The first quarter financial report of Nvidia released in the early morning, in my personal opinion, is beyond expectations when the market expects it to be average or not, but in the end, it is not as bad as everyone thought. Nvidia's Q1 revenue growth slowed to a two-year low of 69%, still higher than analysts' expectations; The quarterly EPS was 13% lower than expected, but if H20 and tariff effects are excluded, it is still 3% higher than expected. The gross profit margin dropped to 61%, but after excluding H20 effects, it is still higher than 70%; Prior to the restrictions, H20 sales revenue for the quarter was $4.6 billion, with H20 related expenses of $4.5 billion due to export restrictions, lower than Nvidia's estimated provision of $5.5 billion in April; The revenue guidance for the next quarter is expected to increase by 50%, and considering the impact of reduced H20 revenue, the gross profit margin guidance is expected to rebound to 72%; Blackwell NVL72 AI supercomputer is fully put into production, and Blackwell Ultra is expected to start shipping in Q2. According to the financial reports of several technology companies in the first quarter, there is an explosive growth in the demand for AI inference. Top cloud providers such as Microsoft, Amazon, and Google have all disclosed "unexpected token (AI computing unit) usage growth," directly reflecting a surge in actual calls from end users to AI models. The inference requirements are driven by real business rather than pure capital expenditure to support the construction of training clusters. Behind the growth of inference lies the increase in practical applications and user demands, which should be the key areas that AI needs to focus on in the future.