Three Independent Signs That the AI Buildout Is Still Accelerating

Last Updated on 11/09/2026

TSMC reported that its August revenue jumped 53.3% to NT$514.8 billion, equivalent to roughly $16.3 billion. The result was stronger than the 46.8% growth analysts had anticipated for the quarter, although a single month is not enough to confirm a quarterly earnings beat. More importantly, semiconductor capacity remains tight despite TSMC’s aggressive expansion. The company is currently building and equipping around 20 factories, compared with its historical pace of four or five facilities at the same time, yet it still cannot fully satisfy demand. Its requirements for chipmaking equipment have also nearly doubled since the end of last year.

Financial Times AI Market Watch displays rising NVIDIA, Microsoft, Alphabet, and AI sector charts.

TSMC’s investment plans underline the strength of its expectations. In July, the company increased its 2026 capital expenditure forecast to between $60 billion and $64 billion while also raising its revenue outlook. The move reflects TSMC’s confidence that AI-related demand will remain strong through 2027 and beyond. In other words, the world’s leading AI chip manufacturer is not only reporting rapid growth but is also making multibillion-dollar investments based on the assumption that demand will remain elevated for years.

Elsewhere, Bloomberg reported that Huawei’s proposed price for its Ascend 950DT accelerator has climbed by around 60% over the past three months to approximately $37,300. DeepSeek is reportedly planning to deploy at least 160,000 of these accelerators at a data center project in Inner Mongolia. That represents a significant commitment to physical AI infrastructure from a company widely recognized for emphasizing efficiency. While TSMC’s results and capacity constraints provide the strongest evidence of current demand, DeepSeek’s planned deployment points to continued spending ahead.

Meanwhile, Nvidia-backed MediaTek reported that August sales increased 44% to NT$64.2 billion, or around $2 billion. This was well above the 10% quarterly growth rate analysts had expected. MediaTek plans to begin mass-producing its first custom AI chip for a major U.S. cloud provider in the fourth quarter. Although analysts believe Google is the customer, MediaTek has not officially confirmed its identity.

The company forecasts approximately $2 billion in AI chip revenue this year and aims to capture 15% of an estimated $80 billion market next year. That would represent roughly $12 billion in potential revenue, although this remains an ambitious target rather than guaranteed sales. Nvidia has also agreed to purchase up to $3.5 billion in MediaTek convertible bonds, providing additional funding for its expansion.

The broader infrastructure implication is that hyperscalers’ efforts to reduce computing costs are creating another wave of investment in custom silicon. Specialized chips require not only manufacturing capacity but also advanced packaging, memory and other supporting infrastructure. If lower computing costs make it economically viable to run more AI workloads, this additional investment could help prolong the current spending cycle.

Taken together, these three developments provide complementary evidence that the AI infrastructure spending cycle remains strong and could prove durable. TSMC offers evidence of strong realized demand and persistent capacity constraints, highlighting the intensity of the current buildout. Huawei and DeepSeek point to continued accelerator scarcity and future infrastructure spending in China. MediaTek, meanwhile, demonstrates that custom AI silicon is generating new revenue opportunities and expanding the number of suppliers serving the market.

Comments

Leave a Reply

Discover more from THE ETERNAL SOVEREIGN

Subscribe now to keep reading and get access to the full archive.

Continue reading