Nvidia just started a new chip war

What It Covers

Nvidia launches the RTX Spark, a new ARM-based system-on-chip targeting mainstream Windows laptops, with 20 CPU cores, up to 128GB unified memory, and roughly RTX 5070-mobile GPU performance. The chip signals Nvidia's strategic push to own local AI computing alongside its dominant cloud AI position.

Key Questions Answered

• •ARM market shift: Nvidia's RTX Spark entry means ARM-based CPU manufacturers now outnumber x86 vendors — Apple, Qualcomm, and Nvidia versus Intel and AMD. Consumers evaluating new laptops should recognize that ARM-based Windows machines are no longer niche; 30+ RTX Spark laptops and 10 mini desktops are already in development for near-term release.
• •Chip architecture tradeoff: The RTX Spark uses a monolithic single-chip design manufactured by TSMC on a 3nm process with MediaTek as hardware partner. This delivers thin, cool, efficient laptops but eliminates discrete GPU support, capping graphics at roughly RTX 5070-mobile tier — below what a dedicated gaming laptop with a separate RTX 5090 mobile chip provides.
• •Strategic hedging against cloud dependency: Nvidia's move into device-level silicon is partly a hedge against scenarios where cloud AI providers — Google, Meta, Microsoft, Amazon — back competing AI chip suppliers or where consumers demand local AI processing for privacy reasons. Being inside the device protects Nvidia if server-side AI dominance erodes, mirroring Intel's costly absence from smartphones.
• •Developer ecosystem leverage: Nvidia secured launch partners including Lenovo, HP, Dell, Asus, and Microsoft, which announced the Surface Laptop Ultra running RTX Spark. Major game titles including Valorant, PUBG, League of Legends, and Fortnite already committed ARM support, suggesting developers respond to Nvidia's market weight before a user base even exists — accelerating Windows on ARM viability.
• •Pricing expectations for early adopters: No pricing has been disclosed for RTX Spark laptops, but based on the existing DGX Spark AI box priced at $4,000–$5,000 using the same core specs, first-generation consumer devices will likely land in the $3,500–$5,000 range. Buyers seeking value should wait for second-generation SKUs targeting the 16GB RAM, lower-core-count configurations Nvidia has indicated are planned.