TSMC Posts Record July Revenue on AI Orders

TAIPEI–The monthly revenue report from Taiwan Semiconductor Manufacturing Co. has become a ritual for the global chip industry, a monthly check on whether the artificial-intelligence boom is still feeding through the world’s largest contract chipmaker. The July numbers, released Aug. 10, came in strong: revenue of 467.58 billion New Taiwan dollars, about $14.5 billion, up 44.7 percent from a year earlier and the second consecutive monthly record.

The growth was driven by advanced-process orders from customers including Nvidia, whose AI accelerators are manufactured in TSMC’s newest fabs, according to people familiar with the company’s order book. The company’s 3-nanometer and 2-nanometer lines are running at or near full capacity, and customers have queued for space in its upcoming plants.

The result essentially confirms the guidance TSMC issued earlier this year, when it told investors to expect revenue growth of more than 40 percent for 2026. July’s figures, which followed a record June, put the company on track to meet or exceed that target, analysts said, barring a sharp slowdown in the second half.

The report rippled through Asian markets. Semiconductor equipment makers rallied on the news, with shares of companies that supply TSMC’s fabs rising in Tokyo and Taipei. The logic, traders said, was simple: if TSMC is running at capacity and adding more, the equipment makers that build its lines have years of orders ahead.

The numbers also reflect a broader imbalance in the chip market. Memory prices have surged this year amid a shortage that has pushed DRAM and high-bandwidth memory to record levels, but the constraint on advanced logic is different: it is capacity, not materials. TSMC’s leading-edge fabs are effectively sold out, and the company has told customers it will allocate space based on long-term commitments.

That scarcity has given TSMC unusual pricing power. The company raised prices on advanced processes this year and has moved to sign multi-year agreements with major customers that lock in both volume and price, according to people familiar with the negotiations. The strategy protects the company from the cyclical downturns that have historically punished chipmakers.

TSMC’s financial position has strengthened accordingly. The company’s operating margin has expanded through the year, and its cash pile has grown as customers prepay for capacity. The company has committed to spending more than $40 billion annually on new fabrication plants, including its expansion in Arizona and new facilities in Japan and Germany, funded largely from operations.

The record month comes with a caveat that investors have learned to track: the concentration of demand. A handful of AI-chip designers, led by Nvidia, account for a growing share of TSMC’s revenue, and any stumble in that group would hit the company hard. TSMC’s management has acknowledged the risk while arguing that the diversity of AI applications, from data centers to edge devices, spreads the exposure.

The company is also managing a geopolitical balancing act. Washington has pressed TSMC to expand in the United States, while Beijing watches the company’s exposure to China, which still accounts for a meaningful share of its revenue. The company has proceeded cautiously, expanding in Arizona while keeping its most advanced production in Taiwan, and has told investors it can reconfigure capacity if tensions escalate.

For the broader industry, TSMC’s July numbers are a leading indicator. The company is the first major chipmaker to report monthly results, and its figures often presage the quarterly reports of suppliers, customers, and competitors. The strength of July suggests that the AI buildout, which some analysts worried would cool after an extraordinary 2025, retains momentum into the second half.

The memory shortage has reinforced TSMC’s position. With DRAM and HBM prices at historic highs, system makers are paying more for memory even as they buy logic chips from TSMC, and the company’s ability to supply the scarce advanced-logic capacity makes it the bottleneck partner of choice. Analysts said the dynamic effectively transfers pricing power up the supply chain to the most constrained nodes.

TSMC’s customer list for its newest process generation is telling. Beyond Nvidia, the company is manufacturing AI accelerators for AMD, Google, and Amazon, and has won orders from a new wave of startups designing custom chips. The 2-nanometer node, which enters volume production this year, is already fully allocated, according to people familiar with the company’s planning.

The company’s stock has responded. TSMC shares are among the best-performing large-cap technology stocks this year, and the ADR market in New York has tracked the Taiwan-listed shares higher on each record monthly report. Options activity suggests investors are positioned for continued strength, though the company’s valuation, at a premium to its historical range, leaves little room for disappointment.

What could change the picture? A sharp decline in AI spending by the largest cloud companies, a further tightening of U.S. export controls, or a fast resolution of the memory shortage that frees system makers to renegotiate prices. For now, none of those scenarios is visible in the order books, and TSMC’s August report will be the next data point.

The company has also begun to raise prices for 2027, according to people familiar with the discussions, signaling that it sees the capacity crunch lasting into next year. Customers have pushed back but are largely accepting the increases, given the absence of alternatives at leading-edge scale. That pricing power, reflected in each month’s record revenue, is the clearest measure of how the AI boom has redrawn the economics of chipmaking.

Related Posts

  • September 6, 2026
  • 10 views
Anthropic Moves Its IPO Filing to Late September

The bankers and lawyers running Anthropic’s initial public offering had told investors to expect the company’s registration documents as soon as this week. The calendar has moved. Anthropic now plans…

  • September 6, 2026
  • 11 views
OpenAI Quietly Revises GPT-6 Astra Scores After Launch

When OpenAI released GPT-6 Astra on Sept. 3, the launch post carried the usual furniture of a modern model debut: coding results, speed comparisons and a figure for how often…