Counterpoint Research
AI overview
Sign in and the AI will write an overview from our coverage.
Headlines · 11
- Rokid opens world's first unstaffed AI glasses experience store in Hong Kong
Rokid announced on Oct 15 that the world's first unstaffed self-service AI glasses experience store has opened at Harbour City in Tsim Sha Tsui, Hong Kong — its first overseas flagship store. The space has eight immersive zones with no sales staff following shoppers; each zone pairs a mirror-like interactive screen with a pair of Rokid Glasses. The 49-gram glasses combine shooting, teleprompter, translation, navigation and AI Q&A, and buyers can check out and collect the product from a robotic arm after trying it.
IT之家 AI · 🔥 8 - NVIDIA Rubin said to stick with 12-layer HBM4 as HBM prices seen up 121% in 2027
A supply-chain survey cited by Counterpoint Research indicates most of NVIDIA's next-generation Rubin AI accelerators will keep the originally planned 12-layer HBM4 stacks, rather than shifting broadly to 8-layer HBM4 as earlier rumors suggested. NVIDIA's published roadmap lists 12-high HBM4 with up to 288GB of memory per GPU and bandwidth up to 22TB/s; Korean media had reported that NVIDIA asked Samsung Electronics and SK hynix for more 8-layer HBM4 supply, and industry observers say the two accounts need not conflict — some products may use 8-layer stacks to ease capacity pressure while flagship Rubin parts stay at 12 layers. Separately, TrendForce forecasts HBM average selling prices could rise 121% year over year by 2027, driven mainly by the mix shift from HBM3E to pricier HBM4 rather than pure shortage; because the costly base die is spread over fewer memory layers, 8-layer HBM could cost 10% to 20% more per Gb than 12-layer in 2027.
cnBeta · 🔥 4 - CXMT's 4F² DRAM breakthrough triggers global memory chip selloff
On Oct. 10 CXMT president Cao Kanyu said at the 4th Integrated Chip and Chiplet Conference that the company had made key progress on a 4F² DRAM architecture built around its self-developed VSG vertical-surrounding-gate transistor, and expects to ship next-generation DDR5 RDIMM products combining 4F² with wafer-level hybrid bonding by year-end. Memory stocks then fell hard: Samsung closed down 7.39%, SK Hynix dropped 10.09%, the KOSPI lost 5.46% and triggered a SIDECAR program-trading halt, and Micron slipped below $1,000 a share, while CNBC tied the wider Asia tech rout partly to Fed rate-hike fears and worries about AI investment returns. The report also says CXMT posted first-half revenue of 150.3 billion yuan, up 873.64% year on year, with net profit of 77.6 billion yuan and a global DRAM revenue share that rose from 4% in Q2 2025 to 10% in Q2 2026, though its HBM3E is not expected to reach mass production until 2027, roughly three to four years behind Samsung and SK Hynix.
36氪 人工智能 · 🔥 0 - Counterpoint: Nearly half of prescription hearing aids to have on-device AI by 2030
A Counterpoint report on global on-device AI wearables forecasts that worldwide prescription hearing aid sales will exceed 31 million units by 2030, a 4% CAGR from 2025 to 2030, with nearly 50% of those devices carrying on-device AI. The report argues that traditional digital signal processing relies on preset rules and struggles in crowded settings with multiple speakers, whereas deep neural networks trained on millions of voice-and-noise samples can separate speech from noise in real time and preserve fine phonetic cues such as "s", "f" and "th" or easily confused numbers like fifteen and fifty. According to Phonak testing cited in the report, fixed directional microphones with conventional settings improve signal-to-noise ratio by 6.4 dB over omnidirectional mics, while adding AI noise reduction raises that to 9 dB by default and up to 10.2 dB at maximum; however, continuous neural network processing costs battery life — Phonak's 2025 Audéo Infinio Ultra Sphere lasts 56 hours normally but drops to 10 hours with AI always on — and hearing aids require end-to-end latency of 10 milliseconds or less, so the AI must run locally.
IT之家 AI · 🔥 0 - AGIBOT reported raising new round at ~50B yuan valuation, delaying IPO
Citing Caijing magazine, reports say embodied-AI company AGIBOT recently opened a new funding round of roughly 1–1.5 billion yuan that is nearing completion, at a latest valuation of about 50 billion yuan. Investors said Hong Kong Exchange rules require a company that raises money after filing but before listing to wait 120 days after the round closes, meaning AGIBOT cannot complete an IPO this year. AGIBOT was founded in 2023 and said on Sept. 24 that its 20,000th embodied-AI robot rolled off the line; a Counterpoint Research report put global humanoid robot shipments in the first half of 2026 above 22,000 units, up nearly 300% year on year, with AGIBOT first at 9,700 units and over 43% share, and the top five all Chinese vendors at 86% combined.
