AI

Samsung Opens the Door to Multiple AI Partners

Samsung device chief TM Roh told the FT on March 8, 2026 that Samsung is open to strategic co-operation with more AI groups such as OpenAI — a multi-model bet to outmaneuver Apple.

Samsung Opens the Door to Multiple AI Partners — article cover

On March 8, 2026, the Financial Times published an interview with TM Roh, Samsung’s consumer device chief, and the message came down to one line: Samsung is “open to strategic co-operation” with more AI groups, with OpenAI named as an example. The Economic Times headline said it more bluntly — Samsung is opening the door to multiple AI partners as it races to gain an edge over Apple globally.

Two years after tying Galaxy AI tightly to Google Gemini, Samsung has formally put “multiple models side by side” on the table. For AI product teams this is not smartphone gossip: when the world’s largest Android shipper decides several models will appear in the native experience at once, the model vendors’ channel war moves from APIs to device preloads — the biggest distribution battleground since search engine defaults.

What Roh Told the FT

Roh’s framing revolves around choice. “We are open to all solutions … choice, I believe, is how Galaxy AI appeals to consumers,” he told the FT, adding that “consumers are not bound to one AI platform, they are utilizing multiple AI models.”

This is not just talk. Beyond Google Gemini, Samsung recently integrated Perplexity’s AI search engine into its mobile OS. In other words, multi-model is not a future plan but a present fact; Roh simply made the direction explicit: the door stays open, and vendors of OpenAI’s caliber are on the list.

Against Apple, Roh’s argument is that Samsung started earlier. He said Samsung “got into the preparation earlier than others,” which is how it has maintained what he called leadership in mobile AI. The FT notes that Apple has yet to roll out many of the AI features it announced the previous year — a window Samsung intends to exploit.

From Gemini-Centric to Multi-Model

The shift has a traceable arc. In January, Samsung declared it would double devices running Galaxy AI to 800 million units this year, with Google’s Gemini as the core engine. The picture then was a single spine: Galaxy AI as Gemini’s on-device incarnation.

Three months later the story changed. The Galaxy AI Roh describes is a multi-model surface: Gemini handles some tasks, Perplexity covers search-shaped needs, and OpenAI or others may join. For Google this is a warning — its exclusive slot on the device is being diluted. For OpenAI it is a scarce ticket: native touchpoints across hundreds of millions of devices are a scaling path beyond the ChatGPT app.

PYMNTS added the demand side from its own research: consumer AI adoption among U.S. adults reached 54% in January, up five points from the prior month, and “mainstream” users — people who treat AI as a low-complexity complement to search — make up slightly over one in three. Consumers already juggle multiple models; Roh’s read has data behind it.

Why Now: Apple’s Gap and the Memory Headwind

The timing has two layers. The first is competition: Apple’s AI timeline is behind, leaving Samsung room to claim the category. The second is cost structure: Counterpoint Research forecasts a global smartphone shipment downturn driven by AI-related memory chip demand, with the squeeze expected to last “well into 2027.”

When unit shipments are under pressure, the AI experience is one of the few selling points that still supports premiums and upgrade cycles. Samsung’s calculation: rather than betting on one model’s capability curve, package choice itself as the feature. The harder the models compete, the more bargaining power the Galaxy has as the carrier.

What It Means for AI Teams

Three practical effects. First, device-level preloads become a mandatory skill for model vendors — beyond API price and quality, “can you live in the device’s native layer” becomes the next distribution variable. Second, multi-model experience design is a new discipline: which task routes to which model, how cloud and on-device work splits, and where data and privacy boundaries sit across models all move from backend architecture into UX. Third, the hardware resource war intensifies — AI memory demand is squeezing the whole device supply chain, and the best model is worthless on hardware that cannot run it.

Samsung committed to nothing specific; Roh’s statement is a direction, not a contract. But the direction alone matters: smartphone AI in 2026 has officially moved from picking sides to offering a menu.

Sources

AI-assisted summary compiled from the sources above, reviewed by a human before publishing.

FOUND_THIS_USEFUL?

Support more practical AI articles, tutorials, and build notes.

BUY_ME_A_COFFEE
SHAREXEMAIL