Nvidia Stock: Buy Or Sell?

Nvidia stock has climbed roughly 40% over the past 12 months, yet it trades at just 25x FY’27 estimated earnings, a modest multiple relative to its growth trajectory. As we approach the FY’26 year-end results, (January fiscal year), investors are faced with a key question: Does this low valuation represent a big buying opportunity, or is the market correctly pricing in a looming cyclical peak?

The bull case rests on the belief that the market is structurally underestimating the scale of the AI infrastructure cycle.

Revenue is expected to scale from $213 billion in FY’26 to $330 billion in FY’27, with a credible path toward $480 to $500 billion by FY’28 if spending remains elevated. Big Tech’s capex surge for this year and recent developments relating to the increasingly capable AI agents point to sustained demand for Nvidia’s computing power. At a 50% net margin, $480 billion in revenue translates into $240 billion in net income. Against a $4.6 trillion market capitalization, that implies a forward multiple of 19x earnings.

If Nvidia compounds its earnings power this quickly, the stock does not need multiple expansion to justify upside; earnings growth alone will drive it higher. If confidence in the durability of this growth increases, the multiple could actually expand, adding even further upside.

Despite the strong math, the debate centers on the durability of this expansion. related: Nvidia Stock’s Cheap 25x Multiple The Loudest Warning Yet?

If hyperscalers don’t see clear software revenue from their massive AI investments, a “CapEx cliff” could follow the current build-out. Major customers (Alphabet’s TPU, Amazon’s Trainium) are increasingly shifting internal workloads to their own chips to reduce reliance on Nvidia.

The Bull Case: The Math of Hyper-Growth

The Path to 50% Annual Growth

To support a 50% annual growth rate, Nvidia would need to grow its revenue from approximately $213 billion in FY2026 (ending Jan 2026) to roughly $330 billion in FY2027 and $480 billion plus in FY2028. So where will those additional $100 to $150 billion in new annual revenue dollars come from over the next few years?

1. Hyperscaler Capex (+$60 to 80B incremental revenue)

The core of the bull case rests on the four largest cloud platforms: Microsoft, Amazon, Alphabet, and Meta. Estimated impact: +$60–80 billion in additional revenue.

Collectively, they are signaling a sharp increase in capital expenditures for calendar year 2026, largely directed toward AI infrastructure. For perspective, Amazon is projected to spend roughly $200 billion in 2026, up from about $132 billion in 2025. Alphabet has guided to $175 to $185 billion. Meta expects $115 to $135 billion. Microsoft is also likely to see a significant step-up in spending. In aggregate, 2026 capex could surpass $600 to $650 billion, with an estimated 75% tied directly to AI infrastructure. If Nvidia maintains its historical share of GPU and AI networking spend, this expansion alone could account for the majority of incremental revenue growth over the next two years.

2.  Sovereign AI (+$20 to $30 B)

Nvidia is decoupling from purely corporate demand. Nations like Japan, France, Canada, and Middle Eastern states are investing in domestic AI clouds to ensure data sovereignty. This segment is on track to double from $20B to $40B+ over the next two years.

3. Enterprise And Agentic AI (+$15 to $20 B)

The narrative for 2026 shifts from experimental chatbots to production-grade AI agents that can execute complex workflows. As enterprises move from pilot programs to scaled deployment, infrastructure requirements rise materially. The AI inference market, which supports the ongoing operation of trained models, is projected by several analysts to reach roughly $255 billion by 2030. As large enterprises deploy agentic AI systems across operations, they will require dedicated on-premise clusters or private cloud capacity, supporting demand for Nvidia’s DGX systems and high-performance networking solutions.

Besides this, there are other avenues such as physical AI and automotive, that could add incremental revenue.

Reasons to Sell Nvidia? 

1. ROI concerns and Hyperscaler “Cash Crunch”

The bear case rests on the argument that hyperscalers literally cannot afford to sustain this spending without degrading their balance sheets. Here is the specific “cash vs. capex” mismatch that supports the sell thesis:

The “cash vs. CapEx” mismatch is stark: Alphabet has guided to $175 to $185 billion in spending, roughly 1.4x its total liquidity of $126.8 billion. Meta plans to spend $115 to $135 billion, about 1.5x its $81.6 billion cash position. Amazon is expected to spend $150–$200 billion against an estimated $88 billion in liquidity, while free cash flow fell 69% in 2025, increasing its dependence on debt funding.

The broader risk is capital intensity. If software and services monetization does not scale quickly enough to rebuild cash balances, these companies may be forced to rein in CapEx. For Nvidia, that scenario would translate into a sharp contraction in orders and a rapid compression of its backlog.

2. Competitive Landscape

Even if demand remains high, Nvidia’s near-monopoly is facing its most significant challenge yet from two distinct fronts:

Nvidia’s largest customers could also become its biggest competitors. To escape the “Nvidia Tax,” hyperscalers are aggressively pivoting to internal chips, which also excel at inference tasks. For example,

• Google’s TPU v6: Now reportedly handles a decent chunk of Gemini’s training and inference.
• Amazon’s Trainium2: AWS is scaling “Project Rainier” to over 1 million internal chips to reduce reliance on H100/B200 clusters.
• Meta’s MTIA: Mark Zuckerberg has signaled that Meta’s custom silicon will play an “increasingly outsized role” in their 2026 infrastructure plans.

AMD has emerged as the essential alternative for those seeking to avoid Nvidia’s pricey chips. Its Instinct MI350 series provides 1.6x the memory capacity of Blackwell, making it a favorite for high-efficiency inference. Supported by the maturing ROCm 7 open software stack and a massive 6-gigawatt supply deal with OpenAI, AMD is successfully commoditizing the hardware layer and capturing the spillover demand Nvidia simply cannot fulfill.