Nvidia says real demand can double, but it only dares to provide 70% guidance—why?
Last night, Nvidia released its Q2 earnings report for fiscal year 2027, and the numbers were explosive. Quarterly revenue hit $96.2 billion, up doubled year over year, with four straight quarters of acceleration. Even more aggressive: for the first time, Huang Renxun provided full-year forward guidance—fiscal 2028 revenue growth of 70%.
But he said one thing, and it left a deep impression on me. The real demand is far more than 70%—it’s just that supply can’t keep up, so they only dare to promise 70%.
That sentence carries a lot of information.
A company isn’t unwilling to sell more—it’s that capacity is bottlenecked. Cloud providers already have orders on hand stacked up to $2 trillion, and the Big Five’s capital expenditures next year are set to reach $1.3 trillion. They’re rushing to deploy compute power; how many chips Nvidia can ship depends on whether foundry output, HBM memory, and data center power can keep up. The entire supply chain is running at full capacity.
That is to say, the current bottleneck is not in demand, but in supply. This is a rare and comfortable position in the business world.
There’s another detail worth noting. The newly mass-produced Vera Rubin platform will become next year’s core growth engine, contributing about 20% of data center revenue just in the third quarter. The server CPU business is also hitting an inflection point—next year revenue is expected to more than double. Nvidia is shifting from selling GPUs to selling an entire AI factory.
Let’s look at gross margin. Because storage chip prices are rising too aggressively, the company’s gross margin will face short-term pressure. In Q4 it will bottom out between 71% and 72%, and then gradually recover throughout next year. Management explains the storage price increases as benign inflation—meaning prices are rising because industry conditions are favorable. That explanation holds up, because costs are going up while demand is also rising.
A few data points you should remember.
The commercial value per gigawatt of compute: from $18 billion in the Hopper era, to $25 billion for Blackwell, and then $40 billion for Vera Rubin. On the same piece of land and using the same amount of electricity, the revenue that can be extracted keeps doubling. That’s why customers are rushing to upgrade hardware—they’ve run the numbers, and the payback period on investment is now even less than a year.
There’s another easy-to-overlook point. Nvidia isn’t just selling chips. It has partnered with top-tier capital players like Blackstone and BlackRock to set up a revenue-sharing model, unlocking $500 billion in third-party funds to build AI factories. This shifts the story from one-time hardware sales to hardware plus long-term rent sharing. It’s a bigger picture: turning compute into real estate income.
Shareholders also haven’t been left out. In the second quarter, Nvidia returned a record $26 billion to investors, including $20 billion used for stock buybacks. The company said it would return more than half of free cash flow to shareholders; in reality, it returned 60% this year.
What does this mean for people in the crypto market?
AI infrastructure and crypto are getting more tightly intertwined. When compute is revenue, this narrative boosts overall risk appetite across the entire tech sector. Every time Nvidia beats expectations, the market’s imagination for future growth gets opened up again, and capital is more willing to move into high-risk assets.
But I need to pour a bit of cold water. No matter how beautiful the guidance is, most of the optimistic assumptions are already built into the pricing. A 70% growth rate is a conservative figure under supply constraints. The claim that real demand will double is just a verbal judgment, not a commitment. In investing, the most dangerous thing is taking management’s most optimistic line as certainty.
My view is very direct. This earnings report confirms one thing: the cycle of AI capital expenditures is far from over, and the supply shortage will run through at least all of 2028. But confirming a trend and confirming a buying opportunity are two different things.
Don’t obsess over a single headline—watch an entire cycle.