Circular AI Deals Fuel Bubble Debate | Bloomberg Tech: Asia 11/28/25

Key Concepts

• Circular Deals: Transactions where companies invest in their suppliers, creating a closed loop of funding that raises concerns about artificial inflation and sustainability in the AI industry.
• AI Bubble: The fear that the rapid growth and massive investment in Artificial Intelligence are unsustainable and could lead to a market correction or crash.
• High-Bandwidth Memory (HBM) Chips: Crucial components for AI processors, particularly NVIDIA's GPUs, experiencing surging demand.
• Monetization of AI: The challenge of generating sufficient revenue from AI products and services to justify the significant infrastructure and development costs.
• Asset-Liability Mismatch: A financial situation where a company's assets do not match its liabilities in terms of timing or amount, potentially leading to liquidity issues.
• Commoditization of Large Language Models (LLMs): The trend of LLMs becoming increasingly accessible and standardized, potentially reducing profit margins for individual developers.
• AI Ecosystem: The interconnected network of companies, technologies, and services that support the development and deployment of AI.

The AI Money Machine and Asia's Role

Bloomberg Tech Asia, with Shery Ahn in Tokyo and Annabelle Droulers in Hong Kong, is examining the multibillion-dollar circular deals fueling concerns of a bubble in the global AI industry and Asia's position within this AI money-making machine.

Concerns Over AI Bubble and Circular Deals

A dominant narrative in the tech industry has been the fear of bubbles in the AI space, coupled with concerns around circular deals. These deals involve companies investing in their suppliers, creating a loop where capital flows into companies that may not yet have proven business models, cycles into data centers, and returns to chip manufacturers like NVIDIA as revenue. This raises questions about the sustainability of current spending, with some experts predicting a significant correction by 2030 if revenue doesn't justify the AI revolution.

Asia's Tech Titans and the AI Boom

While some Asian tech players stand to benefit from the AI boom, the sustainability of this spending for others is under scrutiny.

• NVIDIA's Role: NVIDIA is central to this discussion, often described as "selling the shovels" in the AI gold rush. The company is striking mega-deals to invest in the very companies that buy its products, leading to concerns about suppliers funding customers and vice-versa, alongside murky revenue-sharing agreements. NVIDIA's Chairman, Jensen Huang, has attempted to temper criticism regarding monetization, emphasizing disciplined execution despite large ambitions.

• SoftBank's AI Exposure: SoftBank, a major financier, has active investments in AI, including OpenAI, which feed into its products. The Japanese conglomerate sold its entire stake in NVIDIA for $5.8 billion to fund its founder's AI ambitions, amplifying market anