The chipmaker raises capital spending outlook after cloud providers continue signaling robust demand for AI infrastructure.
Taiwan Semiconductor Manufacturing Co. plans to invest another $100 billion to expand its Arizona manufacturing campus, bringing its planned US investment to $265 billion, as surging AI demand drove record quarterly results and a higher capital spending outlook.
The expansion will add several leading-edge semiconductor fabrication plants and advanced packaging facilities, further positioning the Arizona campus as one of the world’s largest AI chip manufacturing hubs, Chairman and CEO C.C. Wei said on the company’s July 16 earnings call. “Our customers and customers’ customers, who are mainly the cloud service providers, continue to provide us with their very strong signal and positive outlook,” he said. “Thus, our conviction in the multi-year AI megatrend remains very high.”
For data center operators, the combination of higher capital spending, expanded investment in Arizona, and aggressive buildout of leading-edge manufacturing and packaging capacity suggests the semiconductor supply chain is still scaling for another wave of AI infrastructure. Jack Gold, president and principal analyst at J.Gold Associates, told Data Center Knowledge TSMC's latest U.S. investment reflects both sustained customer demand and a broader effort to diversify semiconductor manufacturing. The company also is betting it can keep its new U.S. fabs fully utilized despite higher production costs than Taiwan, he said.
TSMC raised its 2026 capital spending budget from $60 billion to $64 billion, up from $52 billion to $56 billion, and now expects full-year revenue to grow slightly more than 40% in U.S. dollar terms.
Matt Kimball, vice president and principal analyst for data center technologies at Moor Insights & Strategy, said TSMC's results reflect both unrelenting AI demand and a soft smartphone market. “Gross margins are at 67.7%, up more than nine points year over year,” Kimball told Data Center Knowledge. “This is TSMC pricing the leading edge around AI infrastructure.” He said the company is allocating its fixed manufacturing capacity to the highest-margin AI chips, reinforcing how demand has reshaped its business.
AI and HPC Now Lead TSMC’s Revenue Mix #
High-performance computing, which includes AI accelerators, CPUs, and networking silicon, generated 66% of second-quarter revenue, up from 61% in the first quarter and 60% a year earlier, as TSMC posted record revenue of NT$1.27 trillion ($40.2 billion), up 36% from a year earlier. Net income climbed 77.4% to NT$706.6 billion, and gross margin reached 67.7%. Smartphone revenue declined to 22%, underscoring AI’s growing importance to TSMC’s business.
Wei said the next phase of AI infrastructure is expanding demand beyond GPUs. “The emergence of agentic AI is leading to a resurgence in the role of CPUs in AI data centers, which drives more silicon demand in addition to AI accelerators,” he said.
Gold said agentic AI workloads require more than GPUs. CPUs play a larger role because they coordinate software, data and enterprise systems surrounding AI models, while integrated AI accelerators handle inference. That combination is driving strong demand for both processor types, he said.
Wei added that the trend benefits TSMC regardless of processor architecture because the company manufactures x86, Arm, and RISC-V designs.
2nm Ramps; Advanced Packaging Capacity Expands #
Demand remained concentrated in TSMC’s newest manufacturing technologies. The company said its 2-nanometer process generated 3% of wafer revenue in its first commercial quarter, while 3-nanometer and 5-nanometer technologies accounted for 30% and 33%, respectively. Technologies at 7 nanometers and below represented 77% of wafer revenue.
Wei said third-quarter demand will again be driven by leading-edge manufacturing, including the steep ramp of TSMC’s 2-nanometer process.
Chief Financial Officer Wendell Huang said 70% to 80% of this year’s capital budget will fund advanced process technologies, with another 10% to 20% for advanced packaging, testing, and related manufacturing. “We do not foresee any bottlenecks to our capacity expansion plans,” Huang said.
Kimball said the 2-nanometer ramp also differs from previous technology transitions. Historically, smartphones were the first major adopters of new manufacturing nodes before server processors followed about a year later. “For N2, AI is pulling harder and much earlier,” he said. “A lot of this is hyperscalers designing custom chips on this node.”
Kimball cautioned that much of TSMC's growth still depends on a relatively small group of hyperscale customers. While he sees no signs of demand softening, any pullback in hyperscaler capital spending would have an outsized effect on the company's growth trajectory.