As a consequence of the changing U.S. economic policies, NVIDIA is suffering some headwinds in 2025 as its stock value and profits take a considerable hit. The ExpertStack team has tried to make sense of the chip market – and the role China plays – and who’s on the receiving end of these turbulent winds.
NVIDIA may have surpassed Wall Street targets with Q4 revenue of $39.3 billion coupled with an EPS of 89 cents, but the firm’s growth outpaced sentiment was bearish as it clocked in at its slowest acceleration since 2023. The profitability decline also eroded further dragged the firm’s sentiment as it launched its next-generation Blackwell chip.
Import Taxes: A Death Hook for Supply Chains
NVIDIA may design the world’s most powerful chips, but it doesn’t manufacture them. Instead, it relies heavily on Taiwan’s TSMC, which produces about 90% of its chips – including high-demand models like the H100.
But in January 2025, the U.S. government introduced a sweeping 25% import tax on all goods coming from Taiwan. For NVIDIA, this means every chip sourced from TSMC now carries a steep markup. When you’re producing millions of chips annually, that 25% translates to billions in added costs.
To offset the blow, NVIDIA has two options: hike prices or accept slimmer profit margins. Neither is ideal – especially for data centers and AI firms that buy chips by the thousands. Even a minor price increase multiplies into billions in additional spending for these large-scale users.
So, what can NVIDIA do? One workaround is relocating production to the U.S., and TSMC is already building new facilities in Arizona. The catch: those plants won’t be operational until 2027, and they come with a $100 billion price tag. Worse, domestic production isn’t cheap – U.S. labor costs are about 50% higher, and energy costs around 20% more. This could drive chip prices up by another 20%, pushing the cost of a single high-end unit like the H100 from $40,000 to well over $48,000.
In short, the 25% tariff is not just a policy move – it’s a direct hit to NVIDIA’s cost structure, threatening both affordability for customers and profitability for the company.
NVIDIA is trying to negotiate with the US government to reduce or eliminate taxes on chips. Another option is to find other suppliers, for example, in South Korea or Japan. But this is not fast: you need to check that the new factories make chips just as well, and arrange delivery, and this adds another 5-10% to the costs.
If you look at it more broadly, because of the taxes, the prices of chips are growing, and this concerns not only NVIDIA, but also those who buy its products – companies that produce servers, video cards or systems for smart machines. This makes American technologies more expensive than European or Asian ones, where there are no such taxes. The paradox is that these taxes can help NVIDIA’s competitors from China, because China will start developing its chips faster in order not to depend on America.
A 25% tax could cut NVIDIA’s profits by 5% to 10% if it doesn’t raise prices. Switching to suppliers in South Korea or Japan requires quality assurance and new shipping routes, adding 5% to 10% to costs due to testing and logistics.
Bans on sales to China: Losing a major market
China was once a goldmine for NVIDIA, generating billions in annual revenue as Chinese tech giants raced to build massive AI data centers. But in 2025, that revenue stream was abruptly disrupted. The U.S. government tightened export restrictions, citing national security concerns and the risk of advanced chips being repurposed for military use. As a result, NVIDIA was banned from selling many of its high-performance chips to China—cutting off roughly half its shipments to the region.
In response, NVIDIA tried to adapt by creating lower-performance alternatives that complied with U.S. export rules. One such product is the H20 chip, a downscaled version of the powerful H100. While the H20 is still capable of supporting AI workloads, it delivers only about 70% of the H100’s performance in critical tasks like training large language models. This intentional performance cap ensures it stays below the U.S. government’s regulatory threshold.
However, the reception in China has been lukewarm. These weakened chips offer less performance at a similar price point, making them less attractive for cost-sensitive companies looking to scale AI infrastructure. With domestic demand rising, Chinese firms are increasingly exploring alternatives and building their own.
Huawei, for instance, has developed the Ascend chip series, which already rivals NVIDIA’s offerings in certain machine learning tasks. In some benchmarks, Huawei’s Ascend chips reach up to 85% of the H100’s performance. Bolstered by $50 billion in state-led investments over the past two years, China has made chip self-sufficiency a national priority. By 2030, it aims to domestically produce 80% of the chips it needs, significantly reducing reliance on U.S. technology.
This shift poses a long-term threat to NVIDIA. Even if the company continues to grow in other regions, the loss of the Chinese market could cost it billions in future revenue. Worse still, if U.S. restrictions are tightened further – potentially banning even lower-performance chips like the H20 – NVIDIA may be completely shut out of one of the world’s largest and fastest-growing AI markets.
For China, these bans are both a challenge and an accelerant. Every new restriction gives added urgency and incentive – to develop homegrown alternatives. For NVIDIA, the clock is ticking to find new markets or products to fill the gap.
Growing Competition: NVIDIA Losing Its Lead
NVIDIA has long been the leader in AI chips, but in 2025, competition has become a serious threat. AMD and Intel from the US, as well as Huawei from China, have started to take market share away from NVIDIA. This is important to understand because if NVIDIA does not maintain its lead, it could lose control of the AI industry, where it currently holds 88% of the chip market.
