Nvidia's margins are safe this decade. The company's data center GPU dominance, a potential $250 billion OpenAI data center guarantee, and a $10 billion South Korea expansion with Brookfield and NAVER support a revenue trajectory that makes a demand cliff unlikely before 2030.
Nvidia's margins are safe this decade. The loudest fears on my feed are about a demand cliff that the company's own income statement refuses to confirm.
The bear case for Nvidia Corp. rests on a simple premise: hyperscaler AI spending peaks, GPU supply catches up, and the pricing power that pushed gross margins above 70% evaporates. That scenario assumes the AI infrastructure buildout is finite — a view that looks increasingly at odds with the numbers. Nvidia is now evaluating a $250 billion guarantee to back OpenAI's leasing of a 10-gigawatt data center hub being built by SoftBank Group in Ohio, according to reports from July 26. The total project cost could reach $500 billion, SoftBank founder Masayoshi Son has said. A financial commitment of that magnitude does not suggest a company bracing for a demand cliff.
"Enterprise AI deployment is shifting toward cost-effective open-source models, but that trend will not slow the overall investment cycle," UBS said in a July 26 report. "It may actually bolster demand for Nvidia GPUs and cloud infrastructure as companies focus on optimizing costs." The bank's analysis points to a counterintuitive dynamic: cheaper inference models drive broader adoption, which in turn requires more compute capacity, not less. Companies monitoring token usage and routing workloads to the most affordable models still need the underlying GPU infrastructure to run them.
Nvidia's valuation supports the bull case. The stock trades at $206.84, a 43.8% discount to its GF Value of $368.30, according to GuruFocus data. The trailing 12-month price-to-earnings ratio of 31.68x sits well below the 5-year median of 58.52x, suggesting the market has already priced in significant margin compression that has not materialized. The company's GF Score of 95 out of 100 reflects top marks in profitability and growth, both rated 10 out of 10, with financial strength at 9 out of 10. The valuation rank of 4 out of 10 flags the stock as a possible value trap — but that label applies only if the demand cliff actually arrives.
The $250 Billion Question
The Ohio data center project represents the most ambitious single AI infrastructure commitment to date. A 10-gigawatt facility would consume roughly the output of 10 nuclear reactors, making it one of the largest energy consumers in the country. Nvidia's role as guarantor would enable SoftBank to secure additional financing for the project while ensuring OpenAI has the computing capacity needed for next-generation model training. The arrangement mirrors a broader trend: chip companies are increasingly underwriting infrastructure to guarantee demand for their products.
Nvidia's expansion in South Korea, announced July 27, follows the same logic. The company, along with NAVER and Brookfield Asset Management, plans to scale the NVIDIA DSX AI factory at NAVER's GAK Sejong data center from 55 megawatts to 200 megawatts. Brookfield will invest up to $9 billion as the exclusive capital partner, with Nvidia contributing $1 billion and NAVER providing the remainder of the $10 billion project. The facility will run Nvidia's Vera Rubin and Blackwell platforms, offering production-level AI computing to developers in South Korea and the U.S.
What the Insiders Are Doing
The one data point that gives me pause: insider selling. Over the past three months, Nvidia insiders sold $410.6 million worth of shares with zero purchases. That level of selling could signal caution among executives about near-term stock performance. It could also reflect simple portfolio diversification at a $5 trillion market capitalization where insider holdings represent enormous concentrated wealth. Either way, it is a data point worth monitoring, not a reason to sell.
For investors, the math is straightforward. Nvidia's data center revenue has compounded at more than 100% annually over the past two years. The company's gross margins, which peaked near 78% in early 2025, have settled in the low 70% range — still far above the 50-60% range that prevailed before the AI boom. A return to those pre-boom levels would imply a 20-point margin compression that the current P/E of 31.68x already discounts. If margins hold closer to 65-70% through the decade, as the infrastructure buildout suggests, the stock's current valuation offers a meaningful margin of safety.
This article is for informational purposes only and does not constitute investment advice.
