Faced with NVIDIA's 75% profit margin, AMD is under significant pressure!

Nvidia announced a 75.2% adjusted gross margin, reaching a recent high, but whether this unmatched profitability can be sustained is facing multiple pressures, including narrowing supply bottlenecks, the accelerated rise of self-developed chips, and the yet-to-be-realized returns on AI investments from customers.

Nvidia's latest quarterly report shows that as of the quarter ending in January this year, the adjusted gross margin reached 75.2%, the highest level since the second half of 2024, and the company expects to maintain a similar level in the current quarter. On the demand side, capital expenditures from large-scale AI companies are expected to total about $650 billion this year, an increase of approximately 60% compared to 2025, and Nvidia will benefit significantly from this—this point has been fully anticipated by the market before the earnings report was released, thus the real highlight of this earnings report lies in the profit margin rather than the demand itself.

The competitive landscape is changing. AMD announced this week that it has signed a "hundreds of billions of dollars" supply agreement for data center processors with Meta, directly impacting Nvidia's core GPU business; Alphabet's TPU chips and Amazon's self-developed chips are also accelerating their market share acquisition, with pricing significantly lower than Nvidia's products, making the increasingly cost-effective options drive more customers to explore diversified procurement.

Nvidia CEO Jensen Huang defines "sustained generational technological leaps" as the core lever for maintaining high profit margins and is optimistic about the demand growth driven by agentic AI. However, whether the massive expenditures on AI hardware can bring corresponding commercial returns to customers remains an unresolved question, and this is precisely the biggest variable affecting the sustainability of Nvidia's high profit margins.

Supply Tightness and Cost Pressure: Potential Hazards to Profit Margins

Nvidia's high profit margins do not come without a cost. The rising costs of memory are an unavoidable reality, even though Nvidia holds a priority position in the supply queue for key components.

According to Bloomberg, Nvidia CFO Colette Kress stated that the company has "strategically locked in inventory and capacity to meet demand for several quarters ahead," but also expects that supply "tightness" will persist.

Leading manufacturers of core components have warned that the shortage may continue until 2027 or even longer. The growth rate of AI hardware demand is still far exceeding the pace of corresponding infrastructure capacity expansion.

This means that while Nvidia is under pressure on the cost side, it still needs to provide sufficient supply to customers. Under this dual pressure, whether the 75% profit margin can remain stable in the coming quarters is quite uncertain.

The Rise of Alternatives: Competitors Competing on Price

The pricing differences are particularly significant. According to Bloomberg Industry Research data, the average selling price of Google's TPU is between $8,000 and $10,000 per unit, while Nvidia's H100 chip is priced above $23,000, with the newer Blackwell system starting at $27,000.

The price difference of more than twice makes the diversification of computing power procurement economically attractive.

In terms of specific transactions, the competitive situation is also accelerating in clarity. Meta has signed a "hundreds of billions of dollars" supply agreement for data center processors with AMD; in October last year, OpenAI reached a similar arrangement with AMD In both transactions, AMD has additionally offered some shares to enhance attractiveness.

On the Alphabet side, its TPU has taken on a large computing load for Google Cloud customers and its own AI services like Gemini, leading to a rise in its stock price after the news became clear. Amazon, on the other hand, has won the heavyweight client Anthropic with its self-developed chips.

The Return Rate Dilemma: Hundreds of Billions Invested Yet Not Translated into Revenue Growth

Nvidia's data center business recorded revenue of $62.3 billion, with slightly more than half coming from hyperscale cloud companies. This means that Nvidia's high profit margins largely depend on these customers' continuous and large-scale purchasing willingness.

Jensen Huang expressed optimism about this. "I am confident in their cash flow growth for a simple reason," he said, "We have already seen the inflection point of intelligent agents AI, and the practical value of AI agents in global enterprises is becoming apparent. The huge computing demand you see comes from this." He further stated, "In the new era of AI, computing power is revenue."

However, there remains a significant gap between reality and expectations: the massive computing investments from hyperscale cloud companies have yet to translate into visible revenue returns sufficient to justify their rationale.

If this return continues to be delayed, the market's willingness to continue purchasing high-premium chips will be tested. At that time, Nvidia's profit margins, which leave competitors in the dust, will become the first variable under pressure.

Nvidia's Moat: Dual Arguments of Versatility and Energy Efficiency

Jensen Huang reiterated this logic during Wednesday's analyst call: compared to the custom chips from Google and Amazon, Nvidia's GPUs can handle a wider range of AI-related tasks, rather than being limited to specific scenarios like model training or "inference" (i.e., running constructed AI models).

In the context of increasingly tight energy supply, Nvidia's progress in energy consumption optimization also constitutes a differentiated advantage.

Regarding profit margins, Huang provided his core logic: "The most important lever for our gross margin is actually the continuous delivery of generational leaps to customers."

At the same time, Huang stated, "We also love CPUs," emphasizing that Nvidia's CPU products will surpass competitors in data center scenarios and may become "one of the largest CPU manufacturers in the world."