Image source @ Visual China
Text | Value Institute
Officially completing the acquisition of Finnish semiconductor company Siru Innovations, Intel has added another heavy weight on the way to catching up with NVIDIA.
From the perspective of technology, Siru Innovations has a deep accumulation in the field of computer graphics, covering advanced APIs to low-level GPU architecture solutions; from the perspective of business and market share, Siru Innovations' contacts and customer resources accumulated over the years have all been pocketed by Intel, which also makes the latter more powerful.
Of course, in the view of the Institute of Value, the acquisition of Siru Innovations is only one step in a series of recent big plans of Intel, and the ambitions of the top management for the fat meat of the GPU have long been expanding. The reason why Intel wants to die in the GPU market is, in the final analysis, to catch up with the trend of the times - the high-performance computing technology represented by the GPU is interdependent with cutting-edge technologies such as AI and 5G.
It is undeniable that NVIDIA is still the absolute leader of the GPU industry, regardless of the level of technology research and development, production capacity and market share is far ahead. Especially in the discrete graphics market, there are a wider range of application scenarios, closer to cutting-edge technologies and higher profit margins, which is also a treasure that Intel has been looking for.
However, with the official launch of the Intel Arc series, NVIDIA's sense of crisis has risen, and the direct confrontation between several giants has begun to escalate.
There is a more and more fierce talent war outside the field, the field is a continuous inward roll of technology, looking at the cornucopia of GPUs, semiconductor giants have long been unable to hold back the desire and impulse in their hearts.
With the acquisition of Siru Innovations, Intel continues to increase its GPU business
On Tuesday, local time in the United States, Intel announced on the official Twitter that it has officially completed the acquisition of Finnish semiconductor company Siru Innovations, which will be merged into Intel's graphics business unit.
To make Intel so excited, Siru Innovations naturally has its own superiority.
Founded in 2011, Siru Innovations is a low-power graphics IP developer and mobile graphics service provider focused on markets such as graphics IP, software-defined radios, and industrial touchscreen user interfaces. According to the official introduction, Siru Innovations has a lot of achievements in the field of graphics IP-related technologies, and technology is also one of its biggest selling points.
From the perspective of technology research and development, Siru Innovations does have a deep accumulation in the field of computer graphics: covering high-level APIs to low-level GPU architecture solutions, and also has rich experience in low-power, high-performance SoC graphics IP development.
Of particular interest is the Siru Innovations team's contribution to the development of intel Ponte Vecchio processors.
At an investor conference earlier this year, Intel unveiled for the first time the performance parameters of the Ponte Vecchio, a self-developed supercomputer card. According to Intel's official statement, the performance of the product is at least 160% ahead of Nvidia's ace product 7nm Ampereca King A 100, which has been dubbed "chip monster" by many media.
According to The Register, Siru Innovations' R&D team was a pioneer in image development in the 1990s, with a large number of R&D talents who had worked at giants such as AMD, Qualcomm and BitBoys, and the technology reserves and Intel were highly adapted.
From the perspective of business and market share, Siru Innovations' contacts and customer resources accumulated over the years have all been pocketed by Intel, which has also made the latter more powerful.
For a long time in the past, Qualcomm and AMD have established stable cooperative relations with Siru Innovations, reflecting the latter's technical advantages. In Intel's view, Siru Innovations' experience with Qualcomm and AMD and its understanding of the market will also help it further develop the lucrative PC and mobile graphics markets.
In the face of this marriage like a match made in heaven, it is no wonder that Intel can hardly hide its joy in the announcement:
"This talented team brings decades of experience in developing graphics IP and software services that will help support our customers in MaaS, ADAS, gaming, and more, and welcome to join!"
Of course, in the view of the Institute of Value, the acquisition of Siru Innovations is only one step in a series of recent big plans of Intel, and the ambitions of the top management for the fat meat of the GPU have long been expanding.
