After DeepSeek, It Seems Apple May Have Had the Best AI Strategy

The competition in the field of artificial intelligence (AI) is becoming more intense. Many of the leading tech companies, often called the Magnificent Seven, are pouring vast amounts of money into developing the best AI models quickly.

This aggressive strategy has led to questions about the return on investment. Will these significant expenditures ultimately pay off?

Recently, a group of significant AI companies in the U.S. announced a massive $100 billion project named Stargate. Shortly after, a cost-effective Chinese model called DeepSeek R1 emerged, highlighting that cutting-edge AI models can be trained without breaking the bank.

The launch of DeepSeek brings a surprising insight: it may have vindicated Apple's AI strategy, which faced criticism in recent years. After all, the company seems to have a well-structured plan and could end up ahead in the long run.

The Implications of DeepSeek

There is some debate about the financial investments made by DeepSeek in its model, whether it derived its model from the leading OpenAI's o1, or whether it has more GPUs than it publicly admits. However, the optimizations detailed in its research paper display innovative methods to reduce the costs associated with training advanced models.

Investors can expect a few outcomes from this development. Firstly, look for these new ideas to be rapidly adopted by major competitors. This trend may drive down costs for AI models across the board.

Secondly, DeepSeek's model is partially open-source, meaning that developers can access its optimizations and algorithms. Among American tech firms, only Meta Platforms' Llama models are similarly open-source.

The availability of major American and Chinese companies disclosing their model advancements could encourage more open-source initiatives, further accelerating the development of large and effective models.

While OpenAI still holds a technological edge, that difference has decreased to mere months compared to its open-source rivals. Meta CEO Mark Zuckerberg recently stated that he believes their Llama 4 models will outperform OpenAI's offerings this year.

The increasing trend toward open-sourcing along with algorithm-driven cost efficiencies is likely to lead to commoditization in AI. This means that prices for AI models will likely decrease, benefiting end-users but harming the companies creating these models.

Apple Was Prescient

As the AI competition intensified over the last couple of years, most significant tech companies dramatically increased their capital expenditures to acquire expensive Nvidia GPUs and expand data center capabilities. However, one major tech player, Apple, opted out of this costly spending spree.

Apple actually reduced its capital expenditure over the last two years, contrasting sharply with its competitors, whose expenditures soared by 73% to 182% in the same period. This places Apple in a unique position.

When a resource becomes abundant and commoditized, it's wise to act as a user of that resource rather than a producer. Apple's strategy of focusing on utilizing AI to serve its core customer base, while avoiding expensive Nvidia-based infrastructure, turns out to be a smart move.

Apple Is Still Investing in AI

It's important to note that Apple is indeed investing in its proprietary models. In June, Apple introduced its models, aimed at specific applications for the iPhone and Mac. During a recent conference call, CEO Tim Cook mentioned some of the most used features since the introduction of Apple Intelligence last October.

Apple Intelligence users benefit from various tools, such as Writing Tools for finding the right words, Image Playground and Genmoji for creating images, and a more interactive version of Siri for daily tasks. Additionally, users can create movies of their memories with simple prompts and improve their photos with Clean Up. New camera control features enhance understanding of surroundings.

In contrast to focusing on generic models, Apple has designed its approach with certain customer use cases in mind. Apple Intelligence's base model is built on an open-source framework known as Apple's AXLearn. This allowed Apple to avoid Nvidia chips and its proprietary CUDA software stack. Apple understood the risks of relying too much on Nvidia’s costly ecosystem.

While Apple's AI software framework is open-source for its core model, the company has chosen to use only licensed data, allowing publishers to opt out if they wish. Additionally, Apple has its proprietary web crawler to gather publicly available data. After training, Apple employs proprietary filtering and optimization processes tailored to specific tasks on its devices.

This careful strategy enables Apple to create smaller yet highly useful models for particular purposes at an economical cost.

Apple's Discipline Makes It a Long-term Winner

In the wake of declining stock prices for Nvidia after the DeepSeek announcement, Apple has regained its status as the largest company in the world.

Despite being currently the world's biggest company, Apple has historically never been the first to invent technologies. It did not invent the PC, portable music player, smartphone, or wireless headphones. Instead, Apple has excelled at improving existing technologies to make them intuitive and user-friendly, while also optimizing for costs.

This disciplined approach, coupled with a reluctance to overspend on uncertain innovations, contributes to a low-risk business model. Since large language models are set to become cheaper and more widely available, Apple's disciplined path emphasizes why it has been a favorite choice among investors like Warren Buffett for the last decade.

AI, Investment, Technology, Strategy, Apple