. Copilot Pro subscribers ($10/month) receive a monthly credit allowance to cushion the transition through mid-2026, but heavy agentic users are reporting projected cost increases of 10x to 50x compared to the prior flat rate. (GitHub official changelog, June 1, 2026; The Indian Express, May 31, 2026)

Why it matters: The Copilot pricing shift is the most significant monetization change in Microsoft's AI product strategy since Copilot launched in 2021. By metering agentic AI features — the high-compute capabilities that enterprise development teams use most heavily — Microsoft converts Copilot from a developer subscription into a usage-based infrastructure billing line. For investors, this is a structural revenue expansion: instead of capping Copilot revenue at the $10 to $39 per month subscription tiers, Microsoft can now capture a proportional share of every token an AI agent consumes on behalf of a developer or enterprise team. Enterprise contracts are likely to see Copilot token charges compounding alongside existing seat fees and GitHub Advanced Security bundles. The immediate developer backlash signals that the flat-rate era of AI tooling is ending industry-wide — variable-cost AI infrastructure is becoming the standard billing model.

Stay decoded. See you tomorrow.

— The Get AI Decoded Team