The reason why Intel wants to die in the GPU market is, in the final analysis, to catch up with the trend of the times - the high-performance computing technology represented by the GPU and cutting-edge technologies such as AI and 5G are interdependent. In multiple application scenarios such as autonomous driving, AIoT, and smart terminals, GPUs are indispensable members, and almost all terminals in the future will need more powerful computing power.
Having missed the transition period of mobile Internet and lost its dominance in the chip industry, Intel does not want to fall again. To win the future, betting on GPUs from now on is the only option. And in the past few years, Intel has also made many efforts.
On the one hand, Intel is constantly enriching its product line, from PCs to mobile terminals, for a comprehensive and in-depth layout.
As of March this year, Intel's Arc series of discrete graphics for PC has three major series of products (Arc3, Arc5 and Arc7 series), targeting different application scenarios such as mainstream games, performance games, and catering to games. In addition, the client graphics card Battlemage, Celestial and Drid series, which have been researched and developed at great expense, have also been fully introduced to the market, and Intel has initially built its own GPU product matrix.
Among them, the high-performance PC market mentioned above is the most lucrative, and it is also intel's most eye-catching - this is Siru Innovations' biggest temptation to Intel.
At present, Intel is accelerating the development of high-performance PC processors, and whether it can overcome the technical difficulties related to GPU and accelerators is the key to success or failure. After purchasing Siru Innovations, Intel was more emboldened to endorse Nvidia's Hopper and AMD's CDNA2 enterprise-oriented architecture.
On the other hand, in addition to the acquisition of Siru Innovations, Intel has also integrated internal resources by adjusting the internal organizational structure, strategic adjustment and other ways to bring the GPU to a more important position.
Also at the February Investor Conference, Intel CEO Pat Kissinger announced Intel's latest organizational restructuring, bringing together six business lines: Data Center and Artificial Intelligence (DCAI), Client Computing (CCG), Accelerated Computing Systems and Graphics (AXG), Networking and Edge (NEX), Intel Foundry (IFS), and Mobileye.
Of the six divisions, AXG, IFS, and Mobileye are Intel's next phase of key businesses, and Kissinger expects at least general revenue growth by 2026 to come from these three business lines and achieve revenue growth rates of at least 10%-12% per year.
Recalling the architecture day two years ago, Xe GPU debuted, after 22 years of re-powering the GPU market Intel is still just a humble opponent in front of Nvidia. But just two years later, the situation has been completely turned upside down.
Now, perhaps it's Nvidia's turn to respond.
GPU robbery war started, the former king of Nvidia is not lacking in hidden worries?
It is undeniable that NVIDIA is still the absolute leader of the GPU industry, regardless of the level of technology research and development, production capacity and market share is far ahead.
As we all know, GPUs are currently mainly divided into two tracks: independent GPUs and integrated GPUs, the former is mainly used in high-performance PCs, VR/AR headset devices, and AI servers, and the application scenarios of the latter are concentrated on mobile hardware devices such as smartphones.
According to data released by JPR, as of the fourth quarter of last year, global GPU shipments were 101 million blocks, an increase of 0.8% over the third quarter. Among them, Intel's market share reached 62%, although it fell slightly by 0.1% from the third quarter, and its shipment volume still dominated. AMD's market share for the quarter was about 19%, a slight decline of 0.7% year-on-year, and Nvidia's was down 0.6% to 19% year-on-year.
(Image courtesy of JPR)
However, discrete graphics application scenarios are broader, closer to cutting-edge technology and have higher profit margins, which is a treasure that Intel has been looking for. In the track of independent graphics cards, NVIDIA's overall strength is very strong, and its domination is unshakable.
Data from JPR also shows that Nvidia and AMD have a market share of 81% and 19% in the discrete graphics card market, respectively, basically forming a monopoly by the strength of the two.
