Samsung unveils up to $80B shareholder return as chip giants cash in on the AI boom
Two stories today show where the AI money is actually landing — on the balance sheets of the memory and hardware makers feeding the build-out — plus a bionic hand that pushes robot dexterity another notch.
Samsung plans to return up to $80 billion to shareholders in 2026. The company said Friday it expects 90 trillion to 110 trillion won (about $65 billion to $80 billion) to be available for shareholder returns this year, calling it the largest program ever announced by a Korean company. The package includes roughly 30 trillion won in third-quarter cash dividends, with the rest to be finalized at a board meeting in late October. The announcement lands just days after rival SK Hynix disclosed a 40 trillion won share buyback, making it a blockbuster week for payouts among South Korea's two chip giants.
Why it matters: this is the AI boom converting directly into cash for the suppliers, not just the model labs. Samsung's second-quarter operating profit jumped 19-fold year over year on sustained AI-driven memory demand, and its stock is up roughly 135% in 2026 as it fights to close the high-bandwidth-memory gap with SK Hynix. When the component makers start returning this kind of capital, it's a signal the spending cycle has moved from hype to durable, profit-generating demand. We covered the demand side earlier this month — AI demand sold out all 2027 DRAM and HBM capacity.
Yuequan debuts a bionic "tension-compression" dexterous hand built like human muscle. At a World Robot Conference side event in Changchun, the robotics firm and the Jilin Bionic Robotics Innovation Center unveiled the YingShou Y-Hand series — three iterations (M0, M1, M2) spanning lightweight general tasks up through high-precision manipulation — alongside new motors and artificial-muscle modules. The design, derived from Professor Ren Lei's bionic tension-compression theory, rebuilds robots around a rigid-flexible, muscle-bone system meant to mimic human movement, softness, and safe human-robot interaction, feeding a broader "brain-body-environment" embodied-intelligence vision.
Why it matters: hands are the bottleneck for robots leaving the lab. Most grippers are still rigid and clumsy; a hand that balances load against fine-motor precision, and that slots into a full perception-to-action loop, is exactly the kind of incremental hardware that decides whether humanoids can do real work. It's early — Yuequan paired the launch with industry-signing ceremonies rather than shipped deployments — but the component-level progress is steady and worth watching.
RayNeo ships the iO, a 34-gram AI glasses pair with two-day battery life. Chinese AR brand RayNeo held its 2026 AI-glasses launch on August 21, introducing the iO series and its first model: glasses that weigh 34 grams, run all-day "active AI" for knowledge, expression, and memory assistance, and start at 1,996 yuan (about $280) with prescription lenses from 0 to 1,000 degrees. The pitch is deliberately anti-gadget — RayNeo says it rebuilt the structure, optics, and frame to shed the bulky look of earlier AI glasses and get back to "glasses first."
Why it matters: the consumer-AI wearables race keeps coming back to two unglamorous specs — weight and battery. At 34 grams the iO is close to ordinary eyewear, and a two-day battery removes the daily-charge chore that sank earlier attempts. If the AI actually runs usefully all day at that weight and price, RayNeo is targeting the everyday-wearable form factor Meta, Google, and the Chinese OEMs are all chasing. The open question is whether the "active AI" is genuinely proactive or just another voice assistant in frames.
What to watch: whether SK Hynix follows Samsung with an even larger return, and whether the iO's battery and weight claims hold up in real-world use.
Is "human-enhancement AI glasses" the category that finally makes wearables stick, or just another expensive pair of specs? Tell us in the comments.
Sources: CNBC · Zhidx (智东西) · QbitAI (量子位) · AI Midday — AI demand sold out all 2027 DRAM and HBM capacity
