AMD's AI revenue doubled year-on-year in the second quarter, raising its annual sales outlook for MI300 and rising 9% after hours | Financial Report Insights

After the US stock market closed on Tuesday, in the GPU field of data centers, AMD, a semiconductor giant striving to catch up with NVIDIA, released its financial report for the second quarter of the 2024 fiscal year. Investors will closely monitor the annual sales guidance for its MI300 AI accelerator chip.

Due to AMD's second-quarter revenue and profit slightly exceeding expectations, AI revenue in data centers doubled year-on-year and hit a new high for two consecutive quarters. The midpoint of the revenue guidance for the third quarter is $6.7 billion, higher than the market's expected $6.62 billion, causing its stock price to rise by 5% after hours, leading to a turnaround in NVIDIA's stock price from a 3% decline to an increase of about 3%.

During the earnings call at 5 pm Eastern Time, AMD's management raised the sales expectations for data center GPUs in 2024 from $4 billion to $4.5 billion. They mentioned that Microsoft's usage of the MI300 chip has increased, aiming to refute market rumors about Microsoft cutting orders for the MI300, leading to a rapid expansion of AMD's after-hours gain to 9%, while NVIDIA maintained a 3% gain after hours.

1) Key Financial Data

Quarterly Revenue: $5.835 billion, up 9% year-on-year, up 7% quarter-on-quarter, market expectation $5.73 billion.

Gross Margin: 49% under GAAP, compared to 46% in the same period last year; 53% under non-GAAP, compared to 50% in the same period last year.

Operating Profit: $269 million under GAAP, compared to a loss of $20 million in the same period last year; $1.264 billion under non-GAAP, up 18% year-on-year, up 12% quarter-on-quarter, market expectation $1.25 billion.

Net Profit: $265 million under GAAP, up 881% year-on-year; $1.126 billion under non-GAAP, up 19% year-on-year, up 11% quarter-on-quarter.

Diluted EPS: $0.16 under GAAP, up 700% year-on-year; $0.69 under non-GAAP, up 19% year-on-year, up 11% quarter-on-quarter, market expectation $0.68.

2) Outlook

Third Quarter Revenue: Expected to be $6.4 billion to $7 billion, with a midpoint of $6.7 billion, representing a year-on-year increase of about 16% and a quarter-on-quarter increase of about 15%, analyst expectation $6.62 billion.

Third Quarter Non-GAAP Gross Margin: Expected to be 53.5%, analyst expectation 53.8%.

MI300 Accelerator Sales: Expected data center GPU sales in 2024 to be $4.5 billion. Microsoft's usage of the MI300 chip has increased.

Other Points from the Call: MI300 chip revenue exceeded $1 billion in the second quarter, with a commitment to launch new AI processors/chips every year 3) Segment Business Data

Data Center Division: Revenue reached a record high of USD 2.8 billion, a year-on-year increase of 115% and a quarter-on-quarter increase of 21%.

Client Division: Revenue was USD 1.5 billion, a year-on-year increase of 49% and a quarter-on-quarter increase of 9%.

Gaming Division: Revenue was USD 648 million, a year-on-year decrease of 59% and a quarter-on-quarter decrease of 30%.

Embedded Division: Revenue was USD 861 million, a year-on-year decrease of 41% and a quarter-on-quarter increase of 2%.

AMD Q2 Highlights: Data Center AI Revenue Doubled Year-on-Year, Exceeding Previous Quarter, Significant Contribution to Future

AMD's financial report also stated that the adjusted operating profit margin for the second quarter was 22%, basically in line with analysts' expectations of 21.8%. Capital expenditures for the second quarter were USD 154 million, higher than analysts' expected USD 127.1 million. Research and development expenses for the second quarter were USD 1.58 billion, in line with market expectations.

In terms of total quarterly revenue, the year-on-year growth of 9% compared to the same period last year significantly narrowed from the 18.2% growth in the second quarter of last year, while the non-GAAP gross margin continued to expand from 52% in the previous quarter, and the non-GAAP adjusted earnings per share remained higher than the previous quarter's USD 0.62.

Most importantly, the data center revenue hit a historical high for two consecutive quarters, doubling compared to the same period last year, showing a stronger growth momentum than the 80% year-on-year increase in the first quarter and a significant acceleration from the 2% quarter-on-quarter growth in the previous quarter.