IT之家 AI · 🔥 0 - Lumentum says AI optical part capacity is sold out into early 2029
Lumentum CEO Michael Hurlston said in a Tokyo interview that demand for the company's optical components continues to outstrip supply as tech firms race to build faster AI data centers, with capacity fully sold out and shortages expected to persist into early 2029. Bloomberg reported that about 70% of demand for some products will remain unmet into 2027, while other products face roughly a 30% supply gap that will be hard to close through 2028; constraints have worsened versus six months ago. Counterpoint Research said Lumentum and Coherent, Nvidia's silicon-photonics co-packaged-optics partner, may be unable to fill the gap left by Chinese suppliers such as Innolight even if Washington moves to restrict new Chinese optical transceiver modules.
科技新报 · 🔥 0 - Qualcomm CEO: AI companies ask for phones running 100B-parameter LLMs locally by 2028
Qualcomm president and CEO Cristiano Amon said in an interview with Sources that several companies at the forefront of AI are asking Qualcomm to support smartphones that run at least a 100-billion-parameter language model locally and continuously by 2028, adding that "AI companies are starting to do phones." He did not name those companies, and Qualcomm has not published a complete hardware plan or commercial phone roadmap for reaching that goal. The report notes a 4-bit quantized 100B model would need roughly 50GB of memory for its weights alone, before room for the OS, apps and context cache, while flagship phones typically offer 12GB to 16GB and DRAM supply is tight and expensive. Possible workarounds include more aggressive quantization, MoE expert offloading or keeping weights in flash storage loaded on demand, each with quality or speed trade-offs; Amon said a more realistic near-term step is distilled 20B-to-30B models, with thermals and battery life still limiting always-on operation.
Wccftech · 🔥 0 - AI memory squeeze drives up device prices as the cheap electronics era ends
An opinion piece argues that AI data centers' appetite for HBM is squeezing consumer-grade memory capacity and passing the cost down to ordinary buyers. Citing TrendForce, it says smartphone LPDDR5X contract prices rose 78%–83% quarter-on-quarter in Q2 2026, with a single 12GB LPDDR5X chip going from $77 in Q1 to $146; Samsung, SK Hynix and Micron control over 95% of the DRAM market, HBM margins exceed 70%, and HBM's share of DRAM wafer capacity has climbed from 8% two years ago to 23%, heading for 30% next year. The article also notes Apple's Mac mini entry price moving from $599 to $799, Microsoft's Xbox Series S rising from $299 to $499, and Nothing scrapping the CMF Phone 2 Pro successor because memory was too expensive, and cites IDC forecasts of 10%–20% higher PC, tablet and phone prices in 2026 with smartphone shipments falling 13%. The author calls these increases a hidden "AI tax" consumers pay without noticing.
虎嗅 AI · 🔥 0 - AI glasses remain a 'nice-to-have' as shipments grow without a killer use case
A commentary by Alter published via TMTPost argues that AI glasses are shipping in growing volumes but still lack a common, high-frequency, hard-to-replace core need. It cites IDC data showing 2025 global smart-glasses shipments of 14.773 million units, up 44.2% year on year, with China at 2.46 million units, up 87.1%, and Counterpoint Research figures showing global AI glasses shipments up 263% year on year in the first half of 2026 — though display-less products accounted for about 96% of shipments, with the best sellers remaining camera-and-audio devices closer to ordinary glasses. The author also cites a 2026 Security.org US survey (52% valued hands-free video, 48% photos, 40% AI assistants) and an S&P Global survey of 2,532 US adult internet users in Q3 2026 (44% said the devices improved work efficiency), arguing vendors are racing for the next interface while consumers remain unconvinced; it also flags a roughly 39.13-gram comfort threshold, the social limits of voice interaction, and privacy-driven acceptance issues, such as Norway's proposed temporary bans on camera-equipped AI glasses in parks, beaches and schools.
钛媒体 · 🔥 0 - AI memory squeeze hits PCs: Q3 shipments plunge 20.1%, MacBook down 11%
IDC data shows global traditional PC shipments fell 20.1% year-on-year to 62.7 million units in Q3, down 9.1% quarter-on-quarter and the second straight quarterly decline, breaking the usual seasonal pattern. Apple's MacBook shipments dropped more than 11%, though its share rose to 9.5% from 8.5% because the overall market fell harder; Lenovo, HP and Dell shipped 14.9M, 10.3M and 7.6M units, down 22.6%, 30.9% and 25.0% respectively, with only ASUS declining by a single-digit rate among the top five. Vendors and channel partners say AI data-centre buildout is tightening memory supply and lifting component prices across capacitors, PCBs, ICs and GPUs; IDC's Wang Jiping said prices "won't go back to what they were" and the hardest stretch may still lie ahead.
钛媒体 · 🔥 0 - Counterpoint: Foundry 2.0 market revenue up 25% YoY to $96.6B in Q2
Counterpoint Research's latest foundry supply tracker puts global foundry 2.0 market revenue at $96.6 billion in Q2 2026, up 25% year over year and 11% quarter over quarter. The firm says growth accelerated from the previous quarter, driven by sustained AI GPU demand and AI infrastructure expansion, including custom ASICs, server CPUs, networking chips and power management ICs.
36氪快讯 · 🔥 0
Experience and discussion from the community
Share my Counterpoint Research experienceAsk about Counterpoint Research
Nobody has shared their experience with Counterpoint Research yet.