Nvidia Earnings Announcement D-1: AI Semiconductor Demand Remains Strong

Recent reports from Russian media about a ceasefire agreement with the United States and free passage through the Strait of Hormuz have created expectations of reduced geopolitical risk in the market. This news led to a drop in international oil prices, with West Texas Intermediate (WTI) falling to $82, positively impacting the market. Kim Jang-yeol, Center Head, noted that while the Russian news might be temporary or uncertain, it sparked hopes for easing geopolitical tensions.

The U.S. Treasury Department is pursuing policies to supply market liquidity, including Treasury bond buybacks and utilizing the Treasury General Account (TGA) balance. While these policies are criticized as temporary stopgaps, they are positively impacting market liquidity and helping to lower interest rates. The 10-year Treasury yield, for instance, recently fell to 4.62%.

This phenomenon is interpreted as the market pre-reflecting expectations for earnings. However, on the day of the earnings release, due to these pre-reflected expectations, the stock price fell 8 or 9 out of 10 times.

Kim Jang-yeol, Center Head, explained that on average, revenue was about 6% higher than expected, and EPS was about 10% higher. This demonstrates Nvidia's consistent ability to meet high market expectations.

The consensus for Nvidia's second-quarter earnings is an estimated revenue of $92 billion and EPS of $2.8 cents. This represents a growth rate of approximately 96% year-over-year. Kim Jang-yeol, Center Head, emphasized that since Nvidia has historically achieved surprises of 6-10% on average, revenue in the high $90 billion range and EPS in the low-to-mid $2 range are necessary for a positive market reception this time.

Furthermore, the third-quarter revenue guidance is projected at $108 billion, representing approximately 90% year-over-year growth. Notably, the third-quarter guidance growth rate (90%) might be slightly lower than the second-quarter estimated growth rate (97%). Therefore, the market will closely examine not only whether the consensus is exceeded, but also whether this growth slowdown is a temporary phenomenon or the beginning of a long-term trend.

In this earnings announcement, it will be crucial whether Nvidia can maintain the explosive growth it has previously shown, and whether it can provide clear explanations and future outlooks regarding the anticipated slowdown in growth.

Kim Jang-yeol, Center Head, stressed that Nvidia's stance on 'AI financing' is more important than the earnings figures or guidance in this earnings announcement. This is because Nvidia is positioning itself as the 'Godfather' of AI infrastructure investment, actively engaging in funding and guaranteeing AI investments. This is interpreted as Nvidia's strategy to build a circular financial structure for expanding the AI ecosystem.

Kim Jang-yeol, Center Head, highlighted the potential impact of the political situation on AI investment, stating that if the Democratic Party takes power, there is a higher likelihood of increased regulations or demands for AI infrastructure investment compared to the Republican Party. Specifically, demands related to social values such as environment, power, and water could increase, potentially delaying data center construction and AI investment. The stock market is expected to react sensitively to public opinion data related to the presidential election.

Large-scale capital-intensive data center construction projects require the support of falling interest rates and active financial backing from lenders to be sustainable. In short, money is essential for data center investment to continue.

Kim Jang-yeol, Center Head, raised a fundamental question about whether data centers, even though they will continue to be built next year and the year after, will actually be utilized properly once completed. To maintain actual data center utilization and profitability, the demand for token production in AI services must continue to grow. In other words, the growth of data (token) demand to be processed within the centers is as important as the expansion of the centers themselves.

Kim Jang-yeol, Center Head, pointed out that AI token prices tend to fall by 80-90% annually. For example, a price of 1 in the previous year drops to 0.2. To offset this price decline and maintain investment value, the volume of tokens, or the number of tokens, must increase explosively, he emphasized.

Google, for instance, saw a 300-fold increase in token usage over the past two years, which translates to 15-20 times growth annually. However, given the high rate of price decline, to generate actual investment value, the volume must grow more than 10 times annually compared to the present, explained Kim. He warned that if such volume growth is not sustained, the profitability of AI infrastructure investments could decline.

Therefore, the core of AI investment depends on the continuous and explosive growth of token volume, which can overwhelmingly compensate for the decline in token prices.

Global AI investment is expected to expand further to $1 trillion this year and $1.4-1.5 trillion next year. However, as mentioned earlier, token prices will continue to fall. Therefore, for the sustained growth of the AI industry, the token growth rate must exceed current levels by nine or ten times, or even higher, according to Kim Jang-yeol, Center Head. This means that even with continuous large-scale capital expenditures (CapEx), explosive growth in corresponding token demand must be guaranteed.

Kim Jang-yeol, Center Head, stated that AI semiconductor demand is robust. He emphasized that while some weak links like Oracle account for about 25% of overall AI demand, the remaining 75% of demand has seen no problems over the past two months. The market's instability was due to manipulation by funding sources such as hedge funds, and the fundamental demand for AI itself is not problematic; in fact, it has increased.

He analyzed that if the number of agents increases to 4-5 and operates simultaneously 50-100 times, significantly more memory than anticipated will be needed. This is based on the premise that the number of tokens in the coding field could increase by a minimum of 100 times, and possibly up to 1,000 times.

This surge in token demand will drive demand for various types of memory, including High Bandwidth Memory (HBM), as well as general DRAM and SSDs. Kim described this as 'Memory Tiering,' predicting that AI agents will exponentially increase data processing throughput and computational complexity, significantly impacting the overall memory ecosystem.

In conclusion, the development and proliferation of agent AI will be a growth engine for the AI semiconductor and memory industries, requiring not just quantitative but also qualitative changes.

Kim Jang-yeol, Center Head, stated that institutional investors have not reduced their AI exposure itself. Both hedge funds and general mutual funds commonly bought energy-related stocks (transformers, nuclear power, etc.). In contrast, mutual funds began buying memory-related stocks such as SanDisk and Micron from August, signaling a recovery in investment sentiment for the memory sector.

Kim Jang-yeol, Center Head, emphasized that the recent trend in Nvidia's stock price is not a collapse. He explained that large tech companies like Nvidia and Google have merely shown range-bound movements due to changes in the macroeconomic environment and interest rate hikes, without significant declines. In particular, unlike memory sector stocks that fell sharply, Nvidia and SanDisk stocks maintained relatively robust performance.

While Nvidia shows strength, the growth of the custom ASIC (Application-Specific Integrated Circuit) sector is also drawing attention. Networking chips produced by Marvell are crucially used in Google's TPUs and Amazon's custom chips. This shows that Marvell is forming another growth axis with custom ASICs alongside the AI market led by Nvidia, with both axes showing simultaneous growth.

Kim Jang-yeol, Center Head, assessed that the market's extreme negative scenarios have already been sufficiently priced in. Various political variables, including both Republicans and Democrats, have all been exposed, and he predicted that it would be very difficult for the index to fall below 5,500 to 6,000 unless interest rate hikes, high oil prices, and poor performance from big tech companies all occur simultaneously. In other words, the worst of the market situation has passed, he concluded.

He mentioned that Google's token processing volume increased by 10 to 15 times year-over-year, emphasizing that such figures are objective data readily available. He asserted that as long as token processing volume explodes in this manner, the AI boom will not collapse.

The increase in token processing volume directly translates to an increase in AI service usage and computational demand, which in turn is a direct basis for the demand for AI servers and semiconductors. Kim explained that this data supports the robust growth of the AI industry and shows that the core driving force of the AI boom continues despite short-term market volatility.

Therefore, the increase in AI token processing volume can be interpreted as a crucial indicator that goes beyond mere numbers, demonstrating how actively AI technology is being used in the real world and its significant potential for future growth.