Google's upcoming smartphone, the Pixel 11, may become the first smartphone to use TSMC's 2nm process. Taiwan's Economic Daily News reported on July 13, 2026, that TSMC will manufacture the "Tensor G6" chip in the Pixel 11—expected to launch in mid-August—using its 2nm process. Separately leaked specifications point to changes including a 7-core Arm C1 CPU, a PowerVR GPU, and a MediaTek-made modem. Google has not announced any of these specifications. Still, piecing together publicly available materials suggests the outline of a design in which the G6 updates the CPU and modem while reworking power allocation across the entire SoC.
What's New Isn't 2nm Mass Production Itself, but the Timing of Its Adoption in Pixel
The fact that TSMC's N2 has entered mass production is not new information in itself. The company states on its technology page that mass production began on schedule in Q4 2025. In May 2026, AMD also announced that its server-oriented EPYC "Venice" had entered mass production ramp-up on TSMC's 2nm process. Even if the Pixel 11 leads the way, it would not be the first product to use TSMC 2nm—it would be the first smartphone to do so.
According to Economic Daily News, Google is expected to hold a "Made by Google" event on August 12 (US time) to unveil the Pixel 11 series. The paper reported that a smartphone adopting TSMC 2nm would arrive ahead of Apple's next iPhone. This is the most significant shift in this report. Smartphones featuring TSMC's latest process have consistently been led by Apple, but that order may now be reversed.
That said, neither Google nor TSMC has confirmed the manufacturing process used for the Tensor G6. "First customer" for N2 and "first smartphone to launch" with it are also separate matters. While Mystic Leaks specifies N2, Economic Daily News reports only "2nm" without specifying the derivative process. In 2025, MediaTek announced the design completion of a flagship SoC using the improved N2P process, with mass production planned for the end of 2026. Until the G6's process is confirmed, N2P's performance figures cannot be applied to it.
Cutting Power Further from the 3nm Tensor G5
Google moved manufacturing to TSMC's 3nm process with the Tensor G5. According to official announcements, the G5's CPU is on average 34% faster than the previous generation, with TPU performance improved by up to 60%. It also incorporates a new image processing circuit and security circuit, supporting the Pixel 10's on-device AI and camera features. If the G6 adopts N2, it would mark the second generation of manufacturing at TSMC.
TSMC's stated targets for the base N2 process, compared to N3E, are 10–15% faster speed at the same power consumption, or 25–30% lower power consumption at the same speed, along with a chip density increase of over 15%. The process switches from FinFET to nanosheet transistors, allowing finer control over current flow. If the N2 referenced by Mystic Leaks is accurate, Google could choose to complete the same processing in less time, conserve battery while maintaining speed, or add more circuitry.
However, these figures are not a performance preview of the Pixel 11, nor has it been confirmed that the G6 uses the base N2 process. Raising clock speeds would consume the power savings, and expanding the TPU or image processing circuits would reduce the available die area margin. Memory and modem also consume power, and display and thermal design affect battery life and surface temperature. Even if 2nm adoption is confirmed, what the product actually gains won't be known until device-level measurements are made.
The 1+4+2 CPU Configuration Skips Small Cores
Mystic Leaks, a source of leaked information, describes the Tensor G6's CPU as a 7-core configuration: one C1-Ultra core at 4.11GHz, four C1-Pro cores at 3.38GHz, and two additional C1-Pro cores at 2.65GHz. This is a reduction of one core from the reportedly 8-core G5, while shifting all cores to Arm's C1 generation.
According to Arm, the C1-Ultra offers 25% higher single-thread performance than the previous-generation Cortex-X925. The C1-Pro delivers 16% higher sustained performance at the same frequency as the Cortex-A725, and at equivalent performance, reduces power consumption by up to 12% for tasks like video playback and web browsing. Across the entire C1 cluster, average power consumption for everyday mobile tasks reportedly dropped by 12%. All of these figures come from Arm's reference environment, and there's no guarantee the same differences will materialize in the G6.
What's notable is that instead of using the low-power C1-Nano, the design runs two C1-Pro cores at a lower frequency. In this configuration, even lightweight tasks are handled by the low-clock C1-Pro cores. Meanwhile, standby power consumption and heat generation under sustained load depend on voltage settings, cache design, and TSMC-specific physical implementation. Core names and clock speeds alone cannot confirm improved power efficiency.
The C1 generation also includes SME2, which accelerates matrix operations on the CPU. Arm states that in compatible software environments, it can speed up AI processing by up to 5x and improve efficiency by up to 3x. Some of Google's apps already support SME2. However, Pixel's primary AI processing is also handled by a dedicated TPU. How work is distributed between the leaked new TPU, "Santafe," and the CPU will determine both perceived speed and power consumption.
It's Too Early to Call the GPU "a Step Back to an Older Generation"
The same leak describes the GPU as a "PowerVR C-Series CXTP-48-1536." Based on the name alone, this appears close to the CXT series that Imagination Technologies announced in 2021. The company's CXT-48-1536 RT3 featured three ray-tracing cores and claimed up to 1.3GRay/s. The DXT series, which followed the next year, reportedly offered 20% higher performance per unit area than the CXT, along with reduced power consumption.
But it's premature to conclude from this that the G6's GPU is older than the G5's, or that its performance has decreased. Imagination's public materials contain no configuration matching the "CXTP" derivative name mentioned in the leak. Clock speed and core count remain unknown, and ray-tracing hardware and driver features are also unconfirmed. Even if the latter half of the model number matches the DXT-48-1536 reportedly used in the G5, that doesn't mean the die area or operating frequency are the same.
It's equally plausible that the design intentionally caps GPU peak performance to allocate more power to the CPU, modem, and TPU. Conversely, it could use N2's efficiency gains to run a similarly sized GPU at a higher frequency. There is no evidence yet indicating which path was chosen. Evaluation should be withheld until sustained frame rates in 3D games and GPU power consumption are measured on an actual device.
FCC Documents Reinforce the Switch to MediaTek
Regarding the modem, a stronger clue has emerged than the leak alone provides. Android Authority examined an FCC (Federal Communications Commission) SAR test report related to Google's model number "GZDQ6" and found that page 30 references MediaTek's transmit power control algorithm. The document includes tests in both folded and unfolded states, suggesting it relates to the upcoming Pixel Fold.
While the FCC document increases the likelihood that MediaTek-made wireless technology is included, it does not identify the modem as the "M90." The M90 model name comes solely from Mystic Leaks. MediaTek's officially announced M90 supports downlink speeds up to 12Gbps and handles both sub-6GHz and mmWave bands. It also supports dual SIM dual active and satellite communication. MediaTek claims it reduces average power consumption by up to 18% compared to the previous-generation modem—though this figure is based on MediaTek's own benchmarks and does not guarantee battery life on the Pixel 11.
Reading the Tensor G6 reports through the lens of a single component risks oversimplified conclusions—that 2nm equals high performance, or that a CXT-series GPU equals low performance. If the reports and leaks hold true, the design combines a 2nm process and C1 CPU with MediaTek wireless technology, fitting the TPU and image processing circuits within a single power budget. What we should confirm at the August announcement, beyond process and component names, is how Google has allocated power across different workloads. The answer will show up in sustained performance and heat generation during network use after launch, and will also serve as a basis for judging on-device AI speed and battery life.