The AI Chip wave

What are AI Chips?

The rapid advancement of Artificial Intelligence (AI) in recent years has transformed industries and economies worldwide, delivering limitless applications with  unprecedentedly high speed and precision. As AI becomes increasingly integrated in various sectors, the need for enhanced performance becomes inevitable. But as the development and deployment of AI require immense computational power, demands for more sophisticated hardware solutions are subsequently growing. This changing landscape has given rise to AI Chips - specialized hardware designed for handling the massive volume of data required for AI workloads. Tailored to be more efficient for AI-specific tasks, AI Chips offer superior performance for training and inference of the algorithms:

  • Computation speed: Unlike the traditional hardware with sequential processing (performing calculations in order), AI Chips are packed with parallel processing capabilities, allowing it to process billions of calculations simultaneously. This means AI Chips have higher speed and more computation power than its counterparts, estimated to be 10 times more powerful in ANN applications than traditional hardware [1].
  • Higher bandwidth: McKinsey & Company estimates AI Chips to have 4.5 times more bandwidth than traditional hardware, making it more suited for AI-specific tasks [2]. Higher bandwidth allows for smoother and more consistent transfer of large data between memory and processing units, which is crucial for parallel processing capabilities. 

Undoubtedly, AI Chips play an indispensable role in enabling large-scale and cost-effective implementation of AI innovations across industries. In automotive, AI is expected to drive the sector’s future growth, providing critical applications in elevating driver experience, improving safety, and accelerating the development of autonomous vehicles. Indeed, the automotive AI market is forecasted to grow at a CAGR of 24.1% between 2022 and 2027, with global software providers and carmakers already joining the race [3]. For instance, FPT Automotive, a member of the technology conglomerate - FPT, has rolled out high-performing solutions such as GenAI-integrated IVI systems for personalized driver experience, AI-enabled ADAS for collision avoidance, and Driver Monitoring Systems with AI-enhanced capabilities, including facial recognition, emotion detection, and predictive analysis. These automotive applications require enormous computational power to allow for real-time analysis of data and enable decision-making in split-seconds. Specifically, autonomous vehicle applications involve handling enormous data captured from cameras, LIDAR, radar, and sensors, requiring AI Chips’ parallel processing capabilities to perform multiple tasks in sync such as object extraction, detection, segmentation, and tracking [4]. Without question, AI Chips are becoming highly integrated in the automotive industry, so much so that global chipmakers, such as Intel, are releasing an automotive version of their AI Chips to match this growing demand [5]. 

Similarly, AI is expected to accelerate healthcare innovations, particularly in the field of medical imaging. Renowned for its precision, AI helps improve diagnostics and treatment accuracy by identifying patterns that are often bypassed by the naked eye. For instance, an AI solution by FPT Software assists dental examination and treatment by detecting issues such as missing teeth, dentures, and decayed teeth, as well as identifying implant types and models, helping dentists save time during diagnostics whilst improving accuracy. In addition, AI applications help facilitate personalized healthcare, providing patients with tailored treatment plans and medicine prescriptions instead of the traditional standardized approach. For example, FPT Software’s AI solution detects patients’ muscle and fat areas and helps doctors adjust medicine doses accordingly. Trained with CT images of patients’ abdomen, the AI model analyzes and identifies 2 types of fat area and 4 types of muscle area, and can be optimized to perform on edge devices or CPUs with processing time of less than 3 seconds per slice. As AI innovations for healthcare advance, AI Chips become increasingly important in implementing said solutions. Building effective AI models involves training on large data sets of high-resolution images like MRI, CT, and X-rays, which requires the enormous computational power that only AI Chips can provide. Moreover, AI Chips are critical for AI solutions to perform on edge devices, enabling applications such as real-time monitoring of health indicators captured by wearables and smartphones.  

A challenging landscape

As the adoption of AI applications increases, the demand for AI Chips is expected to continue rising, with revenue forecasted to grow 33% in 2024 [6]. Despite the positive outlook, the industry’s current state remains somewhat uncertain. The AI boom has fueled the race for more computational power, pushing manufacturers to raise production capacities at an exponential rate to meet the skyrocketing demand. The situation is further challenged by the global talent shortage, with estimations revealing that more than 1 million additional skilled workers are needed by 2030 [7].

However, the biggest challenge perhaps is the supply chain disruptions caused by the political scene. With the US imposing sanctions on Chinese tech companies in 2019, the political tension between the US and China has been escalating over the years. Subsequently, a ban on sales of semiconductor chips to China was enforced by the US in 2022, with Japan and the Netherlands following suit by imposing similar restrictions [8]. As a result, the global semiconductor industry is suffering a negative impact, subsequently hindering growth and limiting innovation development to the tech industry as a whole [9]

Vietnam as a rising star

Amidst the challenging landscape, Vietnam is rising as a key player in the semiconductor supply chain. Although a relative newcomer, Vietnam has shown high semiconductor potential with local enterprises already introducing made-in-Vietnam chips. For instance, FPT Semiconductor, a member of the leading technology corporation - FPT, became the first Vietnamese company to design commercial chips and has successfully secured orders from Japan, South Korea, Taiwan, etc. to export 70 million chips. In addition, Vietnam has attracted global chip manufacturers such as Intel, Amkor, and Hanmi Semiconductor. Specifically, the country is home to Intel’s largest factory for chip packaging and testing worldwide, with an investment of $1 billion [10]. More recently, Amkor Technology has increased investment in Vietnam, totaling to $1.6 million to open its most advanced chip packaging factory [11]. With political advantages, human resources, and innovative spirits, Vietnam is proving its potential to solve the global semiconductor challenges. 

