The next key point of Intel's growth

2024-07-17

Is it possible for Intel to once again significantly outperform AMD in terms of sales and profits? The key to this lies in Intel's foundry business.

Intel announced its financial results for the first quarter of 2024 on April 25, while AMD announced its financial results for the same period on April 30. Prior to the release of the financial reports, Intel's stock price was at $35.11, but it fell to $31.88 the next day. Similarly, AMD's stock price fell from $158.38 before the release of its financial report to $146.64 the following day. In short, the financial reports of both companies did not meet the expectations of shareholders, but the content of these reports showed significant differences.

Let's start with Intel. Tables 1 and 2 summarize the sales and profits by division for the five quarters starting from the first quarter of 2023. Intel's current cash cow is undoubtedly its CCG (Client Computing Group), which accounts for 40% of its sales and almost all of its profits and losses. Unconsciously, Intel has transformed into a company that only focuses on client PCs. The operating profit margin of CCG in the first quarter of last year (2023) was as low as 20%, but it has since recovered to the range of 30%. It has already recovered to 35% in the first quarter of this year (2024), and the author believes that this is also a solid business for a semiconductor company.

Advertisement

The problem lies in other businesses. DCAI (Data Center & AI) is also profitable, but its sales are less than half of CCG's, with an operating profit margin of around 21% in the fourth quarter of 2023, which was the best, and slightly less than 16% in the fourth quarter of 2023.

In the first quarter of 2024, this is not a very healthy situation for a semiconductor business. In 2017, the year before AMD entered the server business with EPYC, DCG (Data Center Group) had annual sales of $19.1 billion, operating profit of $8.4 billion, and an operating profit margin close to 44%. In 2023, it will be $12.64 billion and $1.62 billion, respectively, with an annual operating profit margin of about 12.8%. Strictly speaking, in 2017, DCG was equivalent to DCAI + NEX (Network and Edge), so if you consider this, the sales would be about $18.41 billion, but the operating profit would be $1.82 billion, and the operating profit margin would drop to 10%. In fact, even considering the share of NEX, Intel's server business is stagnant, or more accurately, sales are subtly decreasing, which has to be said that this business is being damaged, and this situation will mean there is no improvement. There are signs that this situation will continue at least until the first quarter of 2024.Sales of Gaudi 3

According to Intel, the Xeon 6 based on the E-core (development code name: Sierra Forest) and Gaudi 3 will be released in the second quarter of 2024, and the Xeon 6 based on the P-core will be launched in the third quarter. The sales volume of Granite Rapids is expected to increase significantly (Figure 1), but there are also some concerns.

The Xeon 6 introduces a new platform. That is to say, the method of upgrading the CPU cannot be used on the existing platform. Therefore, it requires a considerable amount of time for verification by equipment manufacturers and customers. Because of this, even if a product is launched, the sales volume is unlikely to increase immediately, and even if there is a big move, it is likely to be at the end of 2024 or the fiscal year of 2025.

In addition to the Xeon 6 based on the P-core being the offspring of the traditional Xeon series, the Xeon 6 based on the E-core of Sierra Forest is the first horizontal expansion of the Xeon, or more precisely, it has some special features. The functions of Cloud Workload that can use existing applications as is to obtain good performance are not many. This means that it takes time for customers to migrate workloads. The reality is that for customers using local or private clouds, the Xeon 6 based on Granite Rapids may be a viable migration target, so the Xeon 6 based on Sierra Forest will not grow so quickly. For cloud providers, it will be tested in the form of comparison with new products based on Arm servers and AMD's "Zen 4c" and "Zen 5c", so it is hard to imagine the Xeon 6 based on Sierra Forest being introduced in large quantities immediately. Depending on the results, there is unlikely to be any major change in 2024, and if there is any change, it may also be after 2025.

The appeal of Gaudi 3 is that it has a better cost-performance ratio or performance/consumption ratio compared to NVIDIA's "H100" and "H200", but on the other hand, it is just the previous generation product. It cannot be compared with NVIDIA's "B100" and "B200" or AMD's "MI300X", which is the first bottleneck. Intel has announced the launch of "Falcon Shores" as the successor to Gaudi 3, but this Falcon Shores is based on GPGPU (the current Intel Datacenter GPU MAX series architecture), which is completely different from the Gaudi series. Unfortunately, the software framework of the Datacenter GPU MAX series and the Gaudi series is not compatible.

In the long run, with the launch of Sierra Forest/Granite Rapids, there is an opportunity to narrow the performance gap with AMD to some extent.

