hankyoreh

US President Donald Trump announces the “Stargate Project” from the White House flanked by Larry Ellison, Oracle founder and executive chairman; Masayoshi Son, the CEO of SoftBank Group; and Sam Altman, the CEO of OpenAI, on Jan 21. 2025. (UPI/Yonhap)

On Jan. 22, US President Donald Trump announced the Stargate Project, a massive initiative expected to cost US$500 billion.

The project represents the US’ ambition of taking the lead not only in artificial intelligence but also, eventually, in artificial general intelligence (AGI), which could reason and solve problems like a human being.

The main corporate partners in the project are OpenAI, Microsoft, Oracle and SoftBank. It’s hardly surprising that the project would include OpenAI, the developer of generative AI chatbot ChatGPT, but participation by Japanese IT giant SoftBank and cloud solutions firm Oracle is more notable.

Security concerns undergird the US’ strategy in the project, which apparently inherits the philosophy of economic security behind the CHIPS and Science Act (enacted under Trump’s predecessor, Joe Biden) while strategically extending that to the domain of AI.

A corollary is that only parties that can make a commensurate contribution in technological, financial and industrial competitiveness can be recognized as security partners in the US-led AI initiative.

For example, SoftBank CEO Masayoshi Son will be chairman of the Stargate Project as well as its financier, implying that SoftBank will be a key player in the US’ effort to take the lead in AI research. It appears that Tokyo’s efforts to accommodate Trump during his first term in office are paying dividends.

Japan has consistently pursued a policy of supporting its semiconductor industry. In particular, the Rapidus Corporation has been strategically cooperating with the US on the goal of producing 2-nanometer chips, which are chiefly in demand as AI processors.

Japanese Prime Minister Shigeru Ishiba, the first Asian leader to visit the US and hold a summit with Trump in January, has promised to increase economic cooperation with the US in AI and other areas. SoftBank and Stargate, IBM and Rapidus, and Micron’s plan to build a chip plant in Japan — they all point to Japan being a partner in the US’ strategy of reshoring semiconductors and AI.
 
US-led AI: No contribution, no participation

Another major partner in the project is Oracle. Its mission goes beyond the AI servers and cloud services it has provided thus far. Surprisingly enough, the primary role Oracle is supposed to play is advanced biotech, which includes developing personalized cancer vaccines using AI.

Oracle proposes analyzing and sequencing individual genetic information on an AI platform provided by OpenAI and SoftBank and then fabricating mRNA-based cancer vaccines within 48 hours through an automated robotic assembly line.

Just a few years ago, that would have sounded like a sci-fi pipe dream, but now such efforts are nearing fruition. The hurdles are no longer technical, but financial. Nor should we forget that robotics, automation, AI, biotech, and big data have played key roles in making all that possible.

The AI hegemony as envisioned by the second Trump administration is one where US partners that share security policies, have advanced AI technology capable of learning and inference, can participate in an AI-specific semiconductor supply chain, can pool sufficient capital and have sufficient competitiveness in other industries such as biotechnology, robotics and manufacturing, become key players.
 
The Stargate Project may be light-years ahead when it comes to technology and capital, but we do have to consider various bottlenecks, in which a single component limits the capacity or performance of an entire system.
 
As more computing resources are needed to build powerful AI systems, AI-focused semiconductors and more efficient AI models become a priority. For example, Chinese AI startup DeekSeek’s large language model, which shocked the world upon its unveiling on Jan. 20, foreshadows how the price/performance ratio will become a competitive factor in the AI industry. Not only will companies toil to develop models that can learn and infer faster, but the cheaper and power-efficient AI computing technologies that specialize in such models will also become a new bottleneck.