The workforce

Developing human resources for the semiconductor industry has been set as a priority by the government, aiming to invest $1 billion to train 50,000 semiconductor engineers by 2030 [12]. The Ministry of  Education and Training expects to have 7,000 students enrolling in chip designing and related majors at universities in the next year, with plans to increase these figures by 20-30% annually [13]. The government has also established a semiconductor and AI center in Da Nang, which will be responsible for research and training for microchips, semiconductors and AI development, as well as fostering technological transfer and international collaboration [14]
Local enterprises are also joining hands. For example, FPT - a pioneering force in semiconductor development in Vietnam, aims to train 10,000 professionals by 2030, equivalent to 20% of the government’s target. The corporation has rolled out its own program, with FPT University - part of the FPT ecosystem and Vietnam’s first private institution, setting up the Department of Semiconductor and Microelectronics to provide in-depth training in IC design and accelerate R&D for the industry. More recently, FPT Corporation has also partnered with Taiwan’s FCC Partners to establish the Vietnam Semiconductor Development Fund, aiming to develop human resources and attract investment for the sector. 

Government policies

Another major advantage lies in the government’s encouraging policies. The Vietnamese government has emphasized the strategic importance of the semiconductor industry on numerous occasions and has begun implementing policies to support its growth. Considered a prioritized product, semiconductors are eligible for attractive incentives offered by the government to boost the development and manufacturing of the industry. This includes exemptions or reductions from land rental for facilities [15], preferential policies for personal income tax and corporate income tax, as well as financial support [16]. For example, semiconductor companies enjoy a preferential corporate income tax rate of 10% in 15 years, compared to the standard rate of 20%. Large-scale and high-impact semiconductor projects are eligible for even lower rates of 9%, 7%, and 5%, with possibilities for tax exemption and reduction [17]

Political stability

Having both a solid domestic political scene and robust foreign relations, Vietnam boasts a prosperous environment for the semiconductor industry. As of 2023, the country has formed diplomatic relations with nearly 200 countries and signed 17 free-trade agreements (FTAs) with nearly 60 economies [18]. More notably, Vietnam has established comprehensive strategic partnerships with the world’s largest economies, including the US, China, South Korea, Japan, Russia, India, and Australia [19].

The country has continuously fostered technological cooperation with major countries in semiconductor development, having attracted 40 companies from Japan, South Korea, the Netherlands, and the US for enhancing research, development, and manufacturing for the sector [20]. In addition, there has been an increase in collaboration on the state level. The recent upgrade of the Vietnam - United States relationship to a comprehensive strategic partnership further enhances its potential for collaboration and integration into the global semiconductor supply chain [21]. Under the newly upgraded collaboration agreement, the two countries announced the establishment of semiconductor workforce development initiatives, with an initial investment of $2 million from the US government [22]. More recently, to commemorate the first anniversary of this relationship upgrade, the two countries have launched the ITSI fund under US Chip and Science Act, in which Vietnam is one of eight selected strategic partners [23]. The fund aims to train 120 instructors & 4,000 students by 2025 and support universities in building chip packaging and testing programs following US standards [24]

The AI nation

Perhaps one of Vietnam’s most powerful motivations stems from its own ambition to become an AI nation. Targeting to be among the top 50 AI countries globally by 2030, the country has introduced a national strategy on R&D and applications of AI, pulling efforts from the government, communities, and local businesses. With the AI movement flourishing across the country, local and international enterprises are setting up dedicated AI centers for enhancing research and development. Most notably, FPT Corporation invested $173 million to build an AI center in Binh Dinh for research, training, & development of AI solutions. Partnerships have also been fostered to strengthen the national AI capabilities. For example, local companies such as FPT are collaborating with leading AI organizations including NVIDIA, Landing AI, and Mila to advance AI solutions and best practices development. Specifically, FPT partners with NVIDIA to build a $200-million AI factory in Vietnam to develop AI and Cloud solutions using NVIDIA’s latest technologies. 

"The most attractive destination for US semiconductor investors"

The President and CEO of the US Semiconductor Industry Association praised Vietnam as the most attractive destination for US semiconductor investors [25], confirming the country’s enormous potential. Indeed, Statista forecasts the revenue for Vietnam’s semiconductor market to steadily rise over the years, reaching as high as $31.39 billion by 2029 [26]. While there are still several underlying challenges, largely due to the country being in its early stages, the government and local organizations, such as FPT, have been proactively making engagement in the industry. While the progress made so far in education, training, research & development establishes a strong foundation, further strategic efforts to foster the next generation of chips - AI Chips, will solidify Vietnam's position as a key player in the global semiconductor landscape. By continuing to invest in education and infrastructure, and emphasizing the participation of local enterprises, Vietnam’s economic and technological outlook is full of promise.

 
Author Nguyen Vu Quynh Trang