Nvidia takes AI battle from the data centre to the laptop

{"text":[[{"start":9.8,"text":"Nvidia’s decision to enter the crowded market for AI computer chips this week puts it at odds with a long-held company philosophy that it only pursues markets that do not yet exist."}],[{"start":20.700000000000003,"text":"The $5.4tn tech giant is betting it can replicate the success of its data-centre chips against established heavyweights Apple, Intel, AMD and Qualcomm in the realm of semiconductors for personal computers."}],[{"start":34.85,"text":"Jensen Huang, Nvidia’s chief executive, argued that the company’s RTX Spark chip, unveiled at the Computex trade show in Taipei this week, would create another “zero billion-dollar market”, referring to promising technologies that make no revenue now, just as Nvidia’s graphics processing units and software helped catalyse the rise of AI."}],[{"start":56,"text":"“We’re not taking market share from anyone,” Huang told reporters on Tuesday. “The only reason we create something is to address a future market that doesn’t exist today.”"}],[{"start":65.85,"text":"Yet analysts said Nvidia was looking to secure a role in an existing and growing market for devices that run AI software locally. They questioned whether its pricey chip, which is set to target a fairly niche consumer base, would succeed."}],[{"start":81.1,"text":"Rivals such as Apple already produce computers with AI-capable chips. But Huang described RTX Spark as the first computer chip designed explicitly for the emerging era of AI agents: software that can operate multiple programmes with minimal human supervision."}],[{"start":98.69999999999999,"text":"Huang envisaged AI PCs handling tasks ranging from managing inboxes to controlling household equipment such as lawnmowers, with users interacting with them remotely through messaging apps."}],[{"start":111.04999999999998,"text":"Competition between Intel and Nvidia intensified this week as both companies signalled ambitions in each other’s traditional strongholds. Intel said it planned to launch an AI data centre graphics processing unit by the end of the year, seeking to challenge Nvidia’s dominance in that market."}],[{"start":null,"text":"

"}],[{"start":127.09999999999998,"text":"Meanwhile, Nvidia’s push into PC chips marks a departure from its core business of selling highly profitable AI accelerators for data centres and graphics cards for computers and games consoles. Consumer devices are a lower-margin, more fragmented market."}],[{"start":144.24999999999997,"text":"Paolo Pescatore, an analyst at PP Foresight, said the move reflected Nvidia’s efforts to extend its reach as AI workloads increasingly moved “closer to users, devices and businesses”, a phenomenon known as edge computing."}],[{"start":160.39999999999998,"text":"The prospect of running AI locally on devices represents a longer-term strategic challenge for Nvidia. As personal devices become more powerful, a greater share of computing can take place outside data centres, potentially offsetting demand for cloud-based processing."}],[{"start":176.79999999999998,"text":"“The PC market may seem small compared with Nvidia’s huge data centre opportunity, but it matters strategically,” Pescatore said. “It gives Nvidia greater diversification, more control over the ecosystem and a stronger position as AI becomes embedded across every device and every experience.”"}],[{"start":194.85,"text":"Aravind Srinivas, founder and chief executive of AI search start-up Perplexity, told the FT that AI PCs could help reduce soaring cloud-computing costs, which many companies have cited as a growing concern."}],[{"start":207.65,"text":"“If part of the data centre comes to your laptop, everyone is far more empowered,” he said. Running more AI workloads locally would reduce spending on tokens and ease concerns about storing sensitive files in the cloud."}],[{"start":221.6,"text":"Perplexity announced this week a partnership with Intel to develop software tailored to the chipmaker’s AI PC platform."}],[{"start":229.4,"text":"Nvidia is working with laptop makers including Dell, Lenovo and Asus on Windows devices expected to launch in the third quarter."}],[{"start":238.15,"text":"While pricing has not been disclosed, several partners said the machines would cost substantially more than conventional PCs, partly because AI workloads require larger amounts of memory, the price of which has risen sharply in recent months."}],[{"start":254.5,"text":"“Initial production will probably be fairly limited,” said a product manager at a laptop manufacturer using the chip. “We’re still negotiating with Nvidia on pricing, but it will be expensive.”"}],[{"start":265.3,"text":"Ian Cutress, a semiconductor consultant at More Than Moore, estimated the computers would cost between $3,000 and $4,000."}],[{"start":274.25,"text":"“That makes this a niche product,” he said. “Even among AI developers, that price range competes directly with high-end Apple MacBooks, which already have strong traction in software development and local AI workflows.”"}],[{"start":287.55,"text":"Several partners also expressed concerns about software issues that often accompany new chip architectures."}],[{"start":293.90000000000003,"text":"RTX Spark is based on Arm architecture, rather than the x86 architecture used by Intel and AMD. Qualcomm has spent years trying to establish Arm-based Windows PCs but encountered compatibility challenges when running software originally designed for x86 systems."}],[{"start":311.85,"text":"Cutress said it could take “two to four generations of products at a minimum” for Nvidia to achieve the level of software compatibility and performance users expect."}],[{"start":322.20000000000005,"text":"“They will trip on many of the same issues Qualcomm did,” he said. "}],[{"start":327.00000000000006,"text":"Huang said Nvidia was committed to building AI PC chips and would “expand the footprint of this architecture for a very long time”. "}],[{"start":335.45000000000005,"text":"“Once we start a new product line, we support it for as long as we shall live,” he said."}],[{"start":null,"text":""}],[{"start":348.75000000000006,"text":""}]],"url":"https://audio.ftcn.net.cn/album/a_1780566672_1634.mp3"}