Nvidia AI Chips Under Pressure From TPUs and AMD

Nvidia’s AI chips face growing competition from Google’s tensor processing units (TPUs), AMD, and Chinese rivals. Elevated valuation, inventory normalization, and tariff volatility are narrowing the margin for error in analyst revenue forecasts as of Dec. 2, 2025.

Market Share and Competitive Pressures

Nvidia held a 92.0% discrete GPU market share in Q3 2025, down 1.2 percentage points from Q2. AMD gained 0.8 points to 7.0%, while Intel rose to 1.0%, marking its first time above 1%. These figures indicate early fragmentation at the top of the discrete GPU market, though Nvidia remains dominant.

Morgan Stanley estimates Google could ship between 500,000 and 1 million TPUs annually by 2027. Meta is negotiating to rent Google Cloud TPUs in 2026 and plans to deploy them in its own data centers in 2027. If fully realized, this shift could represent up to 10.0% of Nvidia’s annual sales, intensifying competition for hyperscaler demand.

AMD is gaining traction in data-center workloads. Its sequential data-center revenue rose 22.0% in Q3 2025, and management targets a 60.0% compound annual growth rate (CAGR) over the next five years. AMD cites its partnership with OpenAI and the ROCm 7 software stack as competitive advantages. The company reported total Q3 revenue of $9.3 billion, up 35.6% year over year.

Chinese suppliers are advancing domestic AI accelerators. Baidu’s Kunlunxin roadmap includes an M100 inference chip launching in early 2026 and an M300 training and inference chip in early 2027. In November 2025, Baidu unveiled Tianchi 256 and 512 supernode systems aimed at local cloud and enterprise demand.

Scale, Valuation, and Demand Risks

Nvidia posted fiscal 2025 revenue of $130.5 billion, a 114.0% increase year over year. Data-center sales accounted for $115.2 billion, about 88.0% of total revenue. Gross margins ranged from 70.0% to 75.0%, reflecting the company’s dominant position in AI infrastructure spending.

Over the past decade, Nvidia’s revenue compounded at about 39.0% annually, while adjusted net income grew at roughly 57.0% CAGR. The stock returned approximately 22,420% from 2015 to 2025, supporting its elevated market multiples.

Analysts forecast continued rapid growth, with revenue estimates near $205 billion for fiscal 2026 and $272 billion for fiscal 2027. Consensus projects a revenue CAGR of about 45.0% and an adjusted earnings-per-share (EPS) CAGR of about 29.0% from FY2025 to FY2028. The shares trade at a forward price-to-earnings ratio near 23 times.

AI accelerator board (AIB) shipments in Q3 2025 totaled about 12 million units, valued at $8.8 billion. This represented a 2.8% quarter-over-quarter increase, below the 10-year seasonal average growth rate of 11.4%. Observers attribute this to inventory normalization after Q2 2025 panic buying and tariff-driven demand volatility.

The data-center GPU market is projected to grow at a 13.7% CAGR from 2025 to 2030, rising from roughly $120 billion to about $228 billion. The broader global AI market is expected to expand at a 31.5% CAGR through 2033. Major cloud operators account for approximately $200 billion of a $290 billion recent infrastructure spending tally. Hyperscaler capital expenditures are expected to rise more than 40.0% year over year in 2025, contributing to estimates of $3.0–$4.0 trillion in global data-center investment by 2030.

This expansion creates a paradox for Nvidia: the total addressable market is growing fast enough that the company could lose share yet still increase revenue. However, the business becomes more sensitive to share erosion, inventory swings, and execution risks as competition fragments the top end of demand.