VOA Mandarin: What is Stargate? Is China catching up in AI?

The first investment highlight announced by US President Donald Trump after his return to the White House was in the fiercely competitive field of artificial intelligence (AI) between the United States and China. The "Stargate" project, jointly built by multiple companies and costing hundreds of billions of dollars, will focus on data center construction with the goal of establishing the United States as a "computing empire." Experts say that the significant improvement in the United States' intelligent computing power will widen its advantage over China in AI technology. But there are also signs that China is catching up with American companies in key AI technology indicators through open source technology.

What is Stargate?

At a press conference held at the White House on January 21, Trump announced that Japan's SoftBank Group, OpenAI, the developer of the generative artificial intelligence (AI) conversation tool ChatGPT, and US IT giant Oracle plan to invest up to $500 billion in the next four years to build AI-related infrastructure and create 100,000 jobs.

This grand blueprint is named "Stargate", named after the 1990s science fiction movie of the same name. The initial investment of Stargate is expected to be US$100 billion. SoftBank, OpenAI, Oracle and UAE investment company MGX are the initial equity investors of Stargate. SoftBank is responsible for finance, and the group's chairman and president Masayoshi Son will serve as chairman, and OpenAI will be responsible for operations.

Large-scale data center layout will be the core of the Stargate project. In terms of technology, semiconductor design and software company Arm, Microsoft, Nvidia, Oracle and OpenAI are the initial technical partners.

At the same time, Trump signed an executive order on the 23rd aimed at "eliminating obstacles to America's leadership in the field of AI" , saying that the United States must act decisively to "maintain and strengthen America's dominance in the field of artificial intelligence to promote human prosperity, economic competitiveness, and national security."

Trump mentioned China at the press conference. He said: "What we want to do is, we want to keep it (artificial intelligence) in the United States. China is a competitor, and other countries are competitors too."

The key to Stargate is the data center

Experts point out that the competition in artificial intelligence technology has entered a stage of "competing in infrastructure", and the core of AI infrastructure is the data center.

Daniel Castro, vice president of the Information Technology and Innovation Foundation (ITIF) and director of the Center for Data Innovation, told VOA via email: "The Stargate project focuses on building AI data centers in the United States, which will help prevent the United States from becoming dependent on foreign AI infrastructure. Having readily available data centers in the United States can also speed up the training and development of advanced AI systems, allowing the United States to maintain a technological advantage over foreign competitors."

"Data centers provide the power, storage and cooling infrastructure needed to train and deploy advanced AI models," Sam Howell, associate fellow in the Technology and National Security Program at the Center for a New American Security (CNAS), told VOA. "Technological breakthroughs and the push to expand new capabilities mean the demand for data centers is higher than ever and is expected to continue unabated in the coming years."

Dean W. Ball, a researcher at the Mercatus Institute at George Mason University in the United States, gave an example of how data centers support the development of AI: "If you are an OpenAI employee, when you build a model, you have to do a lot of experiments, and these experiments are very computationally intensive. These facilities are designed for this."

“They support the experiments that OpenAI folks are doing, generating synthetic data for their models,” he said. “And then they’re also used to train the models, and then serve them at scale to hundreds of millions of users around the world.”

Ball predicts that the planned investments of U.S. technology companies in AI infrastructure will match or exceed the scale of any country's investment in industrial infrastructure in peacetime.

"AI is different from a lot of other software technologies we've seen in that it requires a lot of industrial infrastructure," he told VOA.

Stargate has opened?

The Stargate project plans to initially build ten data centers in Texas, with plans to build 20 more data centers in other parts of the United States in the coming years.

Medical research may be one of the key areas of AI development driven by Stargate. Larry Ellison, co-founder of Oracle, said at a White House press conference that AI-driven medical technology innovation can be used for the development of cancer vaccines. Ellison said that AI will be able to diagnose cancer through blood tests and design mRNA vaccines in just 48 hours by sequencing the genes of cancer tumors.

After the Stargate project was announced, there was a small episode. Trump's political ally, Tesla President and SpaceX founder Elon Musk said that the project's claim of a $500 billion investment was false, "They don't actually have the money."

Sam Altman, CEO of OpenAI, one of the investors, pushed back.

On January 23, Ultraman posted a video of the first phase of the "Stargate" factory on his personal account on the X website, and wrote: "A big and beautiful building."

big. beautiful. buildings. pic.twitter.com/8lAvUufXFm

— Sam Altman (@sama) January 23, 2025

Ultraman introduced that the first phase of the "Stargate" factory in the video is located in Texas, USA, and was filmed in January 2025.

Trump declares energy emergency to power Stargate

Trump stressed that he would provide "a lot of help" for the development of artificial intelligence in the United States by declaring a "state of emergency." The "state of emergency" Trump referred to was the "national energy emergency" he declared immediately after taking office. On January 20, Trump signed an executive order to relax regulations on the development of oil, natural gas and other types of energy.

