NVIDIA delivered again this week. $81.6 billion in revenue, 85 percent more than a year ago. The data center business grew 92 percent and contributed $75.2 billion to the total. Earnings and cash flow at record levels. On top of that, a new share buyback program worth $80 billion, which together with the existing program adds up to roughly $119 billion. The market didn't cheer. But it wasn't disappointed either. The stock remains just below its all-time high. Business as usual.

The AI powerhouse is currently valued at around $5.31 trillion. Operating cash flow in the last quarter came in at $36.2 billion, which annualizes to roughly $145 billion. Market expectations for fiscal year 2026/27 sit at around $190 billion, rising to over $250 billion and nearly $300 billion for 2027/28 and 2028/29. That puts the P/CF ratio at around 28 for the current year, dropping to 21.2 and 17.7 in the following years. Not an absurd valuation for a company growing at this pace. But no margin of safety either. I would be buying on expectations that need to materialize over two, three years. That's too uncertain for me. Though I'm well aware of the opportunities.

One I see in China, which is currently absent from the results. The country has been effectively cut off as a customer for AI chips due to US export restrictions. Yet business is booming. For the current quarter, NVIDIA expects zero revenue from the Chinese data center business, and still projects new record revenues. That makes the recent results all the more impressive. The rest of the world is not just compensating for the shortfall, it's surpassing it. And perhaps more importantly: China no longer plays a role in expectations, so it can't disappoint. Could there even be positive surprises? Jensen Huang was part of the US delegation at the Trump-Xi summit in May. Following the meeting, around ten Chinese companies were approved to purchase NVIDIA's H200 chip. Nothing has been delivered yet, legal questions remain open. But the direction could change.

Then there's Vera Rubin, the next chip generation, set to go into production in the third quarter. NVIDIA is also moving into the CPU market, currently dominated by Intel and AMD. Huang expects nearly $20 billion in revenue from stand-alone CPUs this year alone. The strategy is clear: the industry leader is systematically extending its reach into every corner of AI infrastructure. So far, so good.

What I see at NVIDIA, however, is a habituation effect. When record results become routine and everyone is optimistic, a single negative surprise is enough to catch many investors off guard. Just think back to the DeepSeek moment of January 2025. Sudden fears of declining demand for computing power triggered a correction of over 40 percent from the peak. In hindsight that turned out to be a good entry point, but as an investor you have to be able to stomach those drops first. I consider the risk of further volatile moves real. And that is the most important reason why I don't see NVIDIA as an investment. Because the broader uptrend still appears intact, the stock is, for me, only a trading vehicle. That means small positions and a risk management approach that kicks in at the first signs of weakness.

Is NVIDIA a buy? At current prices, NVIDIA is not a buy for me. The stock trades at a P/CF of around 28 for the current fiscal year, leaving no margin of safety. I would be buying on expectations that need to materialize over two, three years, and that's too uncertain for me.

Disclaimer: This newsletter is for informational purposes only and does not constitute investment advice. I am not a financial advisor. Always do your own research before making any investment decision.

Disclosure: I may hold direct or indirect positions (including options) in any securities mentioned in this newsletter. My opinions are my own and always honest.