Generated by Codex with GPT-5

This summary covers The Economist’s June 13th, 2026 Business article listed in the contents as Apple's second bite and published under the headline A second bite.

Apple’s first attempt to add generative AI to the iPhone was a conspicuous failure. Apple Intelligence, announced in 2024, was supposed to turn Siri into a capable personal assistant with access to a user’s messages, apps and other private information. Two years later, many of those promises remain unfulfilled. Apple is now trying again with a strategy that is less technologically ambitious but potentially more commercially effective: let Google supply the underlying AI models while Apple controls how people use them.

Apple is changing its role in AI

The redesigned Siri can be reached by voice, through a search bar or in a chatbot-style app. Unlike the earlier version, it will not depend mainly on models built by Apple. Google, already one of the leaders in frontier AI, will provide much of the core technology.

That decision turns Apple’s apparent weakness into a different sort of bet. Instead of spending heavily to compete with Google, OpenAI and Anthropic on model development, Apple can use its installed base of devices as the main gateway to their technology. It reportedly pays Google about \$1bn a year, far less than building a comparable model in-house would cost. Apple could also replace Google’s models later, strengthening its position in future negotiations.

This approach resembles Apple’s existing relationship with app developers. The company does not need to build every successful service if it owns the device, operating system and payment channel through which customers reach it. Even during its AI delays, Apple collected commissions of up to 30% on chatbot-app revenue. Its share price rose by more than half over two years despite its lack of a leading model.

The iPhone still gives Apple an advantage

OpenAI is developing an AI device with former Apple designer Jony Ive, while Google and Meta are investing in smart glasses and Amazon is adding AI features to Alexa. These projects aim to weaken the iPhone’s position as the centre of consumers’ digital lives. Apple nevertheless retains two important advantages.

First, Siri can work across a user’s messages and apps. That access may allow it to complete personal tasks that a stand-alone chatbot cannot. Second, Apple designs both its hardware and its chips. It can therefore run many AI features directly on a device, reducing delays, protecting privacy and keeping some functions available without an internet connection. On-device processing also limits Apple’s need to build the enormous data-centre capacity required by frontier-model companies.

There are trade-offs. Computationally intensive features, such as extending or reframing photos, will have daily usage limits, although iCloud+ subscribers may receive more. More fundamentally, Apple’s advantage depends on turning deep access to personal information into genuinely useful services without weakening users’ trust.

Execution remains the test

The new Siri is still not ready. It is expected to launch in America in the autumn, initially only in English. Regulatory obstacles will keep it off iPhones in the European Union and all Apple devices in China at first. Investors marked the announcement by sending Apple’s shares down 2%, a sign that another presentation is not enough after years of delay.

The rollout will mostly fall to John Ternus, Apple’s incoming chief executive and current hardware chief. The company’s deliberate pace has often produced polished products eventually, but AI is developing faster than the markets Apple previously entered late. Its strategy does not require Apple to invent the best model. It does require the company to prove that owning the consumer interface, personal context and hardware can matter more than owning the intelligence underneath.