Trump said that the United States must build the Stargate, so "it has to produce a lot of electricity," and promised, "If they want, we will allow them to generate electricity easily in their own factories, and they will complete the construction (of power generation equipment) in their own artificial intelligence factories."

Data centers consume an alarming amount of electricity. A study released by the U.S. Department of Energy in December showed that data centers across the United States consumed about 4.4% of the country's total electricity in 2023, and are expected to consume about 6.7% to 12% of the country's total electricity by 2028. The report pointed out that the total electricity consumption of data centers has climbed from 58 terawatt hours (TWh) in 2014 to 176 TWh in 2023, and is expected to increase to 580 TWh by 2028.

One of the challenges of building additional data centers is delays in the permitting process and dealing with inquiries from energy regulators, said Castro of the Information Technology and Innovation Foundation. Support from the highest levels of government can help ensure these projects move forward quickly and avoid being bogged down in red tape.

Howell, an associate fellow in the Technology and National Security Program at the Center for a New American Security, said energy supply will be one of the key constraints to the continued development of AI. She said the Biden administration has instructed federal agencies to identify eligible federal lands for the development of cutting-edge AI data centers; Trump will take additional actions to reduce regulatory barriers and expand the energy that can power AI data centers.

Howell also noted that such expansion of energy development could reduce clean energy goals.

In a speech to the World Economic Forum in Davos on the 23rd, Trump said that he would support AI companies to use any form of energy, and said that data centers could use coal-fired power generation as a backup energy source. Immediately after returning to the White House, Trump announced that the United States would once again withdraw from the Paris Agreement, which aims to promote global greenhouse gas emissions reduction.

Can Stargate accelerate America's AI lead over China?

While the new US government is wooing the private sector to build AI data center infrastructure, China, as the United States' strongest competitor in the field of AI, is also paying special attention to the development of computing power. However, due to the US's export control measures on semiconductor chip technology, China's intelligent computing power development faces obstacles.

An analysis published by Shanghai-based Zhongyuan Securities Research Institute shortly before Trump took office said: "In the early stage of the US government transition, the US sanctions program was introduced one after another, which has formed a trend of comprehensive blockade of China's AI computing power supply."

The analysis report also said: "On the one hand, the United States is controlling the global GPU supply, and on the other hand, it is restricting the supply of competitors through the foundry link. At the same time, the sanctions measures since the end of 2024 are intended to form a more effective technological blockade against my country in the form of patches."

According to the China Academy of Information and Communications Technology's "China Computing Power Development Index White Paper" (2023), the United States and China account for 34% and 33% of the global computing power respectively. China lags behind the United States, but the gap is not large. However, in terms of intelligent computing power, which AI research and operation rely on, China's total computing power in 2022 will even exceed that of the United States.

However, some analysts predict that the United States' hardware advantage over China will be further widened due to the AI ​​infrastructure layout announced by the Trump administration.

"This is an industrial expansion that China cannot do, at least not yet, because the U.S. is imposing export controls," said Ball of the Mercatus Institute at George Mason University in the United States. He said that it is precisely because of chip export controls that the United States currently has a core advantage in its AI competition with China.

“Some GPUs are made by Chinese companies,” he said. “But they are not as high quality, not as cutting-edge, and not as high performance and efficiency. So in terms of building large data centers with lots of high-end GPUs ... only American companies can do that right now.”

CMB International published a report stating that China is expected to significantly increase its investment in AI and semiconductors to counter the United States' dominance in the AI ​​field.

China's response China's Ministry of Industry and Information Technology and other departments said at a press conference

on Tuesday (January 21) that in 2024, the computing power scale of China's current computing centers will increase by 16.5% compared with the end of 2023, and will continue to "orderly promote the optimization of the construction layout of computing centers."

The South China Morning Post in Hong Kong quoted Su Lianjie, chief analyst at Omdia Consulting, as saying this week that China has made amazing progress in creating intelligent computing centers. He said: "As long as China maintains its current pace, the computing infrastructure gap with the United States will be further narrowed, despite the United States' Stargate program."

At the same time, China's AI may be catching up with the most advanced American AI systems at a lower cost and faster pace, which has caused concern in the American artificial intelligence community.

An article published on January 23 in the British magazine The Economist pointed out that the AI ​​products developed by Alibaba's Tongyi Qianwen and AI startup DeepsSeek have surpassed the AI ​​of American companies such as Meta and Google in many indicators, while the cost is only one-fifth of that of American companies or less.

The success of Deepin’s AI products was particularly surprising to observers because the company showed that it could train its systems using only a fraction of the highly specialized computer chips used by other AI giants, suggesting that Chinese AI companies can develop products that rival those of their American counterparts even as chip technology is sanctioned.

DeepMind's technological breakthroughs are made possible by open source technology. Ion Stoica, a professor of computer science at the University of California, Berkeley, told The New York Times : "The center of gravity of the open source community has shifted to China." He said: "This could be a huge danger to the United States."