Japan's Slow Growth, China's AI Race

Bloomberg Audio Studios podcasts radio news. Welcome to the Daybreak Asia podcast. I'm Doug Krer. We begin in Japan where economic growth unexpectedly slowed in the second quarter. Real GDP grew at an annual rate of just 1.1%. Now, much of that weakness reflects a slump in capital spending given the uncertainty stemming from the conflict in the Middle East. And that in turn could complicate efforts from the Bank of Japan to fine-tune communications as it weighs the timing of the next interest rate increase. For more, we caught up with Homeman Lee. He is senior macro strategist at Lombard ODA. He spoke with Bloomberg TV host Paul Allen and Heidi Strad Watts. Paul started the conversation with a question on whether the Bank of Japan has now an excuse to keep doing nothing and if the carry trade is still alive and well.

Japan's Economic Outlook and Bank of Japan Policy

Homeman Lee, senior macro strategist at Lombard ODA, pushed back against the notion that Japan's 1.1% GDP growth was anemic. He argued that the consensus 2.0% figure might have been too aggressive, especially considering the impact of energy market disruptions in the quarter. Lee highlighted that positive growth, coupled with deflator series data showing inflation above consensus and a weak yen, provides the Bank of Japan with grounds to consider further rate hikes. He suggested the BOJ might aim to counteract aggressive carry trades by raising short-term rates and potentially accelerating its tightening plans for the remainder of the year. Joint currency interventions further heighten the stakes for policymakers considering yen-supporting measures.

Investment Themes in Japan

Despite the mixed economic signals, Lee remains overweight on Japan, identifying AI as a key theme for the equity market. He also pointed to industrials and materials as sectors poised for growth, benefiting from a global cyclical boom. Lee believes that even with potential headwinds from interest rate adjustments and currency fluctuations, the broad Japanese market is well-positioned due to attractive valuations and a positive earnings outlook. Specifically, financials, some industrials, and materials, alongside the AI theme, are seen as strong contenders for the rest of the year.

The full discussion can be found on Bloomberg Podcast's YouTube channel.

China's AI Race and Competitive Dynamics

The conversation then shifted to the burgeoning AI sector in China. Anthony Stevens, Bloomberg strategist, and Robert Lee, Bloomberg analyst, discussed the rapid advancements in Chinese AI models. They noted that Alibaba's Quanzhou models have achieved over three billion downloads globally in six months, and ZAI is set to upgrade its flagship model soon. This puts China in direct competition with US firms like Anthropic and OpenAI.

Stevens observed that China's AI progress is noteworthy, especially as macro conditions in Japan and the US outside the AI sector remain challenging. He highlighted the intense pressure Chinese models are exerting, leading to market scrutiny of companies like Baidu and Minimax after their results, particularly concerning unmet expectations for continuous technological advancement and monetization.

The Gap Narrows: China's AI Capabilities

Robert Lee elaborated on China's narrowing AI performance gap, stating that Chinese models have significantly improved over the past three months, now trailing US data by only about 5%. He attributed this to a shift towards larger, more complex models and China's inherent software expertise, boasting major software companies and a large pool of developers. Lee believes this trend of closing the gap will continue.

US vs. China AI competition and Geopolitical Risks

When asked if China could eventually overtake the US in AI, Lee suggested it's a neck-and-neck race. However, he questioned the ultimate significance in a bifurcated world, noting that Chinese companies are likely to focus their export efforts on emerging markets. The potential for sustainable monetization in Western markets, dominated by established AI players, remains uncertain. Lee also flagged concerns raised by US Secretary of Commerce Gina Raimondo regarding potential intellectual property theft, referencing accusations against Moonshot AI by Anthropic, which underscores the intertwined geopolitical risks in the AI competition.

Jupai's Performance and Market Challenges

The discussion turned to the performance of specific companies, with Anthony noting that the market's punishment of Jupai today makes sense. He explained that Jupai's new model, while focusing on cybersecurity, hasn't significantly improved its coding performance, a key area where it needs to compete with rivals like Fable 5 and Cursor. The accelerated pace of progress means companies need to offer materially better performance to satisfy investor expectations.

Robert Lee added that Jupai lacks the scale and resources of tech giants like Alibaba, Tencent, and Huawei, whose AI operations are subsidized by their core cash-generating businesses. He pointed out that Jupai's operating losses have increased significantly since its IPO, and its latest model ranks only 27th globally. Lee emphasized the commoditized nature of the large language model market in China, with nearly 1,000 models competing, suggesting a shakeout is needed, along with price increases and the eventual advent of AGI.

Hardware Dominance and Geopolitical Tensions

The conversation concluded with a comparison of Hong Kong and Taiwan's equity markets, noting their similarity in performance driven by hardware companies. Lee stated that it's easier to pick hardware winners than LLM winners, especially given the current shortages. He highlighted companies like Kingboard Laminates, Nan Ya PCB, and Unimicron, whose success stems from producing essential, albeit less complex, components for data centers. The geopolitical dimension of this hardware race, particularly concerning optical component suppliers between the US and China, was also mentioned, with China potentially having an advantage in rapidly scaling up capacity.