If you are asking whether AI smart glasses are worth it in 2026, the honest answer depends entirely on which piece of AI hardware you mean. Meta sold more than 7 million Ray-Ban smart glasses in 2025 alone, tripling its prior pace, and the category is still growing. Meanwhile Humane, the company behind the AI Pin, sold what remained of itself to HP for $116 million after its device generated more daily returns than new sales. Both of these are "AI hardware." Only one of them is actually catching on, and the difference tells you almost everything about what makes AI hardware succeed or fail right now.
The winners: glasses and pins people already had a reason to wear
Meta and EssilorLuxottica, the maker of Ray-Ban, do not talk about smart glasses as a new category. They talk about them as sunglasses that happen to have a camera, speakers, and an AI assistant built in. That framing matters. People were already buying millions of Ray-Bans every year, so Meta did not need to convince anyone to wear something new on their face, only to convince existing Ray-Ban buyers to spend more for the AI version. It worked well enough that EssilorLuxottica has reportedly discussed doubling annual production toward 20 million pairs a year to keep up with demand.
Plaud, a much smaller company, is running the same playbook with note-taking pins instead of eyewear. Plaud has sold more than 1.5 million devices and expects roughly $500 million in sales in 2026, built almost entirely on one narrow job: recording and transcribing meetings for professionals and students who already record meetings some other way. Amazon liked the same idea enough to acquire Bee, a competing $49.99 recording pin, in 2025. None of these products are trying to replace your phone. They are trying to sit next to a habit you already have and make it slightly better.
The losers: hardware that asked you to trust it instead of your phone
The AI Pin from Humane launched in 2024 promising to be a screen-free replacement for parts of your smartphone. It cost $699 plus a monthly subscription, and reviewers found the core functions, like the laser projector display on your palm, unreliable in daily use. Humane cut the price to try to save it. Sales did not recover. Reports before the shutdown indicated the device was seeing more returns than new purchases, and in early 2025 Humane sold its technology and roughly 300 patents to HP for $116 million. Every AI Pin stopped working entirely once Humane shut down its servers.
Rabbit R1 followed a similar arc, though it survived. Roughly 100,000 to 130,000 units sold after a buzzy 2024 launch, but mass returns and reported payroll struggles followed almost immediately. Rabbit is still shipping software updates in 2026 and has found new funding, but it never became the mainstream "AI companion device" its launch keynote promised. The Friend AI necklace had an even rougher debut: about 5,000 units sold at $129, a subway ad campaign that got covered in graffiti protesting always-on listening devices, and a founder who called the backlash "artistically validating" rather than a warning sign. Friend has since relaunched at $249 with a speaker added, betting that a higher price and louder feedback will fix what cheap and quiet could not.
Why the pattern is so consistent
Every product in the "struggling" column asked people to adopt a brand new device category, trust an unproven company with constant audio or visual recording, and give up something they already knew how to use. Every product in the "growing" column attached AI to a device or habit that already existed: glasses people already wore, meeting recordings people already made. Consumer research backs this up in a broader way. Surveys on AI wearables consistently find that around 70 percent of people are concerned about privacy on these devices, while only a small fraction, often cited around 10 percent, actually use the AI-driven personalization features regularly even when they own a device that offers them. People are buying the hardware for one clear job, not for the AI itself.
Are AI smart glasses worth it in 2026, specifically?
For the Ray-Ban Meta glasses, the answer for most people is a qualified yes if you already wanted sunglasses or prescription frames and would use hands-free photos, video, or voice AI occasionally. They start around $299, work as normal glasses even with the AI features off, and do not require you to change any habits. The pricier Ray-Ban Display model, which adds an actual screen in the lens and costs $799 with a wrist-worn Neural Band, is a different bet: it is new, it is expensive, and it has been selling out or requiring a scheduled in-store demo before purchase, which suggests genuine demand but also a much smaller and more curious audience than the base glasses.
Apple has been notably cautious here. Reporting indicates Apple plans to unveil its own smart glasses in late 2026 with a public launch in 2027, and unlike Meta's Display model, the first Apple version is expected to skip a built-in screen entirely, relying only on cameras, microphones, and Siri. That is a telling signal from a company known for waiting until a category is proven: even Apple appears to think a display in your glasses is the harder sell, and a camera-plus-voice-assistant pair of glasses is the safer place to start.
Where does that leave AI in phones and laptops?
Outside of glasses and pins, most "AI hardware" has quietly become a checkbox feature rather than a new device. Microsoft's Copilot+ PC category, which requires a neural processing chip capable of at least 40 TOPS, made up under 2 percent of the entire PC market as recently as early 2025 by the strict definition. Broader definitions of "AI PC," which count any laptop with some dedicated AI chip regardless of power, are projected to cover a majority of PC shipments in 2026. In other words, the AI hardware that is actually spreading fastest is not a new gadget you buy on purpose. It is a spec bump inside the laptop you were going to buy anyway.
That is arguably the real story of AI hardware in 2026: the products asking you to change your behavior are struggling or dead, and the products quietly upgrading something you already own are the ones with real growth numbers behind them.
The takeaway
If you are deciding whether to buy any piece of dedicated AI hardware this year, the sales data suggests a simple filter: would you want this object even if the AI features were mediocre? Ray-Ban buyers would still want sunglasses. Plaud buyers would still want a recorder. Humane Pin and Friend necklace buyers had no reason to want the object itself, only the promise of what the AI inside it might someday do well. That distinction, more than any spec sheet, is what has separated the AI hardware that is actually catching on from the AI hardware quietly heading for the same shelf as last year's failed gadgets.
