I keep buying NVIDIA (NASDAQ:NVDA | NVDA Price Prediction), and I am not going to pretend otherwise. Every paycheck, every rebalance, every dip that lasts more than a week, I add. The reason is simple. The company is selling the one product the entire technology industry is begging for, and the order book keeps getting longer.
My conviction is about a build cycle. Jensen Huang called it “the AI industrial revolution,” and the spending behind it shows up in my account statements. When Meta commits to millions of Blackwell and Rubin GPUs, when OpenAI signs for at least 10 gigawatts of NVIDIA systems, and when CoreWeave plans 5+ gigawatts of AI factories by 2030, I am buying a backlog.
The numbers that keep my finger on the buy button
Start with Q4 FY2026. Revenue landed at $68.13 billion, up 73.2% year over year, with non-GAAP EPS of $1.62 against a $1.52 consensus. Net income rose 94.47% to $42.96 billion in a single quarter.
Data Center revenue alone hit $62.31 billion, up 75%, and Data Center Networking jumped 263%. For full-year FY2026, revenue reached $215.94 billion, up 65.47%, with EPS of $4.77. Companies this size are not supposed to grow this fast.
Second, the cash. NVIDIA generated $96.58 billion in free cash flow for FY2026 and returned $41.1 billion to shareholders, almost entirely through buybacks, with $58.5 billion still authorized. Shareholders’ equity rose 98.28% year over year to $157.29 billion. The penny dividend is symbolic.
The buyback is the real capital return story, and it is funded by operating margins most software businesses would envy. Trailing P/E sits at 43, with forward P/E at 26. For a business compounding net income near triple digits, I am paying a fair price.
Third, the moat. The hyperscalers (AWS, Google Cloud, Microsoft Azure, Oracle Cloud Infrastructure) were the first to deploy Vera Rubin instances. The U.S. Department of Energy named NVIDIA the private industry partner on the Genesis Mission.
Auto, pharma, and engineering software (Mercedes-Benz, Lilly, Synopsys, Siemens, Dassault Systèmes) all run on the same stack. CUDA is the lock. The roadmap, with Rubin offering up to a 10x reduction in inference token cost versus Blackwell, is the key.
The risk I refuse to wave away
China is the real one. The Q1 FY2027 guide of $78.0 billion explicitly excludes Data Center compute revenue from China. Earlier in the year, NVIDIA absorbed a $4.5 billion H20 charge tied to export restrictions. Supply commitments of $95.2 billion sit on the books, and the company depends on TSMC for production. Any of those can bite.
The thesis still holds because the company guided to a record quarter while writing China to zero. That is the punchline. Demand outside one market is enough to keep growth above 70%, and the customer concentration risk is the opposite of fragility when the customers are Meta, OpenAI, Anthropic on Azure, and every sovereign building an AI factory.
Why the buy button stays active
At $216.61, with a $268.61 analyst consensus target and a Rubin ramp ahead, I am holding the picks-and-shovels supplier of the most expensive industrial buildout of my lifetime. I think this stock is just getting started, and I plan to keep buying it like I mean it.
Top Gaining Stocks
Top Losing Stocks