Track Hyper | Apple bets on Gemini: The financial arithmetic of the AI competition

Author: Zhou Yuan / Wall Street News

Apple is in discussions with Google about introducing the latter's Gemini large model into the new version of Siri.

This potential collaboration is both a technological choice and a financial arithmetic problem: how Apple balances the pace of innovation with capital market expectations, and whether Google can regain AI valuation dividends through external implementation.

The more intriguing suspense lies in whether Apple can catch up on experience gaps in the short term through external models without falling into strategic dependency. Capital markets, profit structures, and competitive landscapes are the dimensions that investors and the industry are more focused on.

Innovation Window and Capital Pressure

Apple started late in generative AI.

Compared to Microsoft's partnership with OpenAI and Google's full commitment to Gemini, Apple's self-research path appears inefficient, slow, and unambitious.

In the face of market concerns that "the iPhone has peaked," AI has become the most vocal breakthrough in the capital market.

Joseph Schumpeter proposed "creative destruction" in "Capitalism, Socialism and Democracy," reminding companies that if they cannot self-update in the technology cycle, they will ultimately be eliminated by the market.

Apple's anxiety about AI has actually existed since the low-key "birth" of ChatGPT 3.5, but progress has been slow and attitudes hesitant; introducing external models at this time is essentially another proactive delay against the threat of "destruction."

In this context, this cooperation will inevitably extend beyond mere technical issues.

Apple's R&D expenditure for fiscal year 2024 reached $31.37 billion, but only grew by 4.86% compared to $29.915 billion in fiscal year 2023 (which had a year-on-year growth rate of 13.96%), accounting for 8% of revenue (7.8% in fiscal year 2023).

How does the capital market evaluate such growth? The mainstream view is that Apple lacks "blockbuster results." The stagnation in Apple's GenAI technology progress and actions is triggering a wave of investors discounting Apple's valuation.

If Gemini is indeed embedded in Siri, what answer will Apple's financial arithmetic provide?

According to public reports, Google has begun training a customized Gemini model that can run on Apple's servers. On the surface, this ensures privacy compliance and alleviates data security concerns.

However, the impact at the capital market level is more complex.

On one hand, Apple needs to increase its computing power investment. According to Morgan Stanley's estimates, large-scale access to generative AI will significantly increase Apple's data center CapEx (capital expenditure) in the next two years. In the short term, this may compress Apple's gross margin.

On the other hand, the impact is relatively positive: essentially, this is a way for Apple to gain a "strategic option."

Compared to insisting on self-research, outsourcing ensures that Apple does not fall behind in the AI boom; compared to completely giving up, customized deployment maintains flexibility for future alternatives. Just like the "real options" theory in finance, Apple's move is to exchange limited costs for greater possibilities in the future.

The essence of business is win-win, so what is the capital value of this cooperation for Google? For Google, entering Siri is a potential revenue-sharing business.

The core challenge facing Gemini is the "insufficient application scenarios": Alphabet admitted in its 2023 financial report that the growth rate of search advertising revenue in 2023 was below the historical average, and the capital market is questioning how its AI can monetize.

If Gemini can become the underlying engine for Siri, Google will gain a natural entry point for hundreds of millions of devices.

This not only expands the model's call volume but also provides imaginative space for future subscriptions and service bundling. In other words, this is a key step for Google to transform "R&D expenditure" into "distribution dividends."

Multi-Model Strategy Investment Logic

It must be noted that in the process of embracing GenAI technology (which is still ongoing), Apple has not bet on a single partner.

Previously, Apple also engaged with OpenAI and Anthropic to assess the possibilities of ChatGPT and Claude. This approach is similar to its strategy in the supply chain: decentralized cooperation to maintain bargaining power.

In investment theory, this is precisely the idea of "diversified portfolios."

Faced with the uncertainty of technological evolution, Apple has reduced its dependence on a single partner by introducing multiple model vendors.

This also means that regardless of which large model runs faster, Apple can maintain a relatively safe choice of initiative.

The more pressing question for the capital market is: can this reignite the super cycle of the iPhone?

Apple's revenue structure is transitioning from hardware to services.

According to Apple's fiscal report for 2024, service revenue is expected to account for over 24% to 24.6% in fiscal year 2024, higher than the 22% share in fiscal year 2023.

If Siri evolves again after integrating Gemini, it is likely to give rise to more paid services and app subscriptions, making software services an important pivot for this transformation.

However, investors may also worry: if the external model calling costs are high, will it erode service gross margins? Especially in the context of the high gross margins of the App Store and iCloud, if AI computing costs remain high for a long time, it may put pressure on overall profit margins.

The capital market's attitude towards Apple's AI is nuanced.

Microsoft has been revalued as a "growth tech company" due to its partnership with OpenAI; Google, despite early skepticism about its self-developed Gemini, has at least proven its strategic focus. If Apple overly relies on external cooperation, will it be interpreted as "insufficient innovation"?

From another perspective, Apple's caution actually aligns with its financial discipline.

As management scholar Herbert Simon said, corporate decision-making involves "bounded rationality," and outsourcing is an extension of rational boundaries. Apple's choice of "first cooperation, then self-developed alternatives" is a typical risk-hedging behavior.

If Apple chooses Gemini, it will also face cross-border compliance and regulatory pressures.

The EU's "AI Act" is entering the implementation phase, requiring model interpretability and compliance with data source regulations. The FTC (Federal Trade Commission) is also paying attention to the potential market monopoly issues brought about by large models This means that when Apple introduces external models, it must ensure control over localization and transparency. Otherwise, once it touches the regulatory red line, it will not only face technical risks but also potential fines and reputational risks.

The ultimate answer is still unknown

If the cooperation materializes, Apple and Google will form a "soft alliance": a binding relationship between terminal manufacturers and AI platforms.

This could redefine the division of labor in the industry.

For hardware manufacturers like Samsung, Xiaomi, and Huawei, this trend is worth being cautious about. If Apple leverages Gemini to achieve a generational leap in smart assistant experience, it will force other manufacturers to accelerate cooperation with local or global AI models.

As a result, the industry may move towards a deeply bound pattern of "terminal + model."

There are also positive possibilities, such as Apple's long-term advantage in self-developed chips (M series and A series SoC), which means Apple does not have to rely on external sources for the long term. Once model inference can be completed more efficiently on the terminal in the future, Apple may completely switch back to a "self-developed closed loop."

Ultimately, this cooperation is not just about technology implementation, but also a test for the financial market: Can Apple maintain short-term competitiveness while avoiding long-term profit margin decline by introducing external models? Can Google find a more stable AI business model beyond advertising by entering Siri?

How should investors evaluate this? Is it a "stopgap measure" or the beginning of a "new narrative cycle"?

The answer is still unknown.

But one thing is certain: under the wave of generative AI, every move by Apple and Google is not only a choice of technical path but also a calculation of capital market expectations.

"Time is the scarcest resource for enterprises." This quote from management scholar Peter Drucker may be most suitable for Apple at this moment.

Capital power will not give Apple another minute, allowing Timothy Cook to hesitate again about whether to continue self-developing GenAI or leverage external forces; the market and investors demand immediate responses.

Apple's bet on Gemini means that under the dual pressure of finance and technology, it has chosen a "pragmatic arithmetic": ensuring it does not fall behind in the game of cost, innovation, and capital markets, and then seeking a long-term answer that belongs to Apple