share_log

黄仁勋打消Blackwell供需疑虑 英伟达(NVDA.US)股价再创新高

Huang Renxun dispels Blackwell's supply and demand concerns, nvidia (NVDA.US) stock price hits new high again.

Zhitong Finance ·  Nov 22, 2024 08:11

Blackwell's accelerator chip product line has been "fully put into production"; after the revenue outlook fell short of the highest expectations and caused a decline in the stock price, the stock price ultimately rose on Thursday, hitting a new high.

Fintech News App learned that at the earnings conference, CEO Huang Renxun of NVIDIA (NVDA.US) assured investors that its new product line, the Blackwell architecture AI GPU series products, will continue to lead performance significantly expand the trend driven by the artificial intelligence boom, completely dispelling Wall Street's concerns about Blackwell's capacity and demand, driving NVIDIA's stock price to a historic high of $152.89 on Thursday mid-session, leading most chip stocks to rise. Huang Renxun stated that the design flaws of the Blackwell series products have been completely resolved and are now "fully in production", with shipments expected this quarter, accelerating supply over the next year. Emphasizing that the highly anticipated high-performance AI GPU series product demand is "very strong", demand will far exceed supply for several quarters.

The Blackwell architecture AI GPU series products are the core catalyst that NVIDIA can maintain and even accelerate its unparalleled performance growth pace over multiple quarters, attracting significant attention from Wall Street. At the earnings conference after the strong quarter financial report release, CEO Huang Renxun stated that the eagerly awaited Blackwell series products will ship this quarter, with "very strong" demand. However, NVIDIA's management, including Huang Renxun, also indicated that the production of these cutting-edge chips and some unexpected engineering costs would put pressure on NVIDIA's profit margins.

In addition, NVIDIA's official revenue forecast for the current quarter does not match some extremely harsh forecasts from Wall Street, leading to some Wall Street analysts questioning the limiting nature of the dizzying growth, which also caused NVIDIA to plummet significantly in post-market and pre-market trading after the financial report was released, dropping by over 3% during Thursday's session. However, the market's positive expectations for Blackwell's supply and demand continued to warm up, outweighing the negative impact of strict revenue outlook.

Huang Renxun's optimistic comments on Blackwell's capacity and demand expectations, combined with NVIDIA's unparalleled quarterly performance and an increasing number of analysts believing that "management's revenue forecasts lean towards conservative", eventually sparked a mild reaction from investors. NVIDIA's stock rose by 0.53% on Thursday, closing at $146.670, hitting an all-time high mid-session. NVIDIA did drop by over 3% at one point during the session, but quickly recovered, demonstrating very strong "buying on dips" power.

In 2024, under the wave of the "AI belief" sweeping the globe, AI believers drove NVIDIA's stock price to soar by nearly 200%, making it the world's largest company by market cap. In 2023, NVIDIA's stock price had already surged over 240%. After experiencing a dizzying rise, any degree of underperformance in performance after continuous explosive growth is bound to disappoint.

NVIDIA expects total revenue for the fourth quarter of the current fiscal year (ending January next year) to reach approximately $37.5 billion, with a fluctuation of 2%. Data compiled by institutions shows that despite Wall Street analysts' average forecast of $37.1 billion, some analysts have a highest forecast of $41 billion. The latest financial report data shows that in the third quarter of the 2025 fiscal year, ending October 27, NVIDIA's total revenue skyrocketed by 94% year-on-year, reaching a historic highest revenue scale of $35.1 billion. Under GAAP, NVIDIA's Q3 net income surged 109% year-on-year to $19.3 billion.

big

The datacenter business unit is currently Nvidia's most core business segment, providing the H100/H200 and Blackwell architecture ai GPUs that offer incredibly powerful ai computing infrastructure for datacenters globally. Financial reports indicate that as the wave of global ai moves forward, this business unit's revenue has surged 112% year-on-year, reaching an astonishing 30.8 billion dollars. This figure greatly exceeds wall street's expectations and surpasses the combined revenue of intel and AMD.

Nvidia's performance has comprehensively surpassed Wall Street expectations in recent years. In the past five years, Nvidia has only had one quarter where its performance failed to meet the average expectations of Wall Street analysts. Moreover, in recent quarters, its actual total revenue has even exceeded Wall Street's expectations by at least 20%, leading analysts to set very high performance expectations for it in recent quarters. However, Nvidia consistently manages to exceed these expectations significantly, highlighting that with the emergence of significant generative AI applications such as ChatGPT and Sora, human society is gradually entering the AI era, and the demand from global enterprises and core government departments for the most essential infrastructure of artificial intelligence (AI) – Nvidia’s high-performance AI GPUs – continues to experience explosive growth.

