Nvidia Faces A New Challenge From The Very Companies Driving The AI Boom, JPMorgan Says

According to Cembalest, some competitors are closing the performance gap with Nvidia, particularly in inference workloads, which involve running trained AI models to generate responses.

Advertisement
Read Time: 4 mins
JPMorgan On Nvidia
Source: NDTV Profit
Quick Read
Summary is AI-generated, newsroom-reviewed
  • Nvidia’s share of AI computing deals dropped to 39% from over 80% in 2023-2025
  • AMD and Google are closing performance gaps with Nvidia in AI inference workloads
  • Nvidia’s exposure to China’s AI model ecosystem raises regulatory and revenue risks
Did our AI summary help?
Let us know.

Chipmaker Nvidia is facing a growing challenge from the artificial intelligence (AI) companies that have driven demand for its chips, as major AI developers expand their computing supply chains and explore alternatives to the chipmaker's hardware, according to JPMorgan.

In a report titled 'The Icarus Files: What's Next For Nvidia And AI', Michael Cembalest of JPMorgan said Nvidia's share of future computing agreements with leading AI developers has fallen to 39%, from more than 80% during 2023-2025. The shift, he believes, comes as frontier AI labs increasingly look towards their own chips and competing suppliers.

Advertisement

JPMorgan also flagged Nvidia's reliance on Chinese open-weight AI models, potential changes to its financing of AI companies and the implications of its leadership transition as key issues for investors.

AMD, Google Gain Ground In AI Chip Competition

According to Cembalest, some competitors are closing the performance gap with Nvidia, particularly in inference workloads, which involve running trained AI models to generate responses.

Advertisement

READ | Who Is Backing Jensen Huang? Full List Of Tech Titans Supporting His Nvidia Comeback

The report noted that AMD ranked ahead of Nvidia in an inference performance comparison that considered memory and communication between chips, in addition to processing capacity.

It also cited feedback from an AI developer that the time required to onboard models on AMD GPUs had fallen from six to 12 months a year ago to just three days. The improvement was attributed to the use of newer AI models to handle the engineering work involved.

Advertisement

JPMorgan said these developments are likely to weaken Nvidia's advantage from its CUDA software ecosystem and the cost of switching to alternative chips.

Google's Tensor Processing Units (TPUs) also compared favourably with Nvidia's GB300 chips when running China's Qwen 3.5 397B open-weight model, across most tested speeds, the report said.

China Exposure Adds To Nvidia's Challenges

The report also highlighted Nvidia's exposure to China's open-weight AI model ecosystem, which accounts for a large share of token consumption.

Cembalest pointed to the growing use of US-made chips by Chinese companies through overseas data centres, particularly in Southeast Asia. The proposed Remote Access Security Act (RASA), which has passed the US House of Representatives, seeks to restrict Chinese companies from remotely accessing restricted US chips through overseas facilities.

Advertisement

READ | Jensen Huang To Return As Nvidia CEO? What We Know As Silicon Valley Allies Push For His Comeback

JPMorgan said implementing such restrictions could require US authorities to make Asian cloud providers verify customers and obtain licences. The report also noted that Nvidia's revenue billed to customers with Singapore addresses increased from 8% to 22% in the years after US export controls took effect, before the company stopped disclosing the figures in July 2025.

Nvidia's Financing Of AI Companies Under Watch

Another potential risk, Cembalest said, is Nvidia's role in financing the companies that buy its chips.

According to the report, Nvidia spent 58% of its operating cash flow on equity investments and mergers and acquisitions over the last three quarters. Its off-balance-sheet long-term debt and lease liabilities increased from $214 billion to $422 billion between Q1 and Q2 2026. Including land, power and shell guarantees, the figure rose to $530 billion.

Cembalest said a change in Nvidia's approach to investments and vendor financing could affect AI infrastructure companies such as CoreWeave, Nebius, Applied Digital and IREN. Chipmakers AMD and Broadcom could also benefit if Nvidia's ability to secure memory and logic wafer supplies changes.

Can Nvidia Maintain Its Market Position?

Despite these concerns, JPMorgan noted that Nvidia remained highly profitable, with gross margins of 75% and operating margins of 66% during 2023-2025. The company accounted for 83% of AI chip revenue over the period.

The brokerage also highlighted that Nvidia's price-to-earnings-to-growth (PEG) ratio stood at 0.2 times and that its shares were 32% below their all-time high at the time of the report.

Cembalest said Nvidia's cash flow remained strong and that the long design and production cycles in the semiconductor industry could support its financial performance for some time. However, he flagged intensifying competition and changes in the company's leadership as risks investors need to watch out for.

READ | Nvidia Plot Twist: Jensen Huang Eyes Comeback Days After Resigning, Claims Report

Essential Business Intelligence, Sharp Market Insights, Practical Personal Finance Advice, Daily Fuel, Gold and Silver Prices and Latest Stories — On NDTV Profit.


Loading...