Qualcomm (QCOM.US) Releases Snapdragon 8 Elite Gen 6 Chip: TSMC 2nm Process to Target Apple A20 Pro to Focus on “End Side Intelligence”
The Zhitong Finance App learned that Qualcomm (QCOM.US) has released the latest chipset for Android phones, which will be tuned for end-side artificial intelligence and will compete with the A20 Pro in the Apple (AAPL.US) iPhone. Qualcomm said in a press release on Tuesday that the new chip, branded as “Snapdragon 8 Elite Gen 6,” will be manufactured using TSMC's 2nm process and will be used in flagship smartphones from companies such as Motorola, Xiaomi, and ZTE. Qualcomm is now launching two versions, one of which has additional capabilities and is branded “Extreme (Extreme)”.
The announcement comes at a time when the entire smartphone industry is being hit by a global contraction due to soaring memory costs — rising memory prices are already driving up the price of finished devices. According to Counterpoint Research, the overall smartphone market is expected to shrink by 14% in 2026, and may shrink by another 1% in 2027. The response of smartphone companies is to focus more on the high-end device layer — the target customers are consumers with more disposable income.
Qualcomm positions phones equipped with its high-end chips as “AI hubs” that can process AI models and generate tokens without relying on the cloud. At the press conference earlier this month, Apple described the iPhone as a “smart personal hub” that can perform AI processing on the end side.
“The transformation we are experiencing is from a very phone-centered model to a new experiential model centered on smart devices,” Qualcomm CEO Cristiano Ammon said at the launch event on Tuesday.
In addition to the central processing unit (CPU), Qualcomm's new chipset also includes a graphics processor (GPU) and a neural processing unit (NPU) to handle AI workloads, and can handle agents running in the background. Qualcomm has also added security features to control agents' access to sensitive data.
At the live online press conference on Tuesday, Anmon invited Rick Ostro, head of Google's platform and device business. The latter said that thanks to the chip's computing power and sensors, Google's Gemini will be able to understand the user's situation — such as whether this person is in the car. Qualcomm says the Snapdragon 8 Elite Extreme Gen 6 chipset can run models with 30 billion parameters.
“AI now requires devices to do far more than execute commands; it will be less about navigating between apps and more about completing tasks for you,” Anmon said.
Two chips, one digital
Qualcomm unveiled these two chips at the Snapdragon Summit in Maui, Hawaii on September 22, local time. Both are TSMC's 2nm process: the standard Snapdragon 8 Elite Gen 6 (model SM8950) and the higher-spec Snapdragon 8 Elite Extreme Gen 6 (SM8975). The two chips share the same CPU architecture—custom Oryon cores, 2 Prime cores with a maximum frequency of 5.0 GHz and 6 4.0 GHz performance cores, shared cache with 16MB Flex Cache; the connection part is the same as the X105 5G baseband (downlink up to 14.8Gbps; Qualcomm claims it is the industry's first solution to support 3GPP Release 19) and FastConnect 8800.
The performance improvement numbers are all Qualcomm's official caliber compared to the previous generation Snapdragon 8 Elite Gen 5: the standard Extreme Edition CPU performance is increased by about 10%, the energy efficiency is increased by 37%, GPU performance is increased by 35%, the energy efficiency is increased by 40%, NPU performance is increased by 14%, and AI performance per watt is increased by 20%; the performance improvements of the Super Extreme Edition are 13%, 44%, and 35%, respectively, and the AI performance per watt is increased 33%. The Super Extreme Edition also features 18MB Adreno high-performance video memory, supports 8K 60fps video recording and 4K 240fps slow motion, and the NPU adds an Element Accelerator unit to the Transformer payload. Qualcomm stated in the product brief that the actual test results vary depending on the manufacturer's implementation method — these are manufacturer figures and are yet to be independently tested and verified by the whole machine.
In terms of the pace of launch and first sale, media reports say that the Xiaomi 18 Pro will debut the Standard Extreme Edition and enjoy an exclusive period of about half a year; Motorola released the Signature 27 with the Super Extreme Edition on the same day. The partner brands listed in Qualcomm's official press release include Honor, iQOO, Motorola, OnePlus, OPPO, REDMI, Red Magic, Vivo, and Xiaomi.
