ChatGPT stimulates demand for chips, and the surge in orders shows that the global AI chip race is heating up

ChatGPT, a conversational large-scale language model published by artificial intelligence (AI) lab OpenAI, has brought AI and its various new uses to the flames, and the surge in orders shows that the global AI chip race is heating up.

According to a UBS study, ChatGPT has been used by hundreds of millions of people since its launch in November 2022, making it the fastest growing consumer app in history.

Yasuo Mihashi, co-chief strategy officer of Japanese chip test equipment supplier Advantest, said the global race to develop powerful computing clusters and next-generation AI training systems is prompting chipmakers to buy more of the company's chip test tools.

At present, the demand for consumer electronics is sluggish. But according to data compiled by Bloomberg, the AI surge is driving up orders from Nvidia and AMD, which provide key AI training chips and rely on Advantest as its main test tool supplier.

"We are a leader in the industry, so we benefit when a product like ChatGPT expands the use of high-performance computing." Yasuo Mihashi said.

However, the current variables exist at the regulatory level: for example, Italy, the United States and other European and American countries are considering the upcoming tightening of regulatory restrictions on ChatGPT.

ChatGPT stimulates demand for chips

Advantest's stock has surged 36 percent this year, near record highs, as the promise of ChatGPT and other products aimed at making internet services smarter and more humane captured the imagination of the world. The company's operating profit for the fiscal year ended March is expected to rise 48 percent to 170 billion yen ($1.3 billion).

Microsoft and Google, part of Alphabet, have pioneered the integration of generative AI into their web tools, and the race to build generative AI is underway in other countries and regions around the world.

In simple terms, data centers are essential for training and operating AI models that can inform everything from autonomous driving to advanced chatbot assistants. This requires access to tens of thousands of graphics processing units, which are currently the best hardware for analyzing data in real time.

"The demand for data center GPUs and testers is sure to grow rapidly." Mitsuhiro Osawa, an analyst at Ichiyoshi Research Institute, said: "Hardware is necessary for next-generation social infrastructure such as autonomous driving and AI, and these technologies are being heavily invested by countries around the world. ”

Masahiro Wakasugi, an analyst at Bloomberg Intelligence, said the increasing use of AI is driving chip miniaturization and advanced packaging, although it will also increase the likelihood of defects in the manufacturing process. "This structurally boosts the demand for chip testers, helping manufacturers guarantee performance and quality."

Toyo Securities analyst Hideki Yasuda said global revenue from high-performance GPU chip testers could surpass smartphone chip tester revenue in a few years.

"Server chips are going to be bigger, more complex, and require more time to test," he said. The only solution for chipmakers is to buy more tools and test more chips at the same time. ”

Gokul Hariharan, co-head of technology, media and telecommunications research at JPMorgan Asia Pacific, said: "We expect the overall chip market to see a significant decline in the first half of 2023. But at the same time, driven by high-performance computing, there will be strong long-term growth in the coming years. ”

Regulation tightening is variable

Although ChatGPT is in the limelight, many countries are considering whether to tighten regulatory restrictions on ChatGPT.

On March 31, Italy announced a temporary ban on ChatGPT because it may violate the EU's privacy rulebook, the General Data Protection Regulation (GDPR). Italian regulators said companies that developed ChatGPT have 20 days to address data protection issues or pay a fine of 20 million euros or up to 4 percent of annual revenue.

But this is just the beginning of the ChatGPT troubles. Since OpenAI, the company that created ChatGPT, does not have a local headquarters in one of the EU's 27 countries, this means that data protection authorities in any member state can launch new investigations and enforce bans on it. According to European media reports, the use of ChatGPT's technology has begun to anger European governments, and its risks range from data protection, misinformation, cybercrime, fraud and cheating on school exams.

In Brussels, the European Commission is debating the AI bill. However, Margrethe Vestager, the executive vice president of the European Commission, has been more cautious about disabling artificial intelligence systems.

"Whatever technology we use, we must continue to advance our freedoms and protect our rights," Vestager tweeted, "which is why we do not regulate AI technology, but the use of AI." Let's not throw away in a few years what took decades to build. ”

Policymakers in the United States share the same concerns, and despite the bombardment of AI threat theory in Washington, they worry that disabling or halting AI deployment could cost the United States a competitive advantage in AI research.

Irene Tunkel, chief strategist of the U.S. asset department of BCA Research, an investment consulting firm, told First Financial Reporter: "In order for the United States to maintain its technological advantage over the rest of the world, the U.S. government will need to use policy and fiscal spending to promote the growth of AI over many years." We expect that more government resources will be allocated to AI, both defense and non-defense in nature, while the regulatory environment will facilitate increasing adoption of AI in the public and private sectors across the economy. ”

Tangkel believes that competition in AI will spur policymakers to ensure that AI and related technologies do not encounter major bottlenecks. She said that while many pieces of AI legislation have been proposed at the federal and state levels in the United States in recent years, more legislation does not necessarily mean tighter regulation. "There is a fundamental similarity between these legislations. The goal is to promote the development and use of AI in the public and private sectors in the United States, while also having guidelines to protect and safeguard AI consumers and users. She said.

Tonkel said U.S. budget allocations for non-defense and defense AI research have been growing steadily in recent years. For example, in fiscal year 2021, U.S. nondefense government agencies allocated a total of $1.5 billion for AI research spending, about 2.7 times more than in fiscal year 2018. At the same time, there has been a significant increase in government contracts awarded to private companies in the United States to encourage private companies to produce AI-related goods and services.

As the first step in potential regulation, the U.S. Department of Commerce has formally solicited opinions on relevant accountability measures on the 11th, including whether new AI models should undergo a certification process before release. The comment period is 60 days.