US President Donald Trump announces a US$100 billion investment by TSMC with the company’s CEO CC Wei on March 3, 2025. (Yonhap)

Nurturing the seed of revolution: Partisanship and competition
 
This suggests that the competition over AI-specific GPUs, which has been dominated by Nvidia, will diversify to include neural processing units (NPUs) and tensor processing units (TPUs) as well as traditional CPUs and other parallel computing innovations. Rivalry around high-bandwidth memory (HBM), which Hynix has dominated for the most part, may, for a more efficient AI computing environment, diversify into general-purpose memory from LPDDR to GDDR, or other forms of HBM specialized for other AI hardware. Diversifying bottlenecks also means more competition, which could spawn more innovation from new players, especially outside the US.
 
In this tumultuous environment, the US is likely to keep its competitors, most noticeably China, at bay for the foreseeable future by focusing on its technology and capital while applying its security values to other industries. If the Stargate Project begins to devour existing private investment in AI like a black hole, relevant companies will be divided based on their participation in the project. Intel may be forced to accept the decision to spin off its foundry (IFS), and if the Stargate consortium takes over the Intel foundry, it could force an executive order that AI semiconductors must be made in foundries within the US, not by companies overseas such as TSMC.

The problem is South Korea. In the political chaos that erupted at the end of last year, the country is being run by an acting administration. We’re falling further behind in an intensifying cycle of innovation and competition involving AI and the changes in the industrial landscape AI will bring. Compared to Japan, where the state and private industry have cooperated to quickly place themselves in various positions across the board in preparation for what's ahead, South Korea’s strategy for surviving the AI industrial revolution is nonexistent, a truly unfortunate situation. 

SK Hynix has partnered with Nvidia to generate immense profits in the HBM sector. But if Nvidia’s monopoly falters while the AI computing landscape diversifies amid the rise of DeepSeek, the current structure for profit stability will face a crisis much sooner than expected. 

The level of Samsung Electronics’ involvement in the Nvidia supply network remains opaque, but the volatility of the AI industry could present new opportunities. But if they continue to lose competitiveness in the foundry industry without any clear vision for the AI industry, they could let such an opportunity just pass by. 

Domestic AI hardware startups are also in a state of disorder. They’ve typically opted for developing sovereign AI with objectives tied to NPU or TPU while eyeing the domestic industry as its main market, but they are far behind trends that combine AI with industry, such as physical AI or cutting-edge biotech, which are likely to be accelerated by the Stargate Project.  

We need innovation that leverages our base as a global manufacturer

For South Korea to emerge from the AI revolution as a strong player, we need to innovate in the traditional manufacturing industry, an area that the US cannot manage on its own. South Korea has a strong industrial base in machinery, shipbuilding, energy, petrochemicals and nuclear energy — sectors that are important from a national security perspective. 

While we face extreme competition in manufacturing from China, which has a massive domestic market, the domestic manufacturing sector must be protected for national security purposes. But if we fail to develop the necessary innovations using AI at the right time, we will be forced to relinquish our advantage to countries or firms that innovate before us. If South Korea has any opportunity in the current situation, it lies with manufacturing. 

When considering the volatility of AI trends, the four years of the second Trump administration will exhibit a high degree of intensity that will be qualitatively different from what we saw during the first Trump administration. 

The South Korean government and Korean companies must accurately ascertain the technological and economic context of the Stargate Project and proactively draft technological and investment strategies that will allow us to emerge as a key player in that context. Amid the threats presented by a second and third iteration of China’s DeepSeek and the cutthroat price competition presented by China's manufacturing industry, South Korea needs to reformulate its strategies for industrial conversion with a mid to long-term vision. In particular, we need to quickly define the key connections that AI can have with heavy chemical industry, energy, bio, and shipbuilding and concentrate our efforts to accelerate government-private cooperation on relevant R&D. 

The next South Korean administration needs to form partnerships in the AI industry, which is currently led by both the US government and American firms, and draft policies with a particular focus on AI computing value chains and transitions in conventional industry. We also need to prioritize initiatives that protect our national interests. We need to secure sufficient funding and develop the technological prowess to establish ourselves in irreplaceable positions across the board. We need to get a move on right now. 

By Kwon Seok-joon, professor of chemical engineering and semiconductor convergence engineering at Sungkyunkwan University

Please direct questions or comments to [english@hani.co.kr]