When Intel announced the acquisition of Altera, the second largest FPGA company, for US$16.7 billion in 2015, Intel's management at the time may not have thought that this part of the business would become independent again (note: since this business does not have a new name, let's call it Altera for the time being) and re-list it. Just like Brian Matthew Krzanich, then CEO of Intel, he should not have expected that he would be forced to resign due to some incidents two years later.

However, if you look back at Pat Gelsinger's series of operations during the more than two years since he became Intel - selling the storage business, selling the NUC business, betting on wafer manufacturing at five nodes in four years, and independently listing Mobileye. Intel's independent PSG (Programmable Solutions Group) business is also a matter of time.

However, after Intel recently announced that it would split its PSG business into an independent company to give it more autonomy and flexibility, senior domestic FPGA practitioners still expressed a little surprise. In his opinion, this team with the former second-in-command in FPGA as its team will bring new variables to the FPGA market.

Data center, a beautiful dream

Since its invention in the 1980s, FPGA has a history of forty years. As its scale increases, the process of product iteration also introduces more technological innovations. In addition, the application of FPGA has also entered thousands of industries along with changes in terminal needs.

According to statistics from ACM in 2020, the communications industry has been the largest user of FPGAs in the past decade or so. This is because telecommunications standards are constantly changing, making it very difficult to build telecommunications equipment, and companies that first provide telecommunications solutions tend to occupy the largest market share. This allows FPGAs with huge flexibility to find a "road to wealth." Xilinx and Altera are undoubtedly the biggest winners, and they also have a large say in price.

Interspersed with a story, many years ago, AT&T and Lucent also manufactured their own FPGAs called ORCA (optimized reconfigurable cell arrays), but they were unable to compete with Xilinx or Altera in terms of speed or chip size, and eventually gave up their self-research plans. From this, we can also see the hard power of giants such as Xilinx and Altera, which have survived fierce competition in the past two decades.

Although it can make all the difference in the communications market, FPGA manufacturers do not want to be limited to this, and they have begun to explore more opportunities. For example, FPGA vendors entered the oil and gas field in 2007. At the time, conventional computer simulations of drilling holes in the Earth to find oil would take longer than actually building the drilling sites and drilling the wells themselves. The use of FPGA accelerators greatly changes this reverse timing.

As for the embedded and DSP markets, they are also target markets that FPGA is interested in. However, due to price reasons, FPGA has not been able to succeed in the above-mentioned markets. As a result, FPGA manufacturers turned their attention to markets such as HPC, data centers, and automobiles, which led to Intel's acquisition of Altera and later AMD's acquisition of Xilinx.

According to Intel's acquisition press release at the time, the acquisition will combine Intel's leading products and manufacturing processes with Altera's leading field programmable gate array (FPGA) technology. The merger is also expected to launch new products that meet the needs of customers in the data center and Internet of Things (IoT) market segments. Intel plans to offer Altera FPGA products using Intel Xeon processors as highly customized integrated products. The two companies also hope to enhance Altera's products with design and manufacturing improvements brought by Intel's integrated device manufacturing model.

Brian Krzanich, Intel CEO at the time, said that through this acquisition, we will leverage the power of Moore's Law to make the next generation of solutions not only better, but also able to do more. Whether enabling new growth in networking, large-scale cloud data centers or the Internet of Things, this is what our customers expect.

In contrast, AMD's acquisition of Xilinx also emphasized that the acquisition of Xilinx will allow AMD to bring together highly complementary products, customers and markets, as well as differentiated IP and world-class talents, to create the industry's leader in high-performance and adaptive computing. Xilinx's industry-leading FPGAs, adaptive SoCs, AI inference engines and software expertise enables AMD to deliver the industry's strongest portfolio of high-performance and adaptive computing solutions, positioning us to capture a greater share of the approximately $135 billion market opportunity including cloud, edge and smart devices.