In the past two years, AMD has also made many efforts to steal more market share from NVIDIA, but unfortunately it has not been satisfactory. It will be found that since 2020, NVIDIA's market share has remained above 80%, and although it has been affected by the supply chain crisis and the epidemic lockdown policy, and the production capacity has been significantly reduced, AMD has not been able to leverage its customer resources.
But with the official launch of the Intel Arc series, NVIDIA's sense of crisis has risen, and the direct confrontation between several giants has begun to escalate.
The most direct manifestation is the high-frequency phenomenon of mutual poaching of executives, which has intensified the acquisition war.
When the dust settled on the acquisition of Intel and Siru Innovations, many media outlets emphasized one point in the report: the work experience of the latter's co-founder Mika Tuomi in semiconductor giants such as Qualcomm and AMD.
It is reported that Mika Tuomi once created BitBoys, a GPU company known as "Finnish AMD", which was acquired by ATI, Mika Tuomi successively entered semiconductor giants such as ATI, Qualcomm and AMD, and is also one of the founders of Qualcomm Adreno GPU product lines. In addition, another founder, Mikko Alho, was also the head of Qualcomm Finland, coordinating the hardware planning and resource allocation of Qualcomm GPU projects.
Of course, the turnover of Nvidia, Qualcomm, AMD and Intel giants in the past two years is nothing new, and the two founders of Siru Innovations are just one of them.
Going further afield, when the Xe GPU was first developed, Intel hired Raja Koduri, the head of amd GTG and chief architect, at a high salary. Raja Koduri has not only played a key role in Intel's expansion into the discrete graphics market, but now also plays a key role in the acquisition of Siru Innovations.
Closer together, earlier this year, after the acquisition of ARM was completely bankrupt, Nvidia worked non-stop to build a CPU design and engineering team in Israel, recruiting hundreds of hardware, software and architecture engineers. One of the most important "reinforcements" is undoubtedly the design manager of Intel, Rafi Marom, who served as the main person in charge of architecture CPU research and development.
Even AMD, which is relatively calm and restrained, and Qualcomm, who is not in the center of the battlefield, have no shortage of drama in the off-site robbery war.
After Raja Koduri was snatched away by Intel, AMD poached Sean Pelletier, a senior technology marketing manager from Nvidia; Qualcomm Technology, a subsidiary of Qualcomm, acquired NUVIA, a startup company that was former Apple SoC chief architect and ARM senior engineer Gerard Williams, at a high price early last year, along with NUVIA's senior R&D personnel with apple and Intel work experience.
Perhaps you can already see that several giants have already blurred the original boundaries and cut into each other's hinterland in the process of continuously digging and expanding: Intel wants to go all out to conquer the independent graphics card market, Nvidia has begun to make efforts to do CPU, and AMD and Qualcomm have gone deeper and deeper on the GPU track.
The Value Institute believes that with the Intel Xe series as the representative, more semiconductor giants have launched an impact on the independent GPU market, and Nvidia's opponents are not only more and more, but also more and more difficult.
The application scenario is expected to be further expanded, how much potential is the GPU market?
IDC's report shows that 2018-2024 is a period of rapid development of the GPU market, during which the average annual compound growth rate is expected to exceed 30%, and the overall market size is expected to reach the peak of 200.85 billion US dollars around 2027, which is full of potential.
With powerful parallel computing capabilities, GPUs are now widely used in PCs, servers, game consoles, cars, and mobile hardware. But in the view of the Value Institute, this is not the end and boundary of GPUs: with the continuous development of cutting-edge technologies such as 5G, artificial intelligence, and autonomous driving, there is still a lot of room for imagination in gpu application scenarios.
In fact, as early as 2015, Marc Hamilton, then vice president of engineering architecture for Nvidia solutions, said that the application areas of GPUs are "beyond imagination":
"High-performance computing, data centers, the Internet, cloud computing, big data, deep learning... As long as it is an area that requires strong computing power and acceleration power, there is no shortage of GPUs. ”
In the past two years, the status of GPUs in the field of data centers and cloud computing has been rising, and Raja Koduri, which Intel has dug up heavily, is a master of this, which also reflects Intel's ambitions for these cutting-edge technology fields.
