On July 25, 2026, the US AI computing sector performed strongly, becoming the market focus. By the close, NVIDIA stock rose 5.2% to $168.9, leading the semiconductor sector. AMD, Broadcom, and other computing-related stocks also gained 3% to 4%. On the news front, NVIDIA revealed at its annual AI Technology Summit today that customer orders for its next-generation Blackwell Ultra GPU have increased by over 200% compared to the same period for the previous Hopper architecture product, with order schedules extended to Q2 2027.
Explosive Growth in Computing Demand
NVIDIA CEO Jensen Huang said in the keynote: "We are witnessing a turning point where AI is fully shifting from 'training' to 'inference.' The Blackwell Ultra GPU is specifically optimized for large-scale inference tasks, with a 40% improvement in energy efficiency, prompting cloud service providers and enterprise customers to accelerate upgrades." According to data disclosed on site, the Blackwell Ultra GPU integrates over 300 billion transistors on a single chip, achieving 350 TFLOPS of FP8 computing power, 1.8 times that of the previous generation.
Meanwhile, TSMC announced today that its 3nm process and advanced packaging capacity have been fully booked by major customers, with capacity utilization expected to remain above 95% in the first half of 2027. Market analysts noted that the persistent tightness in TSMC's CoWoS packaging capacity reflects the strong demand for AI chips. Macquarie semiconductor analyst Chen Hao said: "Since the second half of 2025, capital expenditure of the world's top four cloud service providers (Amazon AWS, Microsoft Azure, Google Cloud, Oracle) has increased by more than 50% year-on-year for four consecutive quarters, with most spending on AI infrastructure procurement. As the 'new oil' of the AI era, computing demand is expected to remain rigid through 2030."
Supply Chain Bottlenecks and Investment Risks Coexist
Despite the optimistic industry outlook, some experts have highlighted potential risks. First, the bottleneck in advanced packaging capacity may cause GPU delivery delays, affecting the pace of revenue recognition. According to supply chain sources, the yield rate of NVIDIA's Blackwell Ultra GPU is currently only around 60%, and TSMC is accelerating process improvements, but significant short-term improvement is unlikely. Second, valuations in the AI computing sector are at historical highs. For example, NVIDIA's forward price-to-earnings ratio is still 45 times, far above its historical average of 30 times. Third, geopolitical factors cannot be ignored. The US Department of Commerce recently expanded the scope of export controls on AI chips to China, including some high-end AI servers, which may impact overseas revenue for some companies.
"Computing demand is real, but market expectations may be overly optimistic," said well-known Wall Street short-seller and Muddy Waters founder Carson Block on social media. "When every company claims to have an AI strategy, the actual implementation results need time to verify. Investors should be wary of the possibility of overcapacity, especially after multiple new wafer fabs come online in 2027."
Institutional Views Diverge: Long-Term Bullish but Short-Term Cautious
Goldman Sachs' latest research report maintains an "overweight" rating on the AI computing sector, raising its target price for NVIDIA from $175 to $195. Morgan Stanley holds a neutral stance, believing that current valuations already reflect growth expectations for the next two years. Domestically, CITIC Securities' chief technology analyst Xu Ying said: "The A-share AI computing supply chain, such as optical modules, servers, and liquid cooling, benefited from the spillover of global computing demand, with generally better-than-expected results in the first half of 2026. However, attention should be paid to the volatility risk that may be triggered by the earnings season of US tech giants."
As of the close on July 25, the S&P 500 rose 0.3%, and the Nasdaq Composite rose 0.7%. Market attention is now focused on the upcoming earnings reports of cloud giants such as Microsoft and Google next week, when the return on AI investment will be further scrutinized. Multiple stock commentators remind investors that the long-term trend in the computing track is clear, but short-term chasing highs requires caution, and it is recommended to accumulate high-quality targets with core technical barriers on dips.
