Old Huang "is one step ahead"—while everyone else is scrambling for TSMC's production capacity, Nvidia is "seizing land."

While tech giants are fiercely competing for TSMC's wafer production capacity, NVIDIA CEO Jensen Huang has already set his sights on a longer-term goal—securing land for TSMC's future factories. This strategy highlights NVIDIA's attempt to fundamentally safeguard its supply chain in the midst of a surge in AI chip demand.

On January 15, Tianfeng International Securities analyst Ming-Chi Kuo tweeted, "Most customers negotiate for guaranteed capacity, while Jensen Huang negotiates for land." Kuo pointed out that during his visit to Tainan last November, Huang expressed his willingness to pay for reserved P10 and P11 plots adjacent to the Fab 18 complex, with the ultimate goal of also acquiring the P12 plot. At the time, Fab 18's planning only clearly defined P1 to P9.

This move comes as TSMC announced a record-high capital expenditure for 2026, reaching $52 to $56 billion, far surpassing the market's previous estimate of $45 to $52 billion. TSMC made it clear in its earnings call that NVIDIA's strong demand is a key driver behind this increased spending.

At the same time, traditional major customers like Apple are facing pressure to secure enough production capacity, and NVIDIA may have already surpassed Apple in certain quarters of 2025 to become TSMC's largest customer.

Strategic Planning Behind Frequent Taiwan Visits

Jensen Huang visited Taiwan three times in the second half of 2025, a frequency that has attracted significant market attention.

According to Digitimes, after discussing next-generation Rubin GPU architecture production plans with TSMC on August 22, Huang arrived at Tainan Airport on a private jet on November 7, visited TSMC’s 3nm wafer plant in the Southern Science Park, dined with TSMC chairman Mark Liu and other top executives, and attended TSMC’s annual sports meet the following day.

On November 27, during the US Thanksgiving holiday, Huang was again spotted in Taipei, visiting Siping Street and Dazhi District. While he stated publicly that he was "visiting friends," sources revealed the main purpose was to meet with Quanta Computer. The market is concerned about Quanta's assembly yield and whether it can obtain more orders under NVIDIA's tightened procurement controls; given Huang's close relationship with Quanta Chairman Barry Lam, this visit was seen as a discussion of ongoing cooperation strategies.

This marks Jensen Huang’s fifth visit to Taiwan since early 2025. NVIDIA’s plans for new offices on T17 and T18 plots in Taipei’s Beitou Shilin Science Park have been finalized, demonstrating the company’s long-term commitment to the Taiwanese supply chain.

Capacity Battle Reshaping Customer Landscape

According to analysis by Culpium and supply chain sources, NVIDIA may surpass Apple in at least one or two quarters of 2025 to become TSMC’s largest customer. Although TSMC CFO Wendell Huang declined to comment on changes in customer rankings, public data already tells the story.

TSMC's 2025 revenue is expected to grow 36% to $122 billion, while NVIDIA's sales for the fiscal year ending January 2026 are expected to grow 62%; Apple’s product revenue for the 12 months ending December 2025 is projected to grow just 3.6%. TSMC's high-performance computing business grew 48% last year and 58% the year before, while its smartphone business grew only 11%, down from 23% the previous year.

According to Culpium, last August when Mark Liu visited Cupertino, he informed Apple executives of the largest price hike in years. Even worse, Apple, once dominant on TSMC’s client list, now must fight for production capacity. As the AI boom continues, each GPU from clients like NVIDIA and AMD occupies more wafer area, and Apple’s chip designs can no longer guarantee priority position among TSMC’s nearly twenty wafer plants.

Technology Roadmap Determines Short-Term Competitive Landscape

TSMC is currently mass-producing chips at the 2nm node (N2), with Apple as the main buyer. Later this year, TSMC will begin simultaneous mass production of the N2P variant and a new node called A16. Mark Liu stated that A16, equipped with super power rail technology, "is best suited for high-performance computing with complex signal routing." This technology is TSMC's version of backside power delivery, designed to separate signal and power within the chip.

This technology roadmap favors companies like NVIDIA and AMD over the next one to two years. But according to SemiAnalysis, the A14 node, expected to be mass-produced in 2028, "is designed from the start for both mobile and high-performance computing," meaning the balance will shift back towards Apple at that time.

TSMC's revenue is expected to grow nearly 30% in 2026, with capital expenditure rising about 32%. Looking long term, average annual growth for the five-year period through 2029 is projected at 25%, but AI business will average 55% or more—higher than prior forecasts of just over 40%. TSMC's gross margin for the December quarter last year reached a staggering 62.3%, up 280 basis points from the previous quarter.

Despite pressure to expand capacity, Mark Liu stated at an investor meeting that he is staying cautious: "I'm very nervous too. If we’re not careful, this could spell disaster for TSMC." TSMC’s capital intensity exceeds 33%, with depreciation accounting for 45% of revenue costs, far higher than customers like Alphabet and NVIDIA. This means that if demand decreases, TSMC bears the main risk, which explains why the company must balance speed and risk when expanding production.

Risk Disclosure and DisclaimerThe market carries risk, and investments require caution. This article does not constitute personal investment advice, nor does it take into account individual users' specific investment objectives, financial situations, or needs. Users should consider whether any opinions, views, or conclusions in this article fit their particular circumstances. Investing based on this is at their own risk.