Apple and Google have officially unveiled a 'multi-year agreement,' marking a significant move in the tech landscape. Under this deal, Apple will incorporate Google's Gemini large language model and Google Cloud technology into its Apple Intelligence suite and Siri voice assistant. According to reports from Reuters and The Verge, Gemini is set to become the cornerstone large model for AI functionalities on Apple devices. However, Apple will persist in developing its proprietary large model, with the specific division of responsibilities between the two entities yet to be fully delineated. Bloomberg's report, suggesting that Apple will compensate Google with $1 billion annually, is likely a gross underestimation. Should billions of Apple users globally adopt Gemini through Apple AI, the annual value generated from token consumption could far surpass the billion-dollar mark. The $1 billion figure might merely represent a baseline.

OpenAI, undoubtedly the most disheartened party in this scenario, should have anticipated this development. Since its initial collaboration announcement in 2023, OpenAI has been Apple AI's primary partner, leveraging a blend of 'locally developed small-parameter models + GPT large-parameter models' to cater to user demands. Nevertheless, it's undeniable that Apple AI's performance has left users in most regions worldwide largely dissatisfied. The blame game between Apple and OpenAI is futile; what matters is that Apple has opted to bring in a more seasoned and ecologically integrated partner to rectify the situation.

This decision effectively extinguishes OpenAI's most promising consumer-side potential. By the end of 2025, ChatGPT boasted 1.2-1.5 billion monthly active users (MAUs) and 40-50 million paying subscribers (excluding those paying through third-party APIs)—a substantial user base, yet dwarfed by Apple's colossal reach. Theoretically, if Apple AI had excelled, GPT could have become the 'AI brain' of Apple phones, with Apple's smart devices serving as a ready-made distribution channel for GPT, yielding substantial and sustained revenue streams. Alas, this opportunity now belongs to Google.

Prior to this decision, Apple's management deliberated over various options due to Apple AI's prolonged underperformance. In reality, only four viable paths were apparent:

Maintain trust in OpenAI, believing that time would resolve the issues.

Accelerate the development of its proprietary large model, relying solely on internal resources.

Collaborate with other AI startups like Anthropic or xAI.

Forge a partnership with Google.

The second option was swiftly dismissed—Apple's investment in computational infrastructure and large model training over the past three years has been minimal. While all Silicon Valley giants have been burning through cash, Apple remained on the sidelines, waiting for the AI industry's fruits to ripen and fall into its lap. In generative AI, Apple's accumulation lags significantly behind Google, Amazon, and Microsoft. Without increased investment, the gap will only widen. Apple is not neglecting foundational R&D; its focus lies elsewhere, such as on self-developed chips.

The third option seemed plausible but carried substantial risks—Anthropic's Claude and xAI's Grok occasionally outperform GPT on some large model benchmarks, but their foundations remain shaky, with incomplete ecosystems. Their survival prospects against OpenAI are highly questionable. Should they be acquired by larger firms in a few years, Apple would face new challenges.

This left only the first and fourth options. In my view, Apple's decision was driven by the impressive performance of Gemini 3, released in November of the previous year. Gemini 3 Flash, released in December, achieved comparable results to Gemini 3 Pro at a lower cost, addressing Apple's primary concern about inference costs. Indeed, Google has led the world in reducing inference costs, with its deep reasoning model launched in early 2025 comparable in cost to the groundbreaking DeepSeek R1, albeit overshadowed by the latter's acclaim.

We can say that Gemini 1 (and its predecessor Bard) was a lackluster large model, nearly prompting Wall Street to issue Google's death sentence. Gemini 1.5 marked a comeback, securing Google's ticket to the AI era. Gemini 2 and Gemini 2.5 continuously reclaimed lost ground, restoring Google to normalcy and boosting Wall Street's confidence. Gemini 3, however, was a decisive large model, proving Google's foundational large model capabilities to be world-class, potentially on par with OpenAI.