In fact, looking at the product lines, there are still quite a few differences between Altera and Xilinx. Before the acquisition, the latter had already transformed from a pure FPGA manufacturer into an SoC supplier through the launch of Versal and other types of products. The company also has a variety of other IPs, which gave AMD more room to integrate after acquiring Xilinx.

Of course, for the acquisition of Altera, Intel still responded with heavy responsibility at the beginning, but subsequent development seemed to have deviated from their track.

The seven-year itch, start again

In fact, the story of Intel and Altera did not start in 2015.

As early as 2013, Altera and Intel announced that the two companies had reached an agreement on using Intel's 14nm tri-gate transistor technology to manufacture Altera FPGAs in the future (Altera's foundry partner at the time was TSMC). Targeting ultra-high performance systems for military, wireline communications, cloud networking, and computing and storage applications, these next-generation products will enable breakthrough levels of performance and power efficiency not previously possible.

According to an article by well-known blogger IamCutress, after Altera was acquired, Intel switched all of the latter's roadmaps from TSMC manufacturing to Intel's 10nm process manufacturing. But what is not as good as God is that due to the significant delay in the 10nm process, almost all of Altera's products in the next few years have been delayed, which has caused some of the company's customers to turn to competitors. More importantly, since being acquired by Intel and making PSG part of the Data Center/Artificial Intelligence Group (DCAI), they have shifted their focus from the key traditional FPGA vertical to the DCAI part of the business, which has undoubtedly slowed down the progress of their FPGAs in the traditional market.

Steven Leibson, the author of EEjournal, also stated in his article that the marriage between Intel and Altera did not go smoothly in terms of technology at the beginning. He noted that Altera/PSG initially had difficulty tapering out the Arria 10 FPGA in Intel's 20-nanometer FinFET node, but the next-generation Stratix 10 FPGA performed better at the 14-nanometer node.

At the same time, the two parties also explored Intel's early heterogeneous Chiplet technologies and strategies, including the use of EMIB packaging (embedded multi-die interconnect bridge) and AIB (Advanced Interface Bus) bus technology. Intel PSG then jumped to the successful Agilex FPGA family, which was originally manufactured on Intel's 10nm process technology. Recently, Intel PSG has been expanding its Agilex FPGA family portfolio and appears to be planning to move the Agilex FPGA architecture from the low-end Agilex3 FPGA to the DirectRFAgilex9 FPGA above.


Figure: Intel’s programmable product line

Although it has an impact on Intel's PSG business for one reason or another. But according to statistics from IamCutress. As of Intel's first quarter of 2022, this business will bring Intel an average of about US$500 million in revenue every quarter. The performance has been even better heading into recent quarters, with revenue hitting a record high of over $700 million. This means that PSG has generated nearly $15 billion in revenue for Intel since the acquisition.


Figure: IntelPSG’s revenue data for past several systems

For comparison, according to statistics from statista, Altera's largest competitor Xilinx's revenue in the past few years from 2019 to 2022 has almost been more than 600 million US dollars, and even exceeded one billion US dollars in revenue in Q3 2022.


Figure: Xilinx quarterly revenue data over the past few years

In fact, from a numerical perspective, PSG's performance is not as bad as expected. However, practitioners in the FPGA industry told the author that before the acquisition, Altera and Xilinx competed fiercely in the communications, industrial and military markets. However, after being acquired by Intel, Altera has been losing ground in the domestic market, especially in the communications market.

"We originally thought that Lattice would be the biggest victim after the acquisition, but we didn't expect that it would be Altera." He said.

However, as Intel announced that it will re-independent the PSG business and plans to go public, the story of this FPGA may be rewritten.

According to Intel in the press release, operating PSG as an independent business will give the PSG team the autonomy and flexibility needed to comprehensively accelerate its development and compete more effectively in the FPGA industry, and serve a wide range of markets including data centers, communications, industrial, automotive and aerospace.

