Apple is using its silicon to power new AI features but relies on Google for training its models. As the AI competition intensifies, tech giants are forming unusual partnerships. Google faced a technical crisis after Apple requested more TPUs for AI training.
When Apple unveiled its new AI features, much attention was given to its partnership with OpenAI, which will bring ChatGPT to millions of iPhones—a deal Google had sought for itself.
Apple and Google have long collaborated behind the scenes, with Google providing access to its data centers to train Apple’s AI models. For years, Apple has used Google and Amazon’s cloud services to store data. For example, iCloud backups often reside in Google’s data centers, giving Apple access to vast computing resources that support many beloved iPhone features.
In April, Apple requested additional Tensor Processing Units (TPUs) from Google for AI training. TPUs are specialized chips for AI tasks, which Google leases as an alternative to Nvidia GPUs. This request caused a scramble within Google due to technical issues that nearly prevented timely delivery.
Internally referred to as an “OMG,” this critical incident led to the formation of a crisis team. Apple, nicknamed “Bigfoot” by Google Cloud staff due to its heavy usage of Google’s data centers, narrowly avoided major setbacks.
The incident highlights Apple’s lag in the generative AI race. Most advanced AI features rely on large, energy-intensive data centers built by companies like Microsoft, Google, and Amazon. Consequently, Apple depends heavily on these competitors as it advances in AI.
Apple’s documentation confirms its AI models are trained using a mix of TPUs and both cloud and on-premise GPUs. While much of Apple’s new AI functionality will run on devices, more demanding tasks will be offloaded to data centers with Apple-designed silicon, though their locations remain unclear.
Apple’s partnership with OpenAI will provide a more advanced chatbot, expanding OpenAI’s reach to Apple’s vast user base. The AI competition is driving these crucial, sometimes unexpected, alliances. For instance, Microsoft and Oracle recently partnered to give Microsoft access to Oracle’s cloud servers for OpenAI tasks. Google and Apple have also discussed bringing Google’s Gemini AI to iOS devices, though no agreement has been reached yet.
After the WWDC keynote, Apple executive Craig Federighi stated that Apple is open to various AI partnerships, even mentioning Google, and expressed interest in potentially integrating Google Gemini in the future.
🌐 Sources
- apple.com – Introducing Apple Intelligence for iPhone, iPad, and Mac
- hardwarezone.com.sg – Apple announces Apple Intelligence, its new AI system
- reddit.com – Reddit has struck a $60 Million deal with Google to Use its …
- economist.com – Large language models are getting bigger and better
About The Author
