Samsung might be firmly in Google's software orbit now, but the company has previously spent years trying to build software independence rather unsuccessfully. Tizen, Bixby, Samsung Pay, Samsung Internet, Email, etc were all bricks in a walled garden Samsung was trying to build.
Take Bixby, for example. It was meant to deny space to Google Assistant on the company's devices while also being positioned as the only true rival to Apple's Siri. That's not how the story of Bixby has played out and after a few years in limbo, Samsung is now relying on Perplexity to bring Bixby out of obscurity.
Some of those efforts have worked out, others haven't, not because they were fundamentally flawed, but because Samsung can never have a clean break from Google's ecosystem. Everyone, including Samsung, tried to build an alternative to Android but none exists and it's highly unlikely that one would emerge now.
Apple has more flexibility when it comes to any potential collaboration with Google whereas Samsung, like all other Android OEMs, is obligated to ship Google apps and services on their devices for the privilege of licensing Android.
It's understandable why Samsung made those efforts. Everything that it went through to become the world's top Android manufacturer has helped Google extract enormous value from the Galaxy ecosystem. That holds true even more so for AI as Samsung has effectively solved Google's mobile distribution bottleneck for Gemini.
Google's own Pixel devices don't come close to the sheer volume of Galaxy devices that Samsung pushes across the globe every single year. So when Google needed to fine tune Gemini in the real world, across languages, networks, demographics, use cases, Samsung's customer base was the only place that could provide that at scale.
Every Gemini query, interaction, UI friction point, etc is a training signal that feeds the machine. Samsung's customers are doing the work and the net benefit of that flows back to Google. Samsung's devices have effectively become the world's largest and most diverse AI testing ground, without any control over the direction Gemini takes in the future.
Apple tried to scale the mobile AI mountain on its own but this proved to be more difficult than it may have imagined. While its users were waiting on the promised highly personalized iteration of Siri, Apple announced in January this year that its next-generation AI models will be powered by Google's Gemini models, which will also specifically power the new version of Siri.
This deal did have a privacy carve-out, in that the Gemini-powered Apple Foundation Models will continue to run on Apple's Private Cloud Compute.
Realistically, this means that while training data flows from Samsung devices to improve Gemini across all platforms, potentially even on Apple, similar data may not flow back from Apple as it runs a gated implementation of Google's AI model.
Google is the ultimate beneficiary. It has an absolute chokehold on Android and the platform's largest vendor, Samsung, doesn't have a robust mobile AI offering without Gemini. Apple, the biggest rival of its platform, has essentially admitted that even it requires Gemini to achieve its vision of mobile AI.
From a purely implementation standpoint, Apple comes out on top, too. Gemini branding probably won't feature too prominently in Apple Intelligence so for the end user, the improved AI experience would just be another Apple highlight. Try explaining to an average user what AI foundational models are.
Samsung built the Galaxy into the world's most recognizable Android brand through sheer hardware excellence and relentless scale. But scale, it turns out, is a resource that others can spend. Google spent it to distribute Gemini. Apple spent it to quietly benchmark what good mobile AI looks like before shipping its own.
Samsung, for its part, has ended up exactly where it started, making the best possible version of someone else's vision. The walled garden it tried to build is now someone else's orchard.
Post a Comment