I keep buying NVIDIA (NASDAQ:NVDA | NVDA Price Prediction). Every time cash hits my brokerage, every paycheck, every time the position looks too large to add to, I add to it anyway. If I lost everything tomorrow and had to rebuild a portfolio from scratch, NVDA would be the first ticker I typed.
The thesis is simple. NVIDIA sells the picks and shovels for the most capital-hungry build-out of my lifetime. CEO Jensen Huang calls it the agentic AI inflection point, and the spend behind it shows up in numbers I have never seen on a company this size.
The income statement does the talking
Fiscal 2026 closed with revenue of $215.94B (+65.47% YoY), net income of $120.07B, and free cash flow of $96.58B. The Q4 print was $68.13B in revenue (+73.2% YoY) and EPS of $1.62 vs a $1.52 estimate.
Growth keeps accelerating as the base gets larger: Q1 +69.18%, Q2 +55.6%, Q3 +62.49%, Q4 +73.21%. Data Center alone did $62.31B in Q4 (+75% YoY), and Data Center Networking grew 263% YoY. Management guided Q1 FY2027 revenue to about $78B, and that figure assumes zero Data Center compute revenue from China.
The moat is paid for in advance
Meta has committed to a multiyear program covering millions of Blackwell and Rubin GPUs. OpenAI is set to deploy at least 10 gigawatts of NVIDIA systems. CoreWeave is building 5+ gigawatts of AI factories by 2030.
AWS, Google Cloud, Microsoft Azure, and Oracle are all standing up Vera Rubin instances. Huang put it plainly: “Computing demand is growing exponentially.”
Blackwell Ultra delivers up to 50x better performance and 35x lower cost for agentic AI versus Hopper, and the Rubin platform targets a 10x reduction in inference token cost on top of that. Customers keep buying because the per-token math wins.
The balance sheet and the multiple
Operating cash flow ran $102.72B for the year. The company returned $41.1B to shareholders in FY2026 ($40.1B in buybacks, $974M in dividends), and $58.5B remains on the buyback authorization.
Non-GAAP gross margin sits at 75.2%. Return on equity is 101.5%. At $216.61, the trailing P/E is 43 and the forward P/E is 26. For a company growing quarterly earnings 95.6% YoY with this margin profile, I see a mathematical mismatch I am happy to keep buying.
The honest risk
China is excluded from the model. The Q1 FY2026 H20 charge cost $4.5B. Beta is 2.34, so this position will move violently in both directions. Reliance on TSMC for fabrication is real, and supply-related commitments of $95.2B are real obligations.
I keep buying because every one of those risks was in front of investors a year ago, when the stock was $110.98. Today it sits at $216.61, up 95.17% over twelve months. The thesis bent and held.
Forward conviction is what keeps the buy button warm. The customers funding this build-out are the most cash-rich companies in the world, and they have told us, in dollars, what they intend to spend through 2030.
Analyst consensus sits at $268.61 with 57 Buy ratings against 2 Hold and 1 Sell, and the crowd can debate the next $20 of price action all it wants. I am buying the next decade of compute, and I will explain it to my grandchildren the same way I am explaining it to you now.
Top Gaining Stocks
Top Losing Stocks