Google just made me look like an idiot for believing in its Tensor chips

I’m not a Tensor chipset hater. Most of the criticisms are overblown and stem from a time when Google had legitimate issues with earlier generations. If you were a Pixel 6 or 7 owner, I feel your pain, and there’s nothing Google can do to ease your frustrations. However, Google has cleaned up many of the issues with recent chipsets.

Still, there’s always been a debate about whether Google’s custom SoCs were worthy of a flagship phone. I’ve largely given Google a pass for its lackluster benchmarks compared to Qualcomm’s more powerful chipsets, because it built value in other ways. The strategy is also working. Pixel sales are up, and customer satisfaction has significantly improved.

Unfortunately, the company is testing my patience with the announcement that the latest Gemini Intelligence upgrade will only work on the company’s most recent hardware. It’s a concerning development, and I’m beginning to wonder if Google’s made a fool out of me for defending Tensor.

Google’s Tensor claims go out the window with Gemini Intelligence

It’s almost impossible to comprehend that a Google Pixel 9 Pro XL with a Tensor G4 and 16GB of RAM can’t handle the latest and greatest AI features from Google. Google heavily advertises the AI capabilities of Pixel phones, and some people buy them for those features. It’s why I’ve given Google a pass on poor benchmark performance. There’s only so much room on the die, and Google prioritizes the NPU and other tasks like image processing at the expense of GPU performance.

It’s a trade plenty of people are willing to make, especially if it means their Pixels will get nifty new features like the agentic layer added with Gemini Intelligence. Unfortunately, that’s not the case, and it makes me wonder what the point is. Google had an opportunity to end the Tensor debate and prove that raw performance benchmarks aren’t everything.

If the company came out and said that Gemini Intelligence can only run on the most recent Qualcomm chipsets, but Pixel support would go back to the Tensor G3, I would’ve been sold, and Google’s added value with Tensor would make sense.

But Google didn’t do that, and the resulting conversation is going to get uncomfortable.

Samsung already tried this, and it annoyed me then, too

I remember feeling similar outrage when Samsung announced the then-year-old Samsung Galaxy S24 Ultra wouldn’t get Galaxy AI. It seemed artificial and arbitrary — a way for the company to get you to shell out more for a Galaxy S25 Ultra instead. I didn’t even really care about the Galaxy AI features (and Now Brief is still proving me right almost two years later), but it was the principle.

How could a phone that I just paid $1,300 for not be capable of handling a new feature less than a year after launch? We barely scratch the surface of what’s possible with today’s high-end chipsets. The only time they are truly taxed is when benchmarks are designed to push them to their limits. I used to take comfort in the fact that, while I now pay more for a flagship SoC, that extra power would come in handy, futureproofing my device for years to come. But that’s no longer the case.

If $1,300 phones are going to be rendered obsolete every year, what’s my incentive to spend more? I should just pick up one of the fantastic midrange models that do more than enough and upgrade those more frequently. There is a bumper crop of capable phones available for $500-$600, and it’s getting harder to justify not picking up one of those instead — especially if I’m not going to be receiving the latest features anyway.

There should be a way around this

I understand that generational leaps are possible. I never expected my Super Nintendo to do things only my Nintendo 64 could, but this is different.

I thought Magic Cue’s arrival on the Pixel 9 series was a dress rehearsal for how this gap would be handled going forward. Magic Cue was a Pixel 10 exclusive, supposedly made possible by the Tensor G5 chipset, but it’s now coming to other phones. Newer features could be brought to older devices by offloading more of the computing to the cloud.

Users wouldn’t feel ripped off, and Google would still be able to make a case on why you should upgrade. In most cases, the cloud results are going to be slower and less secure, while new chipsets could handle more and more on the device. Tired of slow performance on that new AI feature? No problem, upgrade to Google’s latest phone. But at least the value of those older devices would be protected, and the entire thing would feel less like a joke at our expense — literally.

Google has reversed course before

Google made a similar announcement with Gemini Nano on the Pixel 8 series. The Pro devices received Nano without issue, but the company claimed it couldn’t come to the base Pixel 8 because the phone only had 8GB of RAM. There was a pretty fierce backlash, and Google came up with a workaround. I’m hoping for a similar outcome here.

It’s unacceptable that features are gatekept, solely to get you to upgrade to another device. If Tensor chipsets truly aren’t up to the task for more than one year, I’m going to start spending my money elsewhere and not look like an idiot in the process.