🤖 AI Summary
Apple is reportedly close to a deal to pay Google roughly $1 billion a year for a custom version of Google’s Gemini model to power a major overhaul of Siri, Bloomberg says. The bespoke model would have about 1.2 trillion parameters—roughly eight times the size of Apple’s current cloud-based “Apple Intelligence” models (≈150 billion parameters)—and is being treated as an interim solution while Apple builds its own, more capable AI. Apple tested options from Google, OpenAI and Anthropic and chose Google’s offering; the revamped Siri is expected to debut next spring, though plans could still change.
For the AI/ML community this is notable on several fronts: it underlines the commercial value of large, customizable foundation models and shows Big Tech is willing to buy turnkey capabilities rather than wait to scale in-house solutions. Technically, integrating a 1.2T-parameter model implies significant inference-cost, latency and infrastructure considerations (cloud hosting, model optimization, safety/alignment tuning and privacy controls), and signals increasing demand for custom model variants rather than off-the-shelf APIs. Strategically, the move shifts Apple temporarily toward cloud dependence on a major competitor, while also validating the business model of selling tailored, high-capacity models to enterprise customers.
Loading comments...
login to comment
loading comments...
no comments yet