At the beginning of the new year, chip giants have successively released their 2023 performance reports.
On January 31st, AMD released its financial reports for the fourth quarter and full year of 2023. The financial report shows that AMD's revenue in the fourth quarter of 2023 was $6.2 billion, a 10% increase year-on-year, and its annual revenue was approximately $22.7 billion, a 4% decrease year-on-year.
Screenshot from AMD financial report data
Coincidentally, on January 26th, Intel also disclosed its financial data for the fourth quarter and full year of the 2023 fiscal year. According to financial report data, Intel's revenue in the fourth quarter of 2023 was $15.41 billion, a year-on-year increase of 10%, and its annual revenue was $54.2 billion, a year-on-year decrease of 14%.
Screenshot from Intel's financial report data
The performance guidelines of the two companies for the first quarter of 2024 cannot be accepted by the market. On the day of performance release, the stock prices of the two companies fell by more than 10% and 6% respectively.
On January 31st, the release of AMD's performance further triggered a decline in the stock prices of AI related listed companies. According to Reuters statistics, the total market value of artificial intelligence related companies evaporated by 190 billion US dollars (about 1.3 trillion yuan) at the close of the day on the 30th.
However, on February 1st, the stock prices of AI related listed companies in the US stock market rebounded. Intel rose 0.65% to $43.360 per share, while AMD rose 1.66% to $170.480 per share.
"As long as the big model finds a successful commercial landing, AI will continue to drive the development of the chip industry," Lin Zhi, Chief Analyst of Wit Display, explained to Time Weekly reporters. Currently, the AI chip market is dominated by Nvidia, and it is difficult for other manufacturers to catch up. From the financial report, AMD has not successfully divided Nvidia's share.
However, as a leading global GPU company, the market is anxiously waiting for Nvidia's answer. According to statistics from Wells Fargo Bank, Nvidia currently has over 90% market share in the data center AI market, while AMD has only 1.2% market share and Intel has less than 1%.
Performance below expectations
Behind the slightly lackluster performance reports of the two major chip giants are mixed mixed results in various sub businesses.
In 2023, Intel's revenue from multiple businesses showed a downward trend. Among them, the revenue of the Client Computing Group (CCG) was $29.3 billion, a decrease of 8% compared to the same period last year; Data centers and artificial intelligence (DCAI) amounted to $15.5 billion, a decrease of 20% compared to the same period last year; Network and Edge (NEX) revenue was $5.8 billion, a 31% decrease from the same period last year.
Only its intelligent driving subsidiaries Mobileye and Intel Manufacturing Services (IFS) achieved growth, with Mobileye's revenue reaching $2.1 billion, an 11% increase from the same period last year; IFS revenue was $952 million, a year-on-year increase of 103%. But the proportion of the two to the total income is less than 10%.
Unlike Intel, AMD achieved revenue growth in the data center and embedded departments last year, but was offset by revenue from the client and gaming departments.
Due to the decline in the PC market, AMD's client business unit achieved a revenue of $4.7 billion in 2023, a year-on-year decrease of 25%; The revenue of the gaming business unit decreased by 9% year-on-year to $6.2 billion, mainly due to a decrease in sales of semi customized products.
In 2023, the annual revenue of AMD's data center division was 6.5 billion US dollars, a year-on-year increase of 7%; The embedded business unit's annual revenue increased by 17% year-on-year to 5.3 billion yuan, mainly due to the merger of the acquired Xilinx's annual revenue, which was completed in February 2022.
From a data perspective, AMD has achieved certain results in its transformation in AI and data centers. During the earnings conference call, AMD Chairman and CEO Lisa Su stated that AMD's data center GPU business significantly accelerated in the fourth quarter of last year, with revenue exceeding expectations of $400 million due to growth in MI300X accelerators and AI customers.
Screenshot from AMD financial report data
In the first quarter of 2024, both AMD and Intel provided performance guidance for the first quarter that was lower than market expectations.
AMD expects an expected revenue of $5.1-5.7 billion for the first quarter of 2024, with a gross profit margin of approximately 52% based on non US GAAP accounting standards; The revenue of the data center department will remain unchanged from the previous quarter, while sales of the client, embedded, and gaming departments are expected to decline. Among them, revenue from semi customized products (mainly including game console SoCs) is expected to show a significant double-digit decline.
Intel expects its revenue range for the first quarter of 2024 to be between $12.2 billion and $13.2 billion, far from the average forecast of $14.25 billion by Wall Street analysts.
Wu Quan, President of Huaxin Jintong Semiconductor Industry Research Institute, analyzed to Time Weekly reporters that from the development pattern of the semiconductor industry, the revenue in the first quarter of each year is basically the lowest. Compared to the fourth quarter of the previous year, there will be a significant decline in revenue, which is a normal phenomenon.
"At present, both companies should be able to see the order volume and revenue situation in the first quarter, and their forecasts should be relatively accurate and realistic, without blindly catering to market judgments," said Wu Quan.
Keep up with NVIDIA
In the past, AMD and Intel competed for the CPU dominance in the chip market, but did not want Nvidia to take advantage of GPUs and become the king.