AMD's management emphasized the significant contribution of AI to current and future growth in the financial statement. Dr. Lisa Su, Chairman and CEO of the company, stated in the financial report:

"The continued acceleration of our AI business, driven by demand for Instinct, EPYC, and Ryzen processors, has positioned us for strong revenue growth in the second half of this year. The rapid development of generative AI is driving the need for more computing in every market, and as we provide leading AI solutions across our entire business, this will create significant growth opportunities."

CFO Jean Hu also mentioned that second-quarter revenue exceeded the midpoint of the company's guidance range, benefiting from strong growth in the data center and client divisions. While maintaining robust profit growth, the gross margin continued to expand, "increasing strategic AI investments to lay the foundation for future growth."

AMD also committed to launching new AI processors/chips every year, catching up with NVIDIA, which had previously moved its new product release schedule from every two years to annually.

Financial Report Highlights: Accelerated Growth in Data Center AI Revenue, Steady Growth in PC Chip Revenue, Continued Weak Performance in Gaming and Embedded

AMD stated that in the second quarter, the most important data center AI revenue doubled year-on-year, benefiting from a significant increase in Instinct GPU shipments and strong growth in sales of the fourth-generation EPYC CPUs. The sequential revenue growth was driven by the strong growth in Instinct GPU shipments.

At the same time, the revenue growth in the client department, including PC personal computer chip sales, was mainly due to the sales of Ryzen processors. The year-on-year halving of revenue in the gaming department was attributed to a decrease in semi-custom revenue, while the decline in revenue in the embedded department was due to continued destocking by customers.

Prior to the financial report, the market expected a year-on-year drop of over 58% to $655 million in gaming revenue for the second quarter, and a year-on-year decline of nearly 42% to $850 million in revenue for the embedded business unit. It can be seen that AMD's gaming revenue in the second quarter was lower than expected, while the embedded revenue exceeded expectations.

In the first quarter, gaming revenue had already dropped by 48% year-on-year and 33% sequentially to $922 million, mainly due to a decrease in semi-custom business revenue and a decline in Radeon GPU sales. Revenue in the embedded business unit fell by 46% year-on-year and 20% sequentially to $846 million, primarily due to customers continuing to adjust inventory levels.

In other words, in the second quarter, AMD's chip sales in data centers and PC clients continued to grow significantly year-on-year, while gaming and embedded revenue continued to decline significantly year-on-year. However, in the first quarter, client revenue had surged by 85% year-on-year to $1.4 billion, higher than the year-on-year growth in the second quarter, but at that time, revenue in that area had decreased by 6% sequentially, unlike the 9% sequential growth in the second quarter.

This year, AMD stock has fallen significantly and seriously underperformed the market and chip stock indices, but some analysts are optimistic about the potential of various business lines

Before the financial report was released, AMD fell 4% on Tuesday and closed down 0.9%, approaching its lowest level in nearly seven months since January 4, highlighting market caution. The stock has fallen nearly 7% since the beginning of the year, significantly underperforming the cumulative gains of about 14% in the S&P 500 index and Nasdaq, as well as the 18% gain in the Philadelphia Semiconductor Index.

However, Intel, a competitor severely hit by AMD in the CPU field, has fallen nearly 40% this year, more severely than AMD, while NVIDIA, which has become the "darling of AI" with its GPU hardware and related software services, has doubled its stock price this year with a cumulative increase of 114%.

Since July 10, AMD has fallen 25% from its four-month high, becoming one of the worst-performing stocks in the S&P 500 index. However, mainstream Wall Street ratings remain "buy," with an average target price of over $190, representing about 40% upside potential.

Raymond James analyst Srini Pajjuri rates AMD as "outperforming the market/hold," believing that AMD's revenue from the MI300 AI accelerator product in the second quarter could reach $900 million, and the full-year sales guidance in this area could be "increased by several billion dollars." Currently, AMD expects the MI 300 to bring incremental revenue of over $4 billion by 2024.

The analyst stated that the personal computer PC market appears relatively stable, the server (CPU) market is showing signs of recovery, the gaming market is expected to rebound from a low point in the second half of the year, but suppliers of automotive/industrial integrated circuits suggest that the recovery in the embedded market of Xilinx may be slower Previously, according to IDC statistics, global PC shipments in the second quarter increased by 3% year-on-year, marking the second quarter of growth after eight consecutive quarters of decline. With Nintendo preparing to launch the next generation game console and video game company Take-Two set to release the highly anticipated "Grand Theft Auto VI" later next year, many institutions predict a rebound in the gaming industry by the end of 2024 and 2025.

