I just recently became interested in something quite intriguing from NVIDIA’s financial report last month. On the surface, everything looks perfect—Q4 revenue reached USD 68.1 billion with 73% year-over-year growth, net profit of USD 42.96 billion, and guidance for the next quarter even raised to about USD 78 billion. But what’s tantalizing is that the market reaction is actually negative. NVDA’s stock price fell by about 5.46% after the announcement, and the market value evaporated by around USD 260 billion in a single day. This isn’t a simple case of “strong fundamentals but weak price”—it’s a shift in the valuation curve that’s happening right before our eyes.

The problem isn’t with NVIDIA’s performance itself. If we look deeper, Q4 data center revenue reached USD 62.3 billion out of a total USD 68.1 billion—meaning about 91.5% of all revenue comes from a single segment. This is extreme concentration. During a bull market phase, this kind of focus is seen as extraordinary efficiency. But once sentiment changes, the same focus instantly turns into a risk. If AI capital expenditure from cloud providers and enterprises starts to slow down, NVIDIA’s volatility will explode in the opposite direction. Other businesses like automotive, gaming, and professional visualization? They’re too small to cushion the shock—automotive revenue per quarter is only about USD 604 million, nowhere near enough to offset fluctuations in the AI capex cycle.

But wait—there’s another deeper layer. NVIDIA’s customer concentration is also increasing drastically. Just two customers account for 36% of total sales in fiscal year 2026. This means NVIDIA’s spectacular growth is actually heavily dependent on a small group of big players. When they expand, NVIDIA can extract a serious “monopoly tax.” But once they start slowing down—or, and this is the most dangerous part, start systematically supporting alternative suppliers or developing their own internal solutions—that monopoly premium will be eroded into nothing more than a normal leadership premium.

The key phrase here is the shift in the curve from “quarterly focus” to “growth duration.” Investors are no longer only concerned about whether NVIDIA can keep beating expectations quarter after quarter. They’re now asking: how long can this momentum last? Under what kind of business structure? And in what competitive environment?

AMD and Meta have already started sending signals through their long-term collaborations. This isn’t about immediately shifting market share, but about demonstrating that major customers are diversifying their suppliers. The consequence is a gradual decline in NVIDIA’s pricing power. Meanwhile, the AI industry is moving from a wasteful training phase to an inference phase that’s sensitive to cost. Here, new players with more specialized architectures are starting to find openings. NVIDIA is trying to close this gap through acquisitions of talent and technology in inference, but the battle right now isn’t just about raw chip performance—it’s about end-to-end system efficiency.

What’s interesting is that NVIDIA is building what you could call a “second curve.” They’re not just selling GPUs anymore. Look at their initiatives in autonomous vehicles, robotics, and industrial simulation—what they call AI fisik. They’re also rolling out open-source platforms and tools for autonomous driving. Near-term contributions from these areas are still limited, but the long-term strategy is clear: lift NVIDIA from being merely a “shovel seller” to becoming a “provider of operating system-level solutions.” Once this transformation succeeds, NVIDIA’s growth duration will no longer be determined entirely by cloud providers’ capex cycle; instead, it will be driven by long-term demand from industrial digitalization, industrial robots, and autonomous vehicles.

But before this second curve truly scales, the market will still use the old valuation framework: “single machine data center + cyclical capex asset.” So for 2026, there are three curves that are actually far more important than a single income statement. First, does the capex velocity from cloud providers continue to accelerate, or has it started to plateau? Second, can the transformation from “selling GPUs” into “selling complete system solutions” continue to increase customer stickiness and value per customer? Third, how quickly are second suppliers and in-house solutions moving from the pilot stage to large-scale procurement?

So in conclusion, NVIDIA’s financial report shows that the AI infrastructure wave is still ongoing. They remain the strongest cash-flow engine in the compute power sector. But the decline in the stock price reminds us that once “beating” becomes normal, pricing logic has shifted from growth to sustainability—from earnings to the duration of growth. NVIDIA is still strong, but the real test is the durability of this growth and whether its business structure can become more resilient. The answers will determine NVIDIA’s valuation ceiling throughout this year.