AMD has taken a step forward
AMD has taken a decisive step forward in the AI and GPU race. Its latest Instinct MI200 chips are already being deployed for AI workloads and offer a compelling value proposition: comparable performance to NVIDIA’s H100 at roughly 20% lower cost. Over the past year, AMD has doubled its market share in the AI accelerator space, rising from 5% to 10% – a clear signal that customers are responding to the price-performance equation.
While NVIDIA still leads in raw performance, much of its edge comes from its mature software ecosystem, particularly the CUDA platform, which boosts chip efficiency in AI applications. AMD is closing that gap with its ROCm (Radeon Open Compute) platform, which saw a 30% development acceleration in 2024. Though ROCm still lags behind CUDA in terms of optimization and developer adoption, AMD is gaining ground steadily.
In terms of raw numbers, the MI200 delivers approximately 80% of the H100’s performance in machine learning benchmarks but at a significantly lower price – around $32,000 compared to $40,000. This makes AMD an increasingly attractive option for organizations looking to build or scale AI infrastructure without overshooting budget constraints.
The competition is also heating up in the gaming market. AMD’s Radeon RX 9070 XT is now neck-and-neck with NVIDIA’s RTX 5070, offering similar performance at a lower price point. If AMD continues this trajectory – improving performance, expanding software capabilities, and keeping prices competitive – NVIDIA may be forced to adjust pricing or risk losing share in both AI and consumer markets.
The message is clear: AMD is no longer just catching up – it’s becoming a credible threat in segments where NVIDIA once had near-total dominance.
Intel is not far behind
Intel is making a calculated move to re-enter the high-performance chip race with its upcoming 18A process node, a next-generation manufacturing technology designed to challenge NVIDIA’s dominance in AI and data-centric computing. Although still in the testing phase, early benchmarks are promising: Intel’s prototypes have outperformed NVIDIA’s H100 in select workloads, particularly in complex scientific applications like climate modeling and large-scale simulations.
The 18A process aims to increase transistor density by approximately 20% over current-generation chips, a leap that could translate into significant gains in performance and power efficiency. In preliminary testing, chips built on this architecture demonstrated speeds up to 15% faster than NVIDIA’s flagship H100 in specific high-performance computing (HPC) scenarios.
Mass production is slated for 2026, and while Intel has yet to fully commercialize this technology, its potential is clear. If the company can meet its manufacturing timeline and deliver consistent performance gains, it could start peeling away enterprise customers—particularly in sectors like big data, scientific research, and government infrastructure, where raw computational power and long-term scalability are critical.
And now China
China is no longer just a consumer in the global chip market—it’s rapidly becoming a competitor. Huawei’s Ascend chip series has already found traction in domestic data centers and is beginning to rival NVIDIA in AI workloads. In response to U.S. export bans on high-performance chips, China has poured over $50 billion into its semiconductor sector in just two years. The strategic goal is clear: produce 80% of the country’s chip demand domestically by 2030.
The results are already visible. Ascend chips currently deliver around 85% of the performance of NVIDIA’s H100 in AI tasks such as natural language processing and large-scale model training. And in the Chinese market, they’re 25% cheaper. Since 2023, China has ramped up local chip production by 40%, driven by aggressive state-backed investment.
So far, these chips are only used domestically, but if Huawei starts exporting its hardware – especially to developing markets or U.S.-restricted regions – NVIDIA could face a double blow: not only losing the Chinese market, but also its influence in neighboring economies.
Scenarios for the future
With increasing tax burdens, export bans, and rising competition from AMD, Intel, and China, NVIDIA is facing its most challenging period in recent years. Company leadership insists it’s pivoting strategically, expanding into new markets like India and Europe, where sales jumped 15% in 2024, reaching $5 billion. Still, the road ahead is far from certain.
Here are three possible scenarios for NVIDIA’s future:
Optimistic:
NVIDIA successfully launches its next-generation chips in 2026, delivering a 40% performance boost (up to 200 gigaflops). Global demand surges, offsetting losses in China. U.S. trade restrictions ease, reducing the cost pressure from tariffs. Sales in Europe, India, and Latin America accelerate, allowing NVIDIA to strengthen its global dominance.
Pessimistic:
China meets 70% of its chip demand by 2028, increasingly relying on domestic suppliers like Huawei. NVIDIA is squeezed out of key Asian markets. Meanwhile, AMD and Intel, offering chips at 25% lower prices, erode 25% of NVIDIA’s U.S. market share. Revenues fall by 20%, and the company struggles to maintain growth.
Realistic:
NVIDIA retains 75% global market share thanks to its robust ecosystem, R&D leadership, and brand strength. However, rising production costs and sustained geopolitical headwinds reduce profitability. Margins drop to 65%, and while competitors continue to chip away, NVIDIA stays ahead through continuous innovation at least through 2030.
The Bottom Line
NVIDIA’s future hinges on innovation, international diplomacy, and its ability to stay ahead of fast-moving competitors. It still leads in AI performance, developer tools, and global recognition – but pressure is mounting. If it can’t adapt quickly enough, its dominance may no longer be guaranteed. What’s keeping it afloat today is a combination of cutting-edge technology, years of experience, and a brand reputation that competitors are racing to match.