Why It's Important and What to Watch: AI Revenue Outlook

AMD was the first of the three major chip companies to release its second-quarter financial report, with Intel and NVIDIA set to announce their reports on August 1st and August 28th respectively. Investors will closely monitor the sales performance of AMD's data center division to speculate whether enterprise spending on artificial intelligence infrastructure continues to rise. Guidance on GPU sales and management comments from the company are also of great interest.

C.J. Muse, an analyst at the U.S. financial services company Cantor Fitzgerald, pointed out that AMD has dropped more than 20% in the past two weeks, mainly due to rumors that Microsoft may reduce orders for the MI300 AI accelerator. If AMD fails to increase its annual sales guidance for this key chip in this financial report, the stock price will face further downside risks. A guidance increase to over $5 billion would be very favorable.

AMD's current top GPU chip is the Instinct MI300X accelerator. In June, the company revealed that partners and customers including Microsoft, Meta, Dell, HPE, and Lenovo have all started adopting this chip. The next-generation MI325X will be launched in the fourth quarter of this year, followed by the updated MI350X in 2025 and the MI400 in 2026.

Currently, there is a difference of opinion among mainstream Wall Street analysts on what constitutes "good enough" AI revenue for AMD.

Wedbush believes that reaching $6 billion in shipments for MI300 this year is the highest expected level. Barclays believes that even with a lot of negative news related to Microsoft in the supply chain, achieving $5 billion in annual sales is feasible. Susquehanna is concerned that after the NVIDIA GTC AI Developer Conference this year, AMD's accelerators faced cancellations, causing some degree of worry.

Wall Street's Main Concern is Whether AMD Can Effectively Compete with NVIDIA

This also indicates that the GPU battle between AMD and NVIDIA will be a highly watched focus in the market.

Rosenblatt Securities believes that the pressure on AMD's stock price this year reflects investors' concerns about whether its products can compete with NVIDIA, especially in AI accelerators and data center GPUs, where NVIDIA holds an absolute leading position with the H and stronger B series chips:

"Contrary to popular belief, AMD's MI300 development roadmap has become more competitive due to software optimization and higher configuration with high-bandwidth memory HBM3e. Compared to NVIDIA's more closed technical stance, AMD's open approach will also pay off In the second quarter report, AMD mentioned the establishment of the Ultra Accelerator Link promoter group with industry leaders, which will leverage AMD Infinity Fabric technology to advance open standard AI network infrastructure systems.

Analyst Stacy Rasgon from Bernstein Securities stated that recent news regarding potential issues with high-bandwidth memory and the planned retirement of AMD's CEO Victor Peng may pose two "less favorable" factors. However, AMD's valuation multiples have eased during recent sell-offs, with a P/E ratio of around 30 times, closer to the institution's target level.

Another analyst wrote on SeekingAlpha that it is currently difficult to evaluate AMD because of the "considerable uncertainty" regarding whether AMD can successfully compete with NVIDIA in the data center GPU market. The three key data points to focus on in the second quarter financial report are the backlog orders for MI300, the production schedule of the next-generation chip MI325, and the customer engagement in the data center GPU.

The article mentioned that in the first quarter, AMD's MI300 backlog orders only increased by about $500 million, showing slow and unsatisfactory growth. This caused the stock price to drop nearly 9% on the second day after the financial report, as "an increase of only $500 million in backlog orders in a quarter seems too small for AMD to maintain revenue growth in data center GPUs. The current surge in demand for data center GPUs should make it easier for AMD to win customers."

Backlog orders are not expected sales volume but rather "firm commitments made by customers based on their demand for AMD." The slow growth in backlog orders may not be related to the difficulty of competing with NVIDIA but rather influenced by macro factors such as customer considerations for AI infrastructure investments. If the backlog orders increase again, it will send a positive signal for AMD to maintain and increase its sales capacity for data center GPUs next year.

Furthermore, if customer engagement in AMD's data center GPUs increases and translates into actual orders more easily, it also signifies AMD's ability to compete with NVIDIA in the coming quarters. During the first quarter financial report, the management revealed that "over 100 enterprises and AI customers are actively developing or deploying MI300x, higher than the dozens in the fourth quarter of last year."

The upcoming MI325 planned for the fourth quarter of this year will upgrade the GPU memory from 192GB of HBM3 to 288GB of HBM3e. The speed of its production increase will also determine AMD's competitive position against NVIDIA's current generation Hopper GPU and the upcoming Blackwell GPU. Some analysts believe that there are no significant architectural changes in MI325, and increasing production may be easier than with MI300