$NVIDIA(NVDA.US) has been on a continuous decline recently, and last night it dropped a whopping 7%. The recent sharp drop is indeed related to tech giants like Google and Meta coming out to say that their and their peers' AI investments may be overdone. However, even so, in the face of a major technological cycle, these giants have a 'better too much than too little' determination.

The real trigger for last night's plunge was actually a research paper from Apple reported by foreign media. The core of the paper is that Apple's two large models—Apple's on-device model and cloud-side model—were pre-trained using Google's TPUs, while inference was performed using Apple's own M2 Ultra clusters.

Specifically, the on-device servers used Google's 2048 TPUv5p, while the cloud-side model used 8192 TPUv4. The research paper did not mention any use of Nvidia's GPUs. Google's TPUs, being dedicated servers, were indeed developed earlier, are more mature, and offer relatively good performance.

As a result, the market began to worry: Is Apple, the biggest player in downstream AI applications, planning to bypass Nvidia by diversifying its supply chain internally and externally to avoid Nvidia's high 'tax'? If Apple sets this precedent, will other giants follow? The result was Nvidia's stock plummeting 7% in a single day.

However, another version of the story is that Apple didn't seek out Google's computing power (PS: Unlike Nvidia, which sells GPU hardware directly, Google sells computing power through subscriptions to its cloud platform services, requiring customers to connect via software to access Google's computing power) to pre-train its models. Instead, Apple had previously poached an AI team from Google and, unable to procure Nvidia GPUs in time, used the TPU architecture they were familiar with. Later, due to insufficient computing power, they did purchase H servers.

In the long run, though, Apple aims to achieve greater computing autonomy with its upcoming self-developed M5 Ultra chips, manufactured by TSMC.

Overall, Nvidia's valuation is indeed high, and with CEOs of major companies admitting that current AI investments may be excessive, Nvidia faces downward pressure. However, the timing of this news feels like adding fuel to the fire for Nvidia.