How ChatGPT-maker OpenAI is left as the biggest loser in Apple’s AI partnership with Google

Apple's decision to partner with Google for AI just handed OpenAI an unexpected reality check. The iPhone maker confirmed Monday it's using Google's Gemini models to power its revamped Siri and Apple Intelligence features—a choice that speaks volumes about where the AI race actually stands."After careful evaluation, Apple determined that Google's AI technology provides the most capable foundation," the companies said in a joint statement. The multi-year deal means Apple's next-gen AI features, including a smarter Siri launching this spring, will run on Google's technology and cloud infrastructure.Apple says nothing's changing with its existing OpenAI arrangement, which lets users tap ChatGPT for certain queries.

But that feels like polite corporate speak. When Apple needed to pick the actual brain behind Siri's biggest upgrade in years, it chose Google. OpenAI's role now looks less like a strategic partnership and more like a helpful side feature.The decision matters because Apple doesn't make these calls casually. This is a company that killed AirPower after announcing it because the charging mat wouldn't work reliably.

It delayed the HomePod. It spent nearly a year searching for the right AI partner before committing. When Tim Cook tells employees "Apple must do this" about AI, as he did at an all-hands meeting last August, the company takes it seriously.

Apple's messy path to admitting it needed help

This wasn't how Apple wanted things to go. The company intended to use its own AI models but kept hitting walls. At WWDC 2024, Apple demoed an AI-powered Siri that could understand personal context and take actions across apps.

The presentations looked sharp. Behind the scenes, the technology wasn't holding up.Software chief Craig Federighi later explained what went wrong: "We initially wanted to do a hybrid architecture. We realised that approach wasn't going to get us to Apple quality." The challenge was trickier than expected—merging everyday smartphone tasks like setting alarms with sophisticated language models that could truly understand what users wanted.By March 2025, Apple publicly delayed the Siri upgrade. "It's going to take us longer than we thought to deliver on these features," the company admitted. Federighi told The Wall Street Journal the team eventually concluded "this just doesn't work reliably enough to be an Apple product."The struggles had consequences. Apple removed AI chief John Giannandrea from Siri development, reassigning the project to Federighi and Vision Pro creator Mike Rockwell.

Meanwhile, Apple's Foundation Models team—the group building on-device AI for features like text summaries—started unraveling.Chief architect Ruoming Pang left for Meta in July, taking a $200 million package and a senior role in the company's Superintelligence Labs. Several colleagues followed. Others who stayed began interviewing elsewhere, either worried about Apple shifting to outside technology or tempted by the massive compensation packages competitors were offering.The talent drain signalled what was coming. Apple wasn't going to build this alone.

The competition that OpenAI quietly lost

Apple didn't just wake up one day and call Google. The company ran a methodical evaluation process starting last summer. Federighi and Rockwell, along with Apple's corporate development chief Adrian Perica, met with Anthropic, OpenAI, and Google about potentially powering the new Siri.Apple built two internal versions to test. One, codenamed "Linwood," used Apple's own models.

Another, called "Glenwood," ran on external technology. Both underwent stress testing against Apple's quality standards.Anthropic looked like the frontrunner early on. Its Claude model impressed Apple's engineers with strong performance and safety features. But Anthropic's pricing proved too aggressive. The demands pushed Apple to widen the search beyond its initial favourite.Google started training a custom Gemini model over the summer that could run on Apple's servers.

The search giant had relevant experience—it already powers much of Samsung's AI functionality on Galaxy phones. More crucially, Google had financial stability and infrastructure that Apple valued.OpenAI participated throughout. The company had existing momentum from its ChatGPT integration in iOS 26, announced at WWDC 2024. Users could already access ChatGPT through Siri for complicated queries. Apple had even aired commercials promoting the feature.But OpenAI couldn't seal the deal. Its GPT-5 release in August underwhelmed users who complained about a colder tone and struggles with straightforward questions. OpenAI spent months adjusting the model to make it warmer and more reliable—exactly the kind of inconsistency Apple was trying to avoid by partnering with someone in the first place.

When Google's momentum met OpenAI's crisis

The timing makes this harder for OpenAI. Google released Gemini 3 in November, and the model topped industry benchmarks.

The launch proved Google's AI bet was paying off. Alphabet's stock surged, climbing 65% across 2025, outpacing other big tech companies.CEO

Sundar Pichai

had been pushing his teams relentlessly. The effort showed—Google's cloud division locked in more billion-dollar deals through Q3 2025 than it had managed in the previous two years combined. Google was also making headway on custom AI chips, creating energy-efficient alternatives to Nvidia's power-hungry processors.OpenAI was moving in the opposite direction. CEO

Sam Altman

issued a "code red" memo to employees in December, acknowledging ChatGPT needed urgent fixes. He identified problems with personalisation, speed, reliability, and the range of questions the chatbot could handle properly.Altman postponed other projects—advertising features, AI agents for health and shopping, a personal assistant called Pulse—to concentrate resources on fixing ChatGPT's core experience.

The company started holding daily calls for teams working on improvements. That's the kind of emergency response that happens when you know you're losing ground.The contrast was stark. OpenAI burns over $100 billion annually without profits, constantly raising new funding rounds. According to its own projections, the company needs roughly $200 billion in revenue by 2030 just to break even. Google funds its AI work from advertising and cloud revenues.

When the race demands relentless spending on computing power and talent, that financial cushion matters.

Why Apple's choice cuts deeper than revenue

Neither company disclosed financial terms, though Bloomberg previously reported Apple would pay around $1 billion yearly for Gemini access. That's significant but not unprecedented—Google already pays Apple an estimated $20 billion annually to remain Safari's default search engine.Apple emphasised that Gemini will process data on its own devices and Private Cloud Compute servers, keeping user information away from Google's infrastructure.

The approach mirrors how Apple handles other cloud services, using Mac chips in data centers to maintain privacy standards.The improved Siri should arrive with iOS 26.4, which is expected to drop in March or April, finally delivering features Apple first demoed at WWDC 2024. Users will ask Siri complex questions that tap personal context—like finding a podcast a friend recommended weeks ago by searching through old texts and emails.For OpenAI, losing this deal stings beyond the revenue. It's a signal that even with 800 million weekly ChatGPT users, the company isn't considered the most reliable AI partner when the stakes matter most. Apple could have chosen OpenAI to avoid deepening ties with Google, a direct smartphone rival. It didn't.The decision also complicates OpenAI's path to profitability. Enterprise customers are important, but consumer deals with companies like Apple provide scale and steady revenue.

Losing that opportunity to Google makes the financial math harder.There's irony here too. At WWDC 2024, Federighi told YouTuber iJustine that Apple eventually wanted to let users choose between AI models. "We think ultimately people are going to have a preference perhaps for certain models," he said. "We may look forward to doing integrations with different models like Google Gemini in the future."Cut to two years later. OpenAI found itself watching from the sidelines while Google took center stage. For a company that dominated AI headlines just two years ago with ChatGPT's launch, the shift is sobering. The AI race isn't slowing down. And right now, OpenAI is learning what it feels like to lose ground to a competitor with deeper pockets and a longer runway.