Not to mention the fields of AI and autonomous driving that have long been inseparable from GPUs. Taking China as an example, according to the IDC report, the share of GPU servers in the core AI technology facilities in data centers will reach 91.9% in 2021, and the overall market size is expected to reach $6.4 billion in 2024.
(Image courtesy of Tianfeng Securities)
Looking at the financial reports of Nvidia and AMD, which dominate the exclusive market, it can also be found that compared with the past period when games and cryptocurrency mining machines occupied most of the country, their revenue structure has quietly changed. The most pronounced increase in the proportion is in data centers, professional visualization and automotive-related businesses.
According to the fourth quarter earnings report, NVIDIA's data center business revenue reached a record $3.26 billion, in line with Wall Street expectations, up 71% year-on-year and 11% sequentially, respectively. Strong demand from customers in industries such as hyperscale data centers, public clouds, enterprise core clouds, and enterprise edge clouds has driven the continuous growth of NVIDIA's data center business.
Conversely, due to cryptocurrency price volatility and tightening regulations, the gaming business, which has been Nvidia's number one cash cow, has performed mediocrely. The game business recorded $3.42 billion in the fourth quarter, falling below 45% of total revenue, and the year-on-year growth rate of 37% was lower than 42% in the third quarter.
Taking NVIDIA as a reference, it is not difficult to speculate that data centers and cloud computing will play an important role in the future of GPUs. Seizing these heights as soon as possible has become the primary task of AMD, Intel, and Qualcomm sniping at Nvidia.
For now, AMD is half ahead of Nvidia's other catchers, winning the trust of customers such as Microsoft, Lenovo, and Dell, and officially entering the metaverse at the end of last year.
According to AMD's plan, its goal is to improve the energy efficiency of data centers by 30 times over the next five years. The GPU Instinct MI200 accelerator, launched last year when it took orders for Meta, has nearly 5 times more performance than the previous generation.
With AMD struggling to catch up behind, Presumably NVIDIA will also fully raise its vigilance. But on the bright side, the crazy inner volume between these giants may promote the development of GPU technology to some extent and further broaden the application scenarios.
This is not necessarily a bad thing for the development of the entire industry.
Write at the end
When almost everyone focuses on international giants such as Nvidia, Intel, AMD, qualcomm, etc., the Value Institute still wants to remind everyone that the domestic GPU that is driving on the fast track of development also deserves more attention and expectations.
Soochow Securities' research report pointed out that the scale of the GPU domestic alternative market exceeds 5 billion US dollars, and it has broad application prospects in the game, industry, advanced manufacturing, biomedical and other industries.
There is no doubt that the development of domestic GPUs lags a lot behind european and American giants. After all, it was not until 2014 that Jingjia micro developed the first domestic high-performance GPU chip JM 5400 in the true sense.
However, in the following years, the research and development process of domestic GPU chips has been accelerating, and Jingjiawei has also shouldered the heavy responsibility of breaking the monopoly of foreign giants. In the past two years, with the launch of the "Fenghua No. 1" sword refers to the high-performance desktop/server-level GPU market, Moore Thread and Biling Technology have also launched full-featured domestic GPUs, and Nvidia has more and more enemies around it.
At this stage, the above-mentioned domestic GPU manufacturers naturally do not have the strength to break the wrist with giants such as NVIDIA, Inter- and AMD. But as mentioned earlier, the GPU industry has been changing, and Intel and AMD have not given up catching up with Nvidia, constantly improving technology and adjusting strategies.
In this era when technology is king, full of opportunities and challenges, the key is only to see who can seize this opportunity to ascend.
![Real Electric Tiger? Nvidia's new gaming graphics card power consumption or over 900 watts[fig]](data:image/gif;base64,R0lGODlhAQABAIAAAP///wAAACwAAAAAAQABAAACAkQBADs=)