This year’s smartphones are not just about artificial intelligence

Apple revealed its iPhone 16 lineup on Monday, and its biggest selling point was Apple Intelligence. The AI ​​system on Apple’s device offers flashy features like the ability to rewrite emails, generate custom emojis and a vastly improved Siri. But underneath it all, the AI ​​is bringing another huge change to the iPhone: more RAM.

Although Apple never mentions RAM in its smartphones, MacPlots discovered that every iPhone 16 model now has 8GB of RAM, up from 6GB in last year’s entry-level models. And Apple isn’t the only one making such changes. Last month, Google made similar changes to its AI-dominated Pixel 9; both the standard and Pro models have increased the amount of RAM, making 12GB least you can get it this year.

The impetus for RAM increases seems to be AI. AI is the fresh must-have feature of the year, and it’s also incredibly RAM-hungry. Smartphone manufacturers are now increasing memory because they have to — whether they say so out raucous or not.

AI models need to respond quickly when users call on them, and the best way to do that is to constantly load them into memory. RAM responds much faster than a device’s long-term memory; it would be annoying if you had to wait for an AI model to load before you could grab a quick email summary. But AI models are also quite huge. Even a “small” one like Microsoft’s Phi-3-mini takes up 1.8 GB of spaceand that means taking away memory from other smartphone features that previously used it.

You can see this happening very directly on Pixel phones. Last year, Google didn’t enable local AI features on the standard Pixel 8 model due to “hardware limitations.” Spoiler: It was RAM. Android VP and GM Seang Chau said in March that the Pixel 8 Pro would do better with the Gemini Nano, the company’s petite AI model, because that phone had 4GB more RAM (12GB) than the Pixel 8. That model had to be memory-heavy at all times, and the implication was that the Pixel 8 would lose too much memory to support the feature by default.

“It wasn’t as simple as just saying, ‘OK, we’ll just enable this on Pixel 8 as well,’” Chau said. Google eventually allowed Gemini Nano on Pixel 8, but only for people who want to run their phones in developer mode — people who, Chau said, “understand the potential impact on user experience.”

These trade-offs are why Google decided to boost RAM across the board for the Pixel 9. “We don’t want the rest of the phone experience to slow down to accommodate a large model, and thus increase the total RAM rather than squeeze it into an existing budget,” Stephanie Scott, a product manager for Google Group, said in an email exchange with Edge.

Will all of that extra RAM be dedicated to AI alone, or will users see performance improvements across the board? A lot will depend on the implementation and size of these models. Google, which added 4GB to support local AI features, says you’ll see improvements in both cases. “Just talking about our latest Pixel phones,” Scott wrote, “you can expect both improved performance and an improved AI experience with the additional RAM.” She added that the Pixel 9 phones “will be able to keep up with future advances in AI.” But if those advances mean larger models, that could easily mean they’ll operate more RAM.

The same trend of increasing RAM is happening in the laptop world as well. Microsoft mandated earlier this year that only machines with at least 16GB of memory could be considered Copilot Plus PCs — laptops capable of running local Windows AI features. That roach that Apple also plans to add more RAM to its next-generation laptops, after offering 8GB of RAM by default for years.

That extra memory will be needed, especially if laptop makers want even larger models to be loaded locally. “I think most operating systems will always load LLM,” Hugging Face CTO Julien Chaumond told me in an email, “so 6-8GB of RAM is the sweet spot that will unblock that in parallel with other things the OS is already doing.” Chaumond added that models could then load or unload “a small model on top to change some properties,” such as the image generation style or domain-specific knowledge for LLM. (Apple describes his approach similarly.)

Apple hasn’t said outright how much RAM is needed to run Apple Intelligence. But every Apple device that runs it, starting with the 2020 MacBook Air M1, has at least 8GB of RAM. Interestingly, last year’s iPhone 15 Pro with 8GB of memory can run Apple Intelligence, while the standard iPhone 15 with 6GB of RAM can’t.

Apple AI CEO John Giannandrea he said in an interview in June with Daring Fireball’s John Gruber that limitations like “on-device bandwidth” and the size of the neural engine will make AI features too sluggish to be useful on the iPhone 15. Apple’s vice president of software engineering Craig Federighi said during the same keynote that “RAM is one piece of the whole thing.”

The 2GB RAM boost in the iPhone 16 isn’t a huge deal after all, but Apple has long been sluggish to expand the base RAM in its devices, and any boost feels like a win in terms of usability, even if the company starts petite.

We still don’t know how useful Apple’s intelligence will be, or whether a petite memory bump will be enough to allow today’s iPhones to support tomorrow’s AI features. But one thing seems certain: We’ll be seeing more of these types of hardware improvements as AI spreads across the industry.