Within two months of Gemini 3's release, two major events unfolded: first, Apple chose to partner with Google, as discussed in this article; second, Google announced collaborations with retailers like Walmart to launch AI shopping guide services. This fully demonstrates that in the AI large model arena, Google has entered a virtuous cycle of 'technology-users-revenue' and may be the first among major tech firms to generate substantial revenue from AI ToC applications. Including TPU sales to third parties, my estimate is that starting in 2026, direct revenue from generative AI could account for a double-digit percentage of Google's total revenue.

Meanwhile, Google's competitors face immense pressure—the launch of Gemini AI shopping guides immediately pressured Amazon; Apple's shift to Gemini has placed tangible pressure on OpenAI. Over the past few months, OpenAI has been aggressively signing computational infrastructure contracts, leaving an impression of 'flashiness without substance.' However, this is precisely a 'defensive offense' move after recognizing the danger. With Google catching up in foundational large model capabilities and far surpassing OpenAI in resources and ecosystems, OpenAI can only respond by heavily investing and signing partnerships to quickly bring more allies onto its side. Even so, the delaying effect on Google's offensive remains limited.

The relationship between Apple and Google is intricate: they are fierce rivals in the smartphone arena but maintain close cooperative ties elsewhere. Google pays Apple substantial fees annually to promote its search engine, while Apple rents significant Google Cloud services. This collaboration is even more delicate: Gemini has already been integrated into Android, and Apple's move effectively declares that 'AI experiences for iPhone users will not be superior to those of Android users and may even lag behind in updates.' This severely damages the premium image iPhone has cultivated over the past decade.

Despite such significant risks, Apple's decision to partner with Google underscores the urgency of the situation, leaving little room for hesitation. Had OpenAI or another large model developer inspired the same confidence in Apple, the latter likely would not have relied on Google for such a critical matter. This represents a priceless vote of confidence in Google's AI technology. Google's investors are probably already celebrating, aren't they?

As for the 'Apple self-developed large model' rumored to continue development, few likely hold high hopes at this juncture. Had Apple's CEO been Steve Jobs, he would have undoubtedly committed to self-developed large models in November 2022 when ChatGPT emerged. In that case, Apple's computational procurement momentum over the past few years might have been among the strongest globally (and would have propelled NVIDIA's market value to new heights even faster). However, Tim Cook is not Steve Jobs; his strengths lie in supply chain and cost control, making it difficult to decide to invest in an uncertain emerging technology at critical moments. By the time he saw clearly, it was already too late to invest.

Everyone is human, not divine, with strengths and weaknesses. Over the past fifteen years, Apple has fully enjoyed the benefits of Cook's supply chain mastery but must now bear the side effects of his excessive caution; the reverse also holds true. Only deified figures are flawless, but such beings exist only in fantasy novels or 'factual' fantasy novels.

So, compared to Google, which has overcome nearly all challenges and possesses advantages across AI full-stack technologies, what cards does OpenAI still hold? Interestingly, its ace has become 'user habits.' Currently, ChatGPT retains a comprehensive user advantage over Gemini:

MAU: GPT 1.2-1.5 billion, Gemini APP 650 million; but including core Google users utilizing Gemini features may exceed 1 billion.

Paying users: GPT 30-40 million, Gemini unclear but generally considered lower.

API: GPT likely has over 1 million developers, Gemini around 500,000 (420,000 in the second half of 2025).

OpenAI still wields significant influence in both consumer and developer communities. After all, it took Gemini over two years to largely close the gap, during which time GPT had ample opportunity to cultivate user habits. Personally, I have continuously paid for GPT for 21 months; even if I switch to becoming a Gemini paying user, I likely will not abandon GPT immediately. The same applies to professional and enterprise-level users who have long relied on GPT.

Reality is richly dramatic: just over three years ago, OpenAI was the disruptor shaking the world with technological innovation, while Google was the entrenched incumbent clinging to user resources. Now, their roles seem reversed. The situation has changed too rapidly; who knows what it will look like in another three years?

This article has not received funding or endorsement from Google, OpenAI, or any of their affiliated parties.

The author does not directly hold equity in any of the companies mentioned in this article but may indirectly hold stakes through funds or trust plans.