If AI software was the biggest tech trend in 2023, it’s only natural that hardware will follow in 2024.

In the year ahead, you can expect generative AI to become an integral part of smartphones, laptops, and other gadgets. Already, rumors of new AI-imbued devices abound, and some companies have been alluding to big plans even if they haven’t announced any specifics yet.

Depending on your view of generative AI, such developments will either be exciting or exhausting, but we can all agree they’re inevitable. Here’s what to expect on the AI hardware front in 2024 based on what we’ve heard so far:

GOOGLE

Google’s plan to build generative AI into its hardware is no secret, and it’s already happening:

• The Pixel 8 line includes a slew of photo editing features that use on-device generative AI to fill in details. Those include “Magic Editor,” which lets users move and resize people after taking the photo, and “Best Face,” which can stitch together successive shots to include everyone’s ideal facial expressions.
• A recent Pixel 8 Pro update enabled generative AI summaries for audio recordings in Google’s Recorder app and smarter auto-reply suggestions in WhatsApp. Google says these are based on a “Nano” version of its Gemini large language model, optimized for on-device tasks.

A broader AI hardware push is reportedly coming in 2024. Sources have told The Information‘s Jon Victor that Google is planning a new AI assistant called “Pixie” for the Pixel 9 and may be planning a pair of glasses with AI object recognition, though the timing for the latter is unclear.

APPLE

Apple’s stance toward generative AI has been more muted than some of its rivals. In an interview with Forbes contributor David Phalen, CEO Tim Cook said the company has worked on generative AI for years, but doesn’t feel the need to be first or slap “AI” on the names of related features.

Still, the rumor mill suggests that Apple does have some big generative AI plans in the works for next year:

• Bloomberg’s Mark Gurman reported that Apple may bring generative AI features to Apple Music and the iWork suite, for instance to create playlists or generate slide decks.
• The Information‘s Wayne Ma reported that Apple wants to bring large language models to Siri in iOS 18. This could allow it to perform complicated, multistep commands such as creating a GIF from a series of photos and sending it through the Messages app.
• As noted by MacRumors, a research note from Haitong International Securities analyst Jeff Pu claimed that Apple has built hundreds of AI servers in 2023 with many more to come next year, and that its AI plans involve a combination of cloud computing and on-device processing.
• Apple researchers published a paper in December describing a novel way to run large language models on computing devices with limited memory, using an unspecified computer with an M1 Max processor as an example.

OPENAI

OpenAI hasn’t announced any hardware plans, but both the Financial Times and The Information reported in September that the company has brainstormed with former Apple design chief Jony Ive on a consumer device of some kind. The FT cited advanced talks with SoftBank CEO Masayoshi Son on $1 billion in funding to build the “iPhone of artificial intelligence,” though the report doesn’t attribute that term to anyone in particular.

WINDOWS PCS

Intel is talking up the “AI PC” with its 14th-gen Core Ultra chips, which have an onboard Neural Processing Unit to speed up on-device generative AI tasks. The company is working with more than 100 developers—including Adobe and Zoom—on optimized software, and it’s already shown off a Stable Diffusion image generator inside the open-source image editor GIMP.

AMD has similar plans for its Ryzen 8000 chips. The company says it’s already enabled AI features on software from Adobe, Blackmagic Design, and Topaz Labs, and it’s working with Microsoft to make sure Windows takes advantage of AMD’s AI processing.

Meanwhile, Qualcomm is planning a big entry into the laptop market with its Snapdragon X Elite chip, but the applications so far are murky.

ANDROID PHONES

On the mobile side, Qualcomm has teased generative AI capabilities for its Snapdragon 8 Gen 3 mobile processor, including on-device versions of the Stabile Diffusion image generator and Meta’s Llama 2 large language model. Honor has already confirmed a virtual assistant based on the latter feature for its forthcoming Magic 6 smartphone.

Samsung, as you might guess, is joining the generative AI party as well, presumably with an assist from Qualcomm. The company proclaimed in a November blog post that “a new era of Galaxy AI is coming,” though it only mentioned real-time call translation as an example. A report by The Korea Times‘s Baek Byung-yeul claimed that Samsung is also developing its own large language model, called “Gauss,” to help edit and summarize text on the Galaxy S24 series.

MISCELLANEOUS GADGETS

Alongside phones, tablets, and computers, we’ll surely see some attempts to build new hardware around generative AI.

One prominent example is Humane, whose AI pin essentially stuffs ChatGPT inside a wearable device with a tiny projector for showing imagery on your palm. It begins shipping in March.

A generative AI makeover is likely coming to smart speakers as well. Amazon has already teased a new version of Alexa built on new large language models, though it hasn’t said on which hardware this new version will run.

PREPARE YOURSELF

As hardware makers race to capitalize on AI hype, the biggest challenge for the rest of us will be to tune out the noise and figure out what’s actually useful.

Already, it seems likely that we’ll see companies use “AI” as a buzzword for features that aren’t new, like offline translation and dictation, or for the kind of image processing that’s been a big part of smartphone photography for years.

You can also expect a lot of cheap parlor tricks that don’t offer much actual utility, along with on-device generative AI features that are markedly worse than their cloud-based counterparts. “Why wouldn’t I just use ChatGPT for this?” will be a useful litmus test for any AI features that device makers announce.

But surely some device makers will figure out how to integrate generative AI in clever and useful ways—or they’ll make a lot of noise trying.