However, Intel and AMD are catching up at full speed.
Last December, AMD released the new MI300 series AI chips in an attempt to challenge Nvidia's dominant position in the field of artificial intelligence chips. AMD claims that the MI300 series chips perform better than Nvidia's H100 graphics cards and have received orders from Microsoft, Oracle, Meta, and OpenAI.
Su Zifeng stated during the financial report conference call that customers have responded very positively to the MI300, and AMD is actively increasing production to support dozens of cloud, enterprise, and supercomputing customers deploying Instact accelerators.
Meanwhile, AMD has also raised its forecast for data center performance in a conference call. Su Zifeng stated that AMD had previously predicted that the revenue of its data center division would remain the same as the previous quarter, with revenue exceeding $2 billion in the first quarter of 2024. However, this figure is now expected to reach 3.5 billion yuan.
Su Zifeng pointed out that the data center artificial intelligence accelerator market will grow to approximately $400 billion by 2027. In the future, AMD will continue to deploy Instact GPU customers. The MI300 is AMD's fastest-growing product in history and will occupy a significant market share in the coming years.
Meanwhile, it is reported that Intel will launch the third-generation Intel AI accelerator Gaudi 3 for deep learning and large-scale generation of AI models in 2024. Gaudi 2 and Nvidia H100 are in the same tier, and Gaudi 3 will benchmark against Nvidia H200 accelerator after its launch.
However, Wu Quan believes that the current performance of AMD's MI300 cannot be compared to NVIDIA's H100. For example, the H100 can achieve a maximum throughput of thousands of TFLOPS, while the MI300 is only a few hundred. The process node of H100 is 4nm, while the MI300 is 5nm. In addition to performance, what is more important is the architecture and ecological construction of the GPU, including the accompanying software and toolset, as well as professional insights and understanding of the industry. The gap between the two companies is even more incomparable.
"But AMD's advantage lies in the fact that compared to Nvidia, which uses CPUs from Intel, it does both GPUs and CPUs, which are relatively integrated. When integrating chips, they are relatively integrated, with better compatibility and smoother performance." He said that Intel, as the former "leader" in the CPU field, also has significant advantages in transitioning to GPUs. Now Intel's transformation is not too early, but it is not too late. Of course, Nvidia, which has been rooted in the GPU field for many years, still has significant advantages at present.
Lin Zhi also believes that it is difficult for Intel and AMD to compete with Nvidia because Nvidia's ecosystem is relatively complete, and the market recognition of other manufacturers is still not as good as Nvidia's.
"At present, Nvidia cannot fully meet market demand and supply is in short supply. This is an opportunity for AMD. If AMD can keep up and provide solutions similar to Nvidia, they can share a portion of the market share." Lin Zhi said.
How long can the "AI story" of chips be told?
On the day Intel and AMD released their performance reports, their stock prices fell by over 10% and 6% respectively.
According to Reuters statistics, on January 31st, the total market value of artificial intelligence (AI) related companies evaporated by trillions of yuan on the 30th, although their stock prices rebounded the next day.
However, in Wu Quan's view, these fluctuations belong to the premium of the capital market for conceptual market values such as "AI" and "chips", and now the market has slowly returned to the essence of the industry.
In addition, Nvidia's fourth quarter and full year financial reports for 2023 will be released at the end of February. As a global GPU leader, the market's expectations for it are bound to be higher than those of Intel and AMD.
"It is expected that Nvidia's answer sheet on AI chips will satisfy investors," Lin Zhi analyzed. Nvidia currently enjoys excess profits from AI chips, which will bring substantial profits to investors, and there will be no strong competitors for a long time.
However, in addition to its own competitors in the chip industry, Nvidia may also face threats from technology giants with extremely strong financial and technological capabilities.
Recently, technology giants such as Google, Microsoft, Meta, and Amazon have increased their investment in self-developed chips, hoping to reduce their dependence on Nvidia.
In the fourth quarter of last year, Google launched a new TPU v5p for cloud AI acceleration, Microsoft launched its own self-developed chip Azure Maia, and Amazon Cloud Technology launched its own Amazon Graviton4 and Amazon Traineum2 chips.
In addition, according to Reuters on February 1st, Meta plans to deploy self-developed custom chips this year to support the development of its AI business. This chip will be produced on Meta's second-generation internal chip production line and will be put into production in 2024, working in conjunction with hundreds of thousands of purchased GPUs.
These tech giants all hope to break Nvidia's monopoly. For example, Google officially showcased the self-developed TPU to the world at the 2016 Google I/O Developer Conference.
Wu Quan believes that from a business perspective, technology giants are more focused on examining strategic windows and measuring input-output. In the past, the application of AI had not yet reached a large scale, which was not enough to enable them to make localized investments. Since the outbreak of AI big model driven by ChatGPT, this AI boom is more like the Internet technology wave in 2000, and AI and its applications will gradually take root and develop on a large scale. Technology giants entered the market by stepping on these milestones. From this perspective, their strategic judgment is relatively accurate.