Summary: The Financial Times ran an interview with TM Roh, the executive in charge of Samsung’s consumer device business, on March 8, 2026. Multi-model is a present fact, not a roadmap item; Roh’s contribution was to spell the direction out: that door will not close, and companies with OpenAI’s heft are on the candidate list. Reduce the conversation to one line and it reads: Samsung is “open to strategic co-operation” with more AI groups, and OpenAI was the example he named.
The Financial Times ran an interview with TM Roh, the executive in charge of Samsung’s consumer device business, on March 8, 2026. Reduce the conversation to one line and it reads: Samsung is “open to strategic co-operation” with more AI groups, and OpenAI was the example he named. The Economic Times headline put it more bluntly: Samsung is opening up to multiple AI partners as it races Apple for a global edge.
Galaxy AI spent two years fastened to Google Gemini; the multi-model arrangement is now Samsung’s official position. For AI product teams this is not handset gossip: the moment the world’s largest Android shipper lets several models sit inside the native experience together, the model vendors’ channel fight spills over from APIs to device preloads — a distribution battleground larger than any since search engine defaults.
Roh’s Message to the FT
Choice is the pivot Roh keeps returning to. “We are open to all solutions … choice, I believe, is how Galaxy AI appeals to consumers,” he told the FT, before adding, “consumers are not bound to one AI platform, they are utilizing multiple AI models.”
The words stand on something concrete. Alongside Google Gemini, Perplexity’s AI search engine has recently been built into Samsung’s mobile OS. Multi-model is a present fact, not a roadmap item; Roh’s contribution was to spell the direction out: that door will not close, and companies with OpenAI’s heft are on the candidate list.
On Apple, his case is that Samsung moved first: it “got into the preparation earlier than others,” he said, and that head start is what sustains the mobile AI leadership he claims. The FT adds that Apple still has not delivered many of the AI features it announced the previous year — a gap Samsung plans to use.
From a Gemini Monopoly to Many Models
There is a traceable arc behind the change. Back in January, Samsung announced it would double the number of Galaxy AI devices to 800 million units this year, with Gemini from Google serving as the core engine. At that stage the blueprint had a single spine: Galaxy AI was, in effect, Gemini living on the device.
Three months on, the storyline differs. The Galaxy AI Roh now describes is a single surface over multiple models: Gemini takes some tasks, Perplexity owns the search side, and OpenAI or others may yet join. Google should read it as a warning — the exclusive berth it holds on the device is thinning out. For OpenAI it is a hard-to-get ticket: a native foothold on hundreds of millions of devices is a scaling route that runs beyond the ChatGPT app.
The demand side comes from PYMNTS’ own research: consumer AI adoption among U.S. adults hit 54% in January, five points above the prior month, and “mainstream” users — those who lean on AI as a low-complexity complement to search — account for slightly over one in three. Juggling several models is already ordinary consumer behavior; the data sits behind Roh’s read.
The Timing: Apple’s Window Plus the Memory Squeeze
Two layers sit beneath the timing. Competition comes first: Apple’s AI timeline has slipped, giving Samsung space to take over the category. Cost structure comes second: Counterpoint Research expects AI-related memory chip demand to drag global smartphone shipments into a downturn, with the squeeze set to run “well into 2027.”
With shipments under strain, the AI experience is one of the few selling points still able to hold up premiums and upgrade cycles. The calculation at Samsung: quit wagering on a single model’s capability curve and sell choice itself as the feature. The harder the models fight each other, the more leverage Galaxy holds as the carrier.
Takeaways for AI Teams
Three practical effects follow. First, device-level preloads turn into required training for model vendors — past API price and quality, whether a model can live in the device’s native layer becomes the next lever in distribution. Second, multi-model experience design is a new discipline in its own right: which tasks go to which model, how the work splits between cloud and device, and where the data and privacy boundaries fall between models all migrate out of backend architecture and into UX decisions. Third, the hardware resource war escalates — AI memory demand is squeezing the entire device supply chain, and even the strongest model counts for nothing on hardware that cannot run it.
Note what is absent: Samsung has committed to nothing specific, and Roh’s remark is a statement of direction rather than a contract. The direction alone still matters — smartphone AI in 2026 has formally moved from picking sides to serving a menu.