Poor performance in business outside CCG

The sales of the former PSG (Programmable Solutions Group) and the current Altera have also deteriorated significantly (Figure 2). It is explained that "from 2024, it is expected to achieve sales of $2 billion per year", but this is not much different from the sales before 2021. However, there is no explanation for the sharp decline in sales since the fourth quarter of 2024. In terms of product lineup, although the old products (Stratix/Arria/Cyclone/Max) continue to sell well, the transition of Agilex, which was launched to update these products, seems to have not progressed, which is also one of the reasons. That is, high-priced products are not selling well. This is in sharp contrast to AMD's embedded division.Intel Foundry Services

On April 2, 2024, a new segment market report webinar was held regarding Intel Foundry Services, where they announced the application of the new business model previously explained (Figure 3). In short, Intel Product Company, which manufactures and sells processors, charges a "fair market price" that appropriately considers manufacturing costs and profits (although this is an internal transaction), and now records profits and losses independently. However, there is currently a significant deficit (Figure 4). This is natural because new factories are being built in Arizona and Ohio, and Intel Foundry Services will be responsible for all of this, so they cannot achieve profitability. However, this is a front-end investment and is inevitable for completing the IDM 2.0 strategy.

In the explanation on April 2, the operating deficit of Intel Foundry Services will peak in 2024 and then gradually decrease; the target for Intel Foundry Services is to return to profitability by the mid-2030s and achieve a gross profit margin of 40% and an operating profit margin of 30% on a non-GAAP basis by the end of 2030; the target for Intel Products is to achieve a gross profit margin of 60% and an operating profit margin of 40% by the end of 2030, calculated on a non-GAAP basis.

Strong sales of AMD client PCs seem to have reached a stable level

AMD's situation is the opposite of Intel's. Table 3 shows sales in the first quarter of 2023, and Table 4 shows operating income. It can be seen that the operating income for each fiscal year is maintained at more than 1 billion US dollars, and regardless of sales, the total operating income is not much different from Intel, or it has already exceeded Intel in the first quarter.

AMD is the driving force behind its revival around 2017, but it seems to have reached a plateau in the client PC area. Figure 6 shows that AMD's CPU segment market share (shipments and value) is 23.6% according to Mercury Research, but its revenue share is 33%. Although the performance is very strong, the client's performance is better. Compared with a year ago, the unit share is only about 20%, and the revenue ratio is only about 15%, which is not that high. In fact, in terms of operating profit, the Client only accounts for about 15% of the Data Center. This is completely the opposite of Intel, and AMD's current challenge is to strengthen this client. The embedded business has declined slightly, partly because the shipment of high-end FPGA products has decreased with the stagnation of 5G development, but for the company itself, the shipment of mid-range products is still strong. The company is currently preparing for the upcoming advanced 5G.Another area worth paying attention to is the gaming sector, which includes not only the consumer-oriented Radeon series GPUs but also the shipment of custom SoCs for gaming consoles like PS5/Xbox. However, the sales volume of these custom SoCs is expected to decline in the second quarter, as previously mentioned in the second quarter. Considering that the bulk shipments of SIE and Microsoft are nearing completion, this is an inevitable situation.

As for data centers, it has been revealed during the conference call that samples of the next-generation EPYC product "Turin" using the Zen 5 core have already begun to be shipped to specific customers. Unlike Intel, Turin does not change the platform, so the introduction threshold is expected to be lower than that of Intel. Taking this into account, AMD's strong position in the server field is not likely to be easily shaken even by 2024. Looking back at Photo06, the Server's Unit Share is 23.6%, but the Revenue Share is 33%, which means that many high-priced high-end products are being sold, while Intel's relatively low-priced products are selling well, thus maintaining this advantage. This is AMD's most important task.

It is also worth mentioning that although the mobile segment has made progress compared to last year (2023), it is still weaker than the desktop segment. Intel announced "Core Ultra" last year, and this year also announced the launch of products based on "Lunar Lake" and "Arrow Lake."

However, in the first quarter of 2024 alone, the sales gap between Intel and AMD narrowed to more than three times, and AMD has already surpassed them in operating profit. By the way, Tables 1 to 4 are all based on a non-GAAP basis, and according to GAAP, AMD's profits will drop to the brink of loss (this is because the amortization of goodwill related to the acquisition of Xilinx and Pensando is taken into account), but according to GAAP, Intel's profits will also decline, so in this sense, there is not much difference between the two companies.

Contract manufacturing business holds the key to Intel's fate

Is it possible for Intel to once again significantly surpass AMD in sales and profits? Intel's foundry will hold the key to all this. As long as AMD outsources production to TSMC, its manufacturing capabilities are limited. However, since there are no other wafer fabs available for outsourcing, it is difficult to significantly increase sales in the future. The basic strategy is to maximize operating profit relative to sales by increasing the share of relatively high-priced products. On the other hand, Intel's advantage lies in the fact that, by having its own Fab, it can sell more products than AMD (which is still the case), but to continue this, the success of Intel Foundry is a necessary condition. If the launch of Intel Foundry in the future is not smooth, Intel can only rely on TSMC to manufacture cutting-edge process chips, and if it really happens, there is no difference in terms compared to AMD, or more precisely, without it, AMD's financial situation will be more robust.

Leave a comment

Submit