As the world watched, Deepseek's emergence in Hangzhou just before the Spring Festival sent shockwaves through Wall Street, leaving investors in a state of unease.

The reason for this turmoil lies in DeepSeek's unveiling of an open-source AI model that surpasses OpenAI's offerings across multiple metrics, all achieved with an R&D budget of under $6 million. In stark contrast, OpenAI's R&D expenditures hover in the billions.

Consequently, Deepseek's sudden rise to prominence had Wall Street grappling with questions: Have tech giants been misguided in spending billions on graphics cards to construct massive AI models? Is the practice of stacking computing power a mere fallacy?

The stock prices of major US tech giants plummeted in unison. NVIDIA took a 17% nosedive overnight, eroding over 4 trillion yuan in market value. Other heavyweights like Broadcom, AMD, Microsoft, and TSMC also experienced steep declines. Many speculated that the conventional wisdom for valuing AI had potentially lost its footing.

These AI chips might not hold the value once envisioned, and even ancillary sectors within the AI landscape, such as power suppliers, have felt the ripple effects.

Many anticipated that future AI innovations would pivot around the Deepseek model, prioritizing cost-effectiveness and efficiency. However, Musk has opted to chart a different course.

In a move that bucks the trend, Musk's xAI company has introduced Grok 3, a colossal model that has emerged victorious in both LMSYS blind tests and AIME competitions, outpacing its rivals. Musk heralds it as the world's smartest AI.

The hallmark of Grok3 lies in its computing prowess, achieved through the stacking of 200,000 H100 graphics cards, embodying the principle that "great efforts yield extraordinary results."

Grok3's success underscores the enduring validity of the scaling law. Put simply, the more graphics cards stacked, the more potent the capabilities of the large model become. For those with the means and resources, there's no need to "reinvent the wheel"; simply stack the computing power.

Thus, despite Deepseek's emergence, it's crucial to recognize that consistent investment in chips, data centers, and cloud infrastructure remains an unwavering truth. While exceptional results can be achieved with limited resources, computing power serves as the bedrock upon which AI stands, a foundation that cannot be ignored.