Wafer Input Costs at TSMC’s U.S. Plant Soar, NVIDIA GPU Pric

Wafer Input Costs at TSMC’s U.S. Plant Soar, NVIDIA GPU Prices “Rise Across the Board” -Digitimes

 

Sources in the supply chain have revealed that NVIDIA is facing multiple crises. Following a reported $5.5 billion write-down in its quarterly earnings due to the ban on downgraded H20 chips being sold to China, CEO Jensen Huang has been shuttling between the U.S. and China in hopes of minimizing the impact and persuading the U.S. government to ease AI chip restrictions.

 

At the same time, market rumors indicate that in order to maintain stable profitability, NVIDIA has recently raised official prices for almost all its products, allowing its partners to increase prices accordingly.

 

For example, ASUS’s high-end RTX 5090 graphics card, launched earlier this year at a retail price of around NT$90,000, has recently surged to NT$100,000 through distribution channels—an increase of more than 10%. Prices for H200/B200 GPU modules and servers have also gone up.

 

Given that AI chip demand outside China remains robust and cloud service providers (CSPs) continue expanding capital expenditures, NVIDIA is expected to deliver strong quarterly earnings within guidance by the end of this month.

 

Since April, U.S. President Donald Trump has imposed reciprocal tariffs and tightened AI chip export rules, effectively blocking NVIDIA’s downgraded H20 chips from entering China. As a result, NVIDIA has become a major target of these trade restrictions.

 

Jensen Huang is doing everything he can to defuse the situation and minimize the impact.

 

In response to Trump’s “Made in America” initiative, NVIDIA announced a $500 billion investment over four years to establish AI server manufacturing sites in the U.S. in partnership with TSMC, Foxconn, and others. Production of 4nm Blackwell chips has already begun at TSMC’s Arizona facility. Huang has also appeared alongside Trump, expressing that U.S. policies should focus on enhancing the competitiveness of American companies. He warned that restricting chip exports to China and other nations would undermine America’s technological leadership. He also visited Chinese government agencies during a sensitive time.

 

As Huang works both sides of the U.S.-China divide, he is seeking a resolution while repeatedly stressing to the U.S. government that China’s AI capabilities are catching up fast. He has warned that broader restrictions on AI chips could seriously harm U.S. chipmakers and accelerate China’s development efforts.

 

In a recent joint appearance with Trump, Huang even mentioned Huawei—blacklisted by the U.S.—noting that it has entered the AI chip field, achieving excellent performance in computing and networking technologies. He reiterated that U.S. policy should prioritize strengthening domestic competitiveness, and that limiting chip exports to China and others would threaten America's current technological lead.

 

It’s not just Huang raising concerns. AMD CEO Lisa Su also recently stated that China represents a "huge opportunity" for the global semiconductor and AI industries, urging a balanced approach between export controls and AI advancement. Rumors have emerged that the Trump administration is considering loosening some of the existing AI chip restrictions.

 

According to the supply chain, NVIDIA’s Jensen Huang is actively participating in public events and issuing warnings to the Trump administration, while reiterating support for the “Made in America” initiative.

 

On the operational side, NVIDIA has historically maintained strong control. Despite uncertainty around tariffs and AI chip restrictions, the company is striving to post stable earnings. Numerous products have reportedly seen official price hikes, and partners have followed suit with their own increases.

 

Take the high-demand RTX 50 series as an example: the RTX 5090, upon global launch earlier this year, was immediately snapped up at premium prices. Chinese AI chip buyers, in particular, paid without question—willing to purchase at any cost. The RTX 5090, originally priced around NT$80,000, quickly doubled in channel pricing.

 

After the AI chip ban was enacted, the RTX 5090 became even harder to obtain, pushing prices higher. Due to the sharp decline in Chinese orders, companies have introduced bundled offerings to maintain profitability and clear inventory. ASUS and others have pushed packages that include power supplies or fully assembled systems. The price of the RTX 5090 surged from around NT$90,000 to NT$100,000 overnight. Other RTX 50 series cards also saw price hikes of 5–10%.

 

Additionally, NVIDIA has reportedly raised prices for H200 and B200 chips and modules in response to rising costs and platform transition strategies. The price increases have rippled out, with server vendors also adjusting their quotes to clients—by approximately 10–15%.

 

The supply chain notes that shifting Blackwell chip production to TSMC’s U.S. plant has led to noticeable cost increases in manufacturing, materials, and logistics. Much like TSMC itself, NVIDIA has the pricing power to pass these costs on to customers, making the price hikes expected.

 

Even though AI chips like the H20 are banned from being sold to China, Chinese buyers still manage to acquire them through other channels outside the mainland. Meanwhile, demand from U.S. CSPs and other international customers remains strong. NVIDIA is expected to post impressive earnings within guidance in its upcoming report.

 

NVIDIA’s fiscal Q4 earnings for the period ending January 26, 2025, beat market expectations.

 

Revenue reached $39.33 billion, up 12% quarter-over-quarter and 78% year-over-year. Data center revenue came in at $35.6 billion, rising 16% QoQ and 93% YoY. The overall gross margin was 73%.

 

Looking ahead to Q1 of fiscal 2026, NVIDIA estimates revenue of approximately $43 billion, a 65% increase YoY, although this growth is slower than the 262% surge in the same period a year earlier. Gross margins are expected to reach 70%.

 

NVIDIA plans to release its earnings at the end of May. Before that, Jensen Huang will visit Taiwan on May 12 to meet with supply chain partners, dine with major manufacturers on the 17th, and deliver a keynote speech on the 19th—reviving what many call the “trillion-dollar banquet,” with full momentum behind the AI boom.