Huang Renxun enthusiastically exclaimed that the demand for Blackwell is "incredible" and is now in full production.

Huang Renxun stated that the Blackwell series products are now in the "full production" stage, and the market demand for the previous Hopper architecture products (H100/H200) remains strong.

In the performance statement, Huang Renxun enthusiastically declared the market demand for Blackwell and Hopper as "incredible". "The era of artificial intelligence is progressing comprehensively, driving the global shift to NVIDIA's computing power." "As base model manufacturers expand the scale of AI pre-training, post-training, and cloud AI inference computing power deployments, the expectations for demand for Hopper and the demand for the fully launched Blackwell are unbelievable." NVIDIA's CFO also revealed that in the third quarter, 13,000 initial open artificial intelligence Blackwell engineering samples were delivered to customers, aimed at starting testing to prepare their AI server systems and software for "bulk shipments starting this quarter."

"Artificial intelligence is changing every industry, every company, and every country," "With the comprehensive breakthrough of physical artificial intelligence, investments in the industrial AI robot field are surging, and countries have realized the importance of developing sovereign-level artificial intelligence large models and infrastructure." Huang Renxun stated in the performance declaration. "The era of artificial intelligence has arrived, it is vast and diverse."

"NVIDIA's latest revenue outlook seems to indicate a slight slowdown in growth, but this may be a conservative forecasting strategy adopted by NVIDIA." Analyst Alvin Nguyen from Forrester Research said. "In the short term, there is no need to worry about the demand for artificial intelligence. NVIDIA is doing everything they should be doing."

In the MLPerf Training benchmark, Blackwell's performance in the GPT-3 pre-training task is twice that of Hopper per GPU. This means that using Blackwell can complete model training faster with the same number of GPUs. For the LoRA fine-tuning task of the Llama 2 70B model, Blackwell's performance per GPU is 2.2 times that of Hopper, indicating Blackwell's higher efficiency in handling specific high-load AI tasks. In MLPerf Training v4.1, in terms of graph neural networks and Text-to-Image benchmark tests, Blackwell's performance per GPU is increased by 2 times and 1.7 times compared to Hopper.

UBS Group, Citigroup, and other investment institutions are very bullish on Blackwell's future sales scale, indicating that the 'Big Six tech giants' in the USA as well as major global companies will continue to invest heavily in purchasing Blackwell for large-scale deployment or strengthening AI training/cloud AI inference computing infrastructure.

Wall Street has been closely following the release of the latest product Blackwell in the AI infrastructure field, known for its speed and comprehensive performance, with a significantly improved connectivity with other semiconductor devices. However, challenges in manufacturing capacity have led to a slower rollout than initially expected. Intel warned at an earnings conference that despite some improvements in the supply situation, there are still constraints preventing the fulfillment of all received orders. Huang Renxun predicts that the demand for Blackwell products in the coming quarters will far exceed supply.

"The key concerns surrounding Blackwell's production growth and customer concentration remain primary focus." Analyst Jacob Bourne of Emarketer stated in a report. "There is hardly any room for execution errors by 2025."

But the shift to Blackwell has impacted the profitability of this chip giant, with Intel predicting a decline in the company's gross margin from 75% in the previous quarter to 73% this quarter. Intel expects this number to rebound as the new products achieve larger-scale production and more favorable economic benefits.

Following Nvidia's explosive performance growth announcement, Wall Street investment institutions have significantly raised their target stock price for Nvidia within the next 12 months, emphasizing the continued demand for Blackwell surpassing supply. Analyst Stacy Rasgon from Bernstein raised Nvidia's target price from $155 to $175 post-earnings announcement. "The demand for Blackwell products is very high, and this demand may exceed supply in the future. With strong growth in the datacenter business, the outlook for the datacenter market in the next year is very optimistic, which is a significant bullish factor for Nvidia," Rasgon stated, giving Nvidia an 'outperform' rating.

Disclaimer: This content is for informational and educational purposes only and does not constitute a recommendation or endorsement of any specific investment or investment strategy. Read more
    Write a comment