PSG Chief Operating Officer Shannon Poulin also pointed out that this is an important turning point for PSG's business - it will help us build a leading end-to-end FPGA product portfolio, while strengthening product launch strategies and increasing market share. The investment the new company will attract and the autonomy it will have to execute are great news for our customers, partners and the entire FPGA industry.

In the author's opinion, the reason why Intel chose independent PSG was first of all because of its huge market potential. As they quoted third-party data, the FPGA market will grow at a compound annual growth rate (CAGR) of more than 9%, from US$8 billion in revenue in 2023 to US$11.5 billion in 2027.

In addition, Intel's vigorous development of IFS business also requires huge financial support. The success of Mobileye prompted them to make this decision. The team at PSG is also confident about the future of FPGA, but can they really get what they want?

Fight again and return to the top?

According to the FPGA practitioners mentioned above, before the acquisition, the FPGA market was mainly controlled by Xilinx, Altera, Lattcie and Actel, and they collectively captured nearly 90% of the market share. Among them, Xilinx and Altera mainly serve high-demand and high-margin markets such as communication computer servers, industrial security vehicles, medical, military and aerospace. Although the third largest Lattice is involved, most of them are mainly small devices for interfaces. They are also involved in part of the consumer electronics market. As for Actel (first acquired by Microsemi and then Microchip), it mainly serves the military and aerospace markets, and involves a small amount of industry.

From a technical point of view, Xilinx belongs to a technical school, has always been technologically leading, and is also excellent commercially, a bit like Qualcomm in the field of mobile phone chips. Altera is a bit like the Wall Street faction, more flexible in business, and has been following for many years, but has been unable to surpass the number one Xilinx. Both manufacturers are also good at various algorithms.

But now, after the acquisition of the eldest, second and fourth eldest brother, the transformation of Lattice, and the rise of Chinese FPGA manufacturers. Although Altera has regained its independence, the world is no longer what it once was.

In the above article, we talked about Altera's steady retreat, while Xilinx became more and more courageous as it fought. As for Lattice, it turns out that they rely on non-volatile FPGAs below 10K (small-capacity XO2-XO3 series products) to contribute more than 80% of the company's revenue. But now they have almost all of these products replaced by domestic ones. However, they have launched ECP5 and new 28nm products, and have replaced many Altera and Xilinx in the mid-to-low-end FPGA market. This can be said to be an example of business war in the chip market.

There is no doubt that Altera no longer has the influence it had at that time, so when asked if they could return to the top, the FPGA practitioner said that this requires their team to start from multiple aspects and do very well in order to have a chance:

First of all, the ability and experience of the leadership team are very important. Although Intel has appointed a new leadership team, it will take time to verify whether they can effectively lead and manage the PSG department. Establishing a new management team and marketing and sales team also requires time and resources, and involves certain risks.

Secondly, market changes are also a factor that cannot be ignored. In the past eight years, the competitive landscape of the FPGA market has undergone great changes, and Altera's brand influence and market share have also undergone certain changes. To return to the top, Altera needs to have a deep understanding of market demand and competitive landscape, and formulate a development strategy that is in line with market trends.

In addition, team reorganization and product innovation are also key. To get back to the top, Altera needs to attract lost talent and reorganize its R&D team to launch competitive new products. This requires a large investment of resources and time, and involves certain risks.

Finally, the stability and competitiveness of the supply chain are also very important. Although Intel emphasized its continued cooperation with IFS in its announcement, for mid-to-large-scale customers, Altera may also need to establish good cooperative relationships with suppliers such as TSMC to ensure product quality and stability and obtain competitive prices.

In short, getting back to the top requires hard work and time accumulation. Although there are many uncertainties and risks, if Altera can do an excellent job in leadership team, market changes, team reorganization, product innovation and supply chain, the possibility of returning to the top is still very high.

For domestic FPGA manufacturers, a new FPGA situation is also opening up. Let’s see how they strategize and climb to the top.

access:

Jingdong Mall