Over the last few years, Nvidia (NVDA -3.01%) has emerged as the leader of America's tech industry -- riding a tidal wave of demand for its cutting-edge chips used to train and run AI algorithms. However, with shares down 17% year to date, the company has hit a roadblock as a combination of trade uncertainty and possible foreign competition calls its growth thesis into question.
Should investors buy the dip or stay far away from this technology leader? Let's dig deeper to find out.
Nvidia's problem is deeper than tariffs
While it's easy to blame Nvidia's recent dip on the Trump administration's trade policy, that's only part of the story. Semiconductors are exempt from U.S. tariffs (although this could change in the future), and Nvidia has already made significant strides to onshore its supply chain -- even sourcing its latest Blackwell-based graphics processing units (GPUs) from Taiwan Semiconductor Manufacturing's new manufacturing facility in Arizona.
NASDAQ: NVDA
Key Data Points
The real problem started in January, with the launch of DeepSeek R1 -- a Chinese large language model (LLM) that reportedly matched the performance of industry leaders like ChatGPT despite using less-advanced H800 chips.
While there has been debate about DeepSeek training costs and possible intellectual property violations, it raises some alarming questions.
Can cheaper alternatives replace Nvidia's most advanced GPUs? Will the high-cost American AI software industry be undermined by more affordable options in China and other parts of the world? It is still too early to have an answer to these questions, but the uncertainty clearly weighs on Nvidia's stock price, even though growth remains strong.
Could clients start diversifying their supply chains?
While DeepSeek's full impact is yet to be seen, there is already evidence that Nvidia's clients are seeking to reduce their reliance on the company by turning to in-house hardware solutions. In February, OpenAI finalized an agreement with TSMC to develop its first generation of custom AI chips with plans to launch them in 2026.
Custom chips are AI chips designed for a specific workload. This characteristic means they can operate with fewer unneeded components, making them more efficient than Nvidia's one-size-fits-all solutions. Perhaps more importantly, they allow clients to skip a costly middleman.
Image source: Getty Images.
Nvidia's products aren't cheap. In the fiscal fourth quarter, the company boasted a gross margin of 73.5%, which is surprisingly high for a hardware company. To put that number into perspective, software-as-a-service (SaaS) giant Microsoft has a gross margin of 69%.
Custom chips can allow customers to buy AI hardware for closer to their production cost. And OpenAI isn't the only major Nvidia client exploring this option. Alphabet, Amazon, and Meta Platforms already make custom chips, while Tesla is working on Dojo D1, which is designed for its AI supercomputer. As a fabless chip designer (that doesn't actually manufacture its chips), Nvidia's business model is a target for competition from its high-tech rivals, who can also make deals with its manufacturing partner, TSMC.
The valuation is at rock bottom
So far, Nvidia's potential challenges from DeepSeek and custom chips have had no noticeable impact on its operational results. Business is booming, with fourth-quarter revenue surging 78% year over year to $39.3 billion, driven by the successful rollout of the company's new Blackwell AI data center chips. Net income soared 80% to $22.1 billion.
Against these stellar results, Nvidia's forward price-to-earnings (P/E) multiple of just 25 looks perplexingly cheap. That said, investors shouldn't rush to buy the stock until there is more clarity on these long-term challenges.
