Taiwan Semiconductor Manufacturing Co. reported revenue of NT$592.5 billion for Q4, up 12% year-over-year, driven by surging orders for Nvidia’s custom AI accelerators. TSMC’s data-processing unit (DPU) and high-performance computing segments grew 18% sequentially, signaling hyperscaler capex for generative AI remains on a steep upward trajectory. Nvidia executives have referenced a roughly US$500 billion backlog for its AI products; TSMC’s outperformance provides independent validation of that order flow and underpins Nvidia’s manufacturing roadmap for 2026 and beyond.

Analyst consensus from FactSet and Goldman Sachs forecasts AI hyperscalers will spend up to US$527 billion on infrastructure capex this year. Nvidia CFO Colette Kress has characterized the company’s order pipeline as growing exponentially. With recent agreements—such as AWS’s US$38 billion GPU cluster arrangement with OpenAI and a US$20 billion license deal with Groq—Wall Street models now project Nvidia could achieve between US$320 billion and US$330 billion in data-center revenue by 2026. Assuming the business captures 60% of big-tech AI spending and maintains a forward price-to-sales multiple near 21x, Nvidia’s market capitalization could expand to US$7–9 trillion, representing roughly 70% upside to current levels.

Nvidia’s forward price-to-earnings multiple, which peaked near 40x during hypergrowth phases, has compressed to approximately 24x following a 19% share-price pullback in early 2025. That sell-off was driven largely by fear over competitive entrants such as DeepSeek and prospective tariff changes, rather than by any deterioration in Nvidia’s core AI revenue growth—data-center sales grew 23% in the last reported quarter. Comparable P/E troughs in 2023 and 2024 preceded subsequent rallies of over 80%. If earnings estimates continue to climb and sentiment rotates back to a growth narrative, history suggests Nvidia shares could re-test prior highs.

On January 17, Chinese customs authorities held shipments of Nvidia’s H200 GPU modules, citing documentation discrepancies tied to export regulations. Key suppliers of HBM memory and power‐delivery components have paused production pending clarification, creating a short-term bottleneck for Nvidia’s next-generation inference accelerator in Greater China markets. While Nvidia is working closely with regulators to resolve the issue, the delay could shave an estimated 2–3% off projected first‐quarter data-center revenue in the region and prompt customers to re-allocate orders to alternative architectures.