In terms of competition, Apple is not the only competitor in this generation of chips. Samsung's self-developed Exynos 2700 has been used as a direct benchmark against the Snapdragon 8 Elite Gen 6 in benchmark testing; MediaTek's Tianji flagship series has long rivaled the Snapdragon 8 series in the high-end Android market. Qualcomm chose to launch two flagships in the same generation rather than one, and was also interpreted by the industry as a response to the memory cost crisis — leaving room for mobile phone manufacturers to make cost trade-offs between the Standard Edition and the Extreme Edition in order to maintain the coverage of the Snapdragon brand on different price flagships.
The engineering details that support the “smart body resides” are placed by Qualcomm on the sensing and memory side: the upgraded Sensing Hub (Sensing Hub) integrates a dual-core AI processor, dual normally open camera and dual micro NPU, and can continuously sense the environment and process a model with up to 200 million parameters in low power consumption mode. Qualcomm claims that its performance is up to 85% higher than the previous generation; the Personal Scribe function creates a personal context layer that users choose to share information and conversations that are continuously updated. These designs are a footnote to Anmon's statement of “shifting from being centered on mobile phones to being centered on smart devices” — the chip manufacturer's narrative is shifting from “whose phone is better at doing things for others.”
Benchmarking A20 Pro: Both companies are betting on “end-side smart devices”
The A20 Pro released by Apple earlier this month is Apple's first mobile phone chip using TSMC's 2nm process. It is equipped with the iPhone 18 Pro series: the number of neural network engine cores doubled to 32 cores (dual 16 cores), added neural accelerators to the GPU, and the FP8 AI load speed was twice that of the previous generation, and the memory bandwidth was increased by 50%. Apple says the AI processing power on the end side is double that of the A19 Pro, and the Siri AI beta in iOS 27 also relies more on local processing than any previous generation.
The narratives of the two companies are almost identical: Apple positions the iPhone as a “smart personal hub,” and Qualcomm positions the Android flagship as an “AI hub.” The common denominator is that it allows smart devices to operate locally and rely less on the cloud. The difference is in the baseband — the iPhone 18 Pro completely abandoned the Qualcomm baseband for the first time in the Pro series and switched to Apple's self-developed C2. This is another Apple order that Qualcomm lost.
2 nanometers are the common foundation of this duel. TSMC's N2 uses a surround gate (GAA) transistor architecture, which can integrate more transistors and significantly reduce electric leakage within the same wafer area; Apple claims that the new process has brought about the biggest increase in battery life in its history. Timeline: On September 9, local time, Apple was the first to release the world's first 2 nm mobile phone chip, and Qualcomm followed up two weeks later — the two companies together formed the first wave of mobile phone customers using TSMC's most advanced manufacturing process. For both companies, the logic of betting on more advanced manufacturing processes is the same: large models need to stay on the end side. Energy efficiency means battery life, and battery life is the premium capacity of high-end models.
Lack of memory: cracks at the foot of the press conference
The industry undertones at this press conference were not easy. Counterpoint anticipates a 14% reduction in global smartphone shipments in 2026, so manufacturers are betting their resources on the high-end — Qualcomm's “dual flagship” strategy itself has also been interpreted by the industry as a response to the memory cost crisis: leaving room for manufacturers to make a cost trade-off between the standard version and the premium version.
The strength of the increase in memory prices has been confirmed by multiple measures: Intel CEO Chen Liwu said on September 15 that the price of some memory products had risen to 5 to 7 times the original, that the memory costs of low-end phones and notebooks account for 70-80% of the total material cost, and “there will be no relief until 2028”; TrendForce data shows that server DDR5 contract prices rose 93% to 98% in the first quarter. The price transmitted to the terminal has already appeared: the starting price of the iPhone 18 Pro rose to $1,299 ($1,099 in the previous generation), and the media generally attributed the price increase to ongoing memory cost pressure. For Qualcomm, this means that the story of the computing power of high-end chips must be fulfilled in a race against memory bills.