Nvidia’s Desktop AI Chip Arrives as Production Constraints Test Record Demand

Nvidia’s RTX Spark superchip brings 1 petaflop AI to laptops, but supply bottlenecks persist as revenue surges 85% to $81.6B. Stock up 6% on Computex buzz.

Jensen Huang used his Computex keynote in Taipei to unveil a direct assault on Intel and AMD’s core territory, but the Nvidia chief also admitted the company is wrestling with a stubborn supply bottleneck. The RTX Spark, a superchip that packs an Arm-based CPU with a Blackwell GPU onto a single system-on-a-chip, is designed to bring personal AI agents to laptops and desktops. Yet even as Nvidia pushes into new markets, physical component shortages remain the primary constraint on near-term revenue growth—a tension that framed the entire presentation.

The RTX Spark integrates up to 20 Arm cores, 6,144 CUDA cores and 128 gigabytes of LPDDR5X memory with a bandwidth of 300 gigabytes per second, delivering one petaflop of AI computing power. That is enough to run complex AI workloads locally without leaning on cloud resources, a shift Huang described as moving “from tool to teammate.” Microsoft is a key partner in the platform, with the two companies building a secure Windows ecosystem that leverages new operating-system security primitives and Nvidia’s OpenShell. First devices from Microsoft, Dell, HP, ASUS, Lenovo and MSI are expected before the end of the year, with the RTX Spark laptops measuring just 14 millimetres thick and weighing about 1.4 kilograms.

Nvidia’s financial performance underscores why it can afford to mount such an ambitious campaign. Revenue for the first fiscal quarter of 2027 hit $81.6 billion, up 85 percent from a year earlier, with the data-centre division contributing the lion’s share at $75.2 billion—a 92 percent jump. Management guided for second-quarter sales of roughly $91 billion while maintaining a gross margin around 75 percent. On the capital-returns front, the company expanded its share buyback programme by $80 billion and raised its dividend.

The stock initially rode the Computex momentum, gaining 6.26 percent on June 1 to close at $224.36 and adding roughly $319 billion in market capitalisation. The following session saw a modest pullback of 0.69 percent to $222.82. In euro terms, the shares trade at €186.20—down about 3 percent on the day but still up roughly 15 percent year to date. That compares with a 52-week high of €202.50 and a low 51 percent below the current level. The relative strength index stands at 52, squarely in neutral territory.

Beyond the desktop push, Huang confirmed that Nvidia’s Vera CPU, designed for AI data centres, has entered mass production with OpenAI, Anthropic and SpaceX among the first customers. The company is also deepening its presence in physical AI, unveiling the Cosmos 3 foundation model that links visual logic with action prediction, and the Alpamayo 2 Super model for robotaxis, which operates with 32 billion parameters. As Nvidia stretches its reach from hyperscale clusters to edge devices and now personal notebooks, the biggest question remains whether the supply chain can keep pace with an order book that shows no sign of cooling.