At the annual Snapdragon Summit held on Maui, Hawaii, Qualcomm China Chairman Meng Pu sat down with Chinese media and disclosed for the first time that the China team has doubled in size over the past eight years, that the Chinese market contributes half of Qualcomm's global automotive revenue, and that the timing of the summit has been repeatedly pushed earlier because of demand from Chinese manufacturers. He also stated candidly that rising memory chip prices have created enormous pressure, and that on-device AI agent collaboration will not truly materialize until 2027-2028.
Summit moved earlier because of Chinese demand
The list of first-launch devices at this year's Snapdragon Summit includes Xiaomi, vivo, RedMagic, and OPPO — almost entirely Chinese manufacturers. Meng did not shy away from the logic behind this: in earlier years, Qualcomm's flagship launches were often delayed until October or even later, but Chinese OEMs wanted new phones on shelves before the National Day holiday, and that demand in turn forced Qualcomm to move its launch cadence up to late September — and this year, because of the Mid-Autumn Festival, it shifted a few days earlier still. "The launch timing of Qualcomm's flagship SoCs and the new product release rhythm of Chinese OEMs have always been very closely coordinated." This year Qualcomm launched dual flagships simultaneously for the first time, which Meng explained as a dual consideration of technology and cost: one chip sets the upper bound of what the industry can achieve, while the other pursues efficiency and serves a broader customer base. Asked about the phenomenon of vivo and OPPO flagships not using Qualcomm chips and Xiaomi's foldable adopting a self-developed chip, he did not dodge the question: "Manufacturers actually run multiple phone brands. iQOO and OnePlus will use Qualcomm chips for their flagship launches in the coming weeks, while the Xiaomi digital series has used Qualcomm platforms for many years." He stressed that Qualcomm will not change its product cadence because of this: "What we need to do is make sure our flagship chips are absolutely top-tier."
China team doubled in eight years
How fast has Qualcomm's China business grown? Meng disclosed that in 2018 Qualcomm had roughly 3,000 employees in China, and that number has now grown to more than 6,000, with new R&D centers added in Beijing, Shenzhen, and other locations. One important driver of this expansion is the unique nature of China's large-model ecosystem. "China may have more large models than the United States, and they have strong localization characteristics that teams in other regions find hard to fully cover," Meng said. In the past, once a phone SoC completed an adaptation, it often did not need adjustment for half a year or more, but now large models iterate every three months or even faster, causing adaptation workloads to surge. During last year's Snapdragon Summit, Qualcomm set up a Beijing satellite venue and launched an "AI Acceleration Program," bringing together carriers, device makers, and large-model companies to push on-device AI adoption; the AI Hub released in 2024 provides developers with a model library optimized for the Snapdragon platform, and Meng said its rollout in China has "basically completed the first step, and we will continue to drive scale adoption." Asked whether Qualcomm would let its China team operate independently, he gave a clear no: "Completely isolating the China team from the R&D system built up over decades of globalization is neither realistic nor conducive to serving Chinese customers."
Memory price hikes pressure low-end phones
A major challenge for the smartphone industry this year is the surge in DRAM and NAND prices, which have risen four to five times within a year. Meng called this a "rare" phenomenon in his decades in the industry. He offered a concrete calculation: a 1,000-yuan phone originally cost about $40, with memory, SoC, screen, and other costs each accounting for roughly $10; now memory alone has risen to $40 to $50, pushing the total device cost up to $70 to $80. "The original 1,000-yuan price segment is very hard to sustain." He judged that this round of price increases will not stabilize until at least the end of next year, and a rapid decline is unlikely. On cost pass-through, he also noted that Qualcomm's fabless model means upstream price increases are transmitted directly. Both flagships released this year use 2-nanometer processes, with unit costs clearly higher than 3-nanometer. "When costs rise too fast and too high, it often also forces technological and product innovation."
China contributes half of automotive business
Meng listed automotive as one of Qualcomm's fastest-growing segments over the next three to five years. He revealed one detail: on September 22 alone, Chinese automakers launched nine new cars, seven of which used Qualcomm solutions. More critically, Qualcomm China currently contributes nearly half of the company's global automotive revenue. At its Investor Day in June this year, Qualcomm raised its fiscal 2029 non-handset revenue target to $40 billion, with handsets, "automotive plus IoT," and data centers each expected to account for one-third. On China's role, Meng was quite candid: "I hope that after the data center business takes off, Qualcomm China's share of the company's overall revenue can remain unchanged." Beyond automotive, embodied robotics is another China example Meng brought up on his own. "Many Chinese companies are positioning themselves in embodied robotics, and Qualcomm maintains broad cooperation with basically all the well-known players in the industry." At two consecutive CES shows, Qualcomm brought Chinese robots to exhibit, and he believes Chinese embodied robots "have global application potential," though whether they can enter different markets depends on local compliance and ecosystem adaptation.
China has clear advantages in the AI cycle
Asked how long it will take for AI agents to truly achieve cross-device collaboration, Meng gave a specific range — around 2027 to 2028. He cited several AI phones that appeared at this year's World Artificial Intelligence Conference: Honor Robot Phone represents traditional manufacturers layering AI capabilities onto existing hardware, Doubao Phone represents a cooperation model between a hardware maker and a large-model company, and StepX Neo from StepFun represents a large-model company making its own hardware. "These paths are all different, and it is still too early to judge which is better." He believes Chinese manufacturers have clear advantages in this technology cycle: a complete industrial chain, capabilities that survived 15 years of elimination rounds, and the world's richest application scenarios. Qualcomm will provide more support to its Chinese partners. Near the end of the group interview, a reporter asked about the biggest risk in the coming year. Meng's answer did not avoid uncertainty, yet carried the confidence of someone with a technical background: "Risk and opportunity actually coexist; the key is whether you can land quickly and execute well." Over the past two years, the industry has rushed frantically toward cloud computing power, while the judgment he has always insisted on is being validated: "On-device AI is the closed loop that AI development must land, and this is becoming increasingly clear." "Truly excellent AI applications will inevitably be the result of device-edge-cloud collaboration. They will not be led by any single type of hardware terminal, but will center on AI agents, flowing across different devices, and ultimately serving every user."
Comments