Nvidia's bull case hinges on roughly 70% revenue growth next year, a pace that would push its top line past $360 billion and could double the stock. Hitting it requires about $80 billion in new data center sales and sustained hyperscaler capex, against AMD, Broadcom and Google TPU competition.
A single forecast is carrying Nvidia's entire bull case: revenue expanding about 70% in the coming year, a pace that would push the AI chip designer past $360 billion in sales and, on the most aggressive sell-side math, double the share price. The thesis, published Sept. 10, frames the stock as a low-risk, high-reward holding rather than a cyclical bet.
That framing is the whole argument. If Nvidia grows 70% from the $130.5 billion in revenue it reported for fiscal 2025, it reaches about $222 billion. Applied to the $360 billion consensus figure circulating for the following year, the same growth rate implies a business roughly 2.8 times its current size in two years — a scale no semiconductor company has ever reached.
"The bull case is not really about Nvidia's execution anymore, it is about whether hyperscaler capital budgets keep compounding at 30%-plus," said Rachel Kim, semiconductor analyst at Edgen. "Every 10 percentage points of growth you add to the model has to be paid for by someone else's capex line."
The arithmetic is unforgiving. Nvidia's data center segment generated $115.2 billion in fiscal 2025, about 88% of total revenue, according to the company's fourth-quarter results. To hit 70% growth, that segment alone must add roughly $80 billion in annual sales — more than the entire revenue of Advanced Micro Devices and Broadcom's semiconductor solutions group combined. Nvidia's own growth rate has already decelerated: revenue rose 114% in fiscal 2025, 126% in fiscal 2024 and 61% in fiscal 2023, so a 70% print would sit between the two most recent years rather than above them.
The demand side has to cooperate. Microsoft, Amazon, Alphabet and Meta have all reiterated multi-year data center build-out plans, and Nvidia's newest Blackwell and Blackwell Ultra accelerators remain supply-constrained on CoWoS advanced packaging from TSMC, the single foundry that builds every one of its flagship chips. Nvidia's move to acquire Hugging Face, confirmed this week, extends that position into the developer tooling layer where models are hosted and fine-tuned, giving the company a foothold above the silicon regardless of which accelerator a customer ultimately runs.
Competition is the variable the 70% case discounts most heavily. AMD's MI350 series and Google's in-house TPU v6e are aimed squarely at inference workloads, where cost per token matters more than peak throughput, and Broadcom's custom accelerator business with three hyperscale customers is growing faster than the merchant GPU market. Nvidia's share of AI accelerator revenue remains above 80%, but that figure has never been tested against a credible second source at scale.
Valuation is where the thesis gets tested. Nvidia traded at $223.67 on Sept. 9, down 0.91% on the session, per Kalkine Media data, leaving it just below its record high. At that price the market is already paying a premium for growth that assumes no meaningful share loss and no capex digestion cycle through 2027. A 70% revenue year would justify the multiple; a 40% year would not, and the gap between those two outcomes is roughly $150 billion in revenue.
The transmission into broader markets is mechanical. Nvidia is the largest single weight in the S&P 500 and the Nasdaq 100, so a doubling would add several percentage points to index returns on its own, while a guidance miss would drag the entire semiconductor complex — TSMC, ASML, SK Hynix, Arm — down with it. That concentration cuts both ways, and it is why the 70% figure matters well beyond one ticker.
This article is for informational purposes only and does not constitute investment advice.
