Summary: On January 2, 2026, TrendForce reported that OpenAI moved the first-device contract from Luxshare to Foxconn. Production is expected in Vietnam or the United States, with a 2026 or 2027 debut and a screenless, pen-shaped device designed with Jony Ive. The reports remain secondhand and do not establish a launch date.
January 2, 2026: TrendForce, citing Taiwan’s Economic Daily News, relayed that OpenAI has moved the contract for its first AI hardware device from China’s Luxshare to Foxconn, with production expected in Vietnam or the United States and the product aimed at a 2026 or 2027 debut. Same-day coverage from The Neuron and the Times of India sketched in the hardware itself: a screenless, pen-shaped device built with Jony Ive, codenamed Gumdrop internally, built to transcribe handwritten notes into ChatGPT and hold voice conversations.
A manufacturing swap is ordinarily the dullest entry in the supply-chain ledger. Not this one — it answers three questions in a single stroke: what OpenAI’s first consumer device actually is, why the company is routing it around Chinese manufacturing on purpose, and what a “third core device” beside the laptop and phone is meant to be.
From Luxshare to Foxconn: How the Order Moved
Per TrendForce’s write-up, the device had been slated for Luxshare, a fixture of Apple’s supply chain. Neither cost nor capacity drove the switch — the driver was OpenAI’s unwillingness to build its inaugural hardware product in China. The order now sits with Foxconn alone, a doubled prize for the assembler, which scoops up both ends of OpenAI’s hardware ambitions at once: the cloud side (AI servers) and the edge side (consumer devices). The Times of India, leaning on industry tipster Smart Pikachu, puts Vietnam forward as the likeliest production site, notes that OpenAI also sounded out Foxconn’s US facilities, and pegs mass production for 2026 or 2027.
Gumdrop: The Screenless “Third Core Device”
The device’s silhouette has been sharpening across successive reports. The Information and South Korea’s Chosun Daily earlier described a pen-style unit fusing handwriting input, voice interaction, and ChatGPT. The Neuron adds that OpenAI has two screenless, audio-first devices in parallel development, with a third concept still on the evaluation bench. The flagship, codenamed Gumdrop, is positioned as a “third core device” to sit beside laptops and phones — pointedly not earbuds, and not a conventional wearable. The second device is a portable “to-go” audio companion.
Design language comes from Ive’s LoveFrom studio. Sam Altman has described the result as “simple and beautiful and playful”; Ive pushes the calmer-than-a-smartphone line, with lower addiction risk — and the absent screen is itself part of the thesis. In a November 2025 interview the pair confirmed a locked prototype and a launch inside two years.
The program’s roots lie in OpenAI’s acquisition of Ive’s hardware startup io: an all-stock deal valued around $6.5 billion, announced in May 2025 and closed in July, ferrying over roughly 55 engineers — former Apple hardware leads Evans Hankey and Tang Tan among them. LoveFrom now sets the design language across OpenAI’s entire product line.
Why Not Made in China
On the surface, geopolitical hedging: with US-China technology controls still tightening, placing the first consumer device on Chinese lines would etch supply-chain risk straight into the product roadmap. One level down sits signal control. OpenAI treats hardware as a dedicated channel for ChatGPT — a device carrying a microphone and a running stream of handwritten content — and where it is made shapes how much trust users and regulators extend to it. The Neuron labels the Luxshare decision without ceremony: a political call.
Audio First: The Q1 2026 Model Architecture
The hardware calendar is welded to the model calendar. TrendForce’s report notes OpenAI’s next audio model targeting the first quarter of 2026; The Neuron identifies the architecture’s lead as Kundan Kumar, recruited from Character.AI, with the work centered on more natural speech, faster responses, and real-time interruption handling. Those are the exact prerequisites for a screenless device: once voice is the only interface, latency and turn-taking quality stop being feature bullets and become the product itself.
Lessons for Product Teams
Three takeaways. First, AI hardware’s graveyard is already crowded — Humane’s AI Pin, the Rabbit R1, the Friend pendant — and every one of them pitched itself as a phone replacement. OpenAI’s differentiation is scope reduction: claim the concrete scenarios of handwritten notes and voice conversation first, not every use case at once. Second, supply-chain choice has been promoted to a product decision; geopolitical risk moves the manufacturing partner, the cost structure, and the launch window, and a hardware roadmap should carry it as a first-class variable. Third, if the “third device” materializes, builders inherit a new input surface (handwriting, voice) and a voice-first interface with long dwell times — designing response pacing and information density for ears rather than eyes is where the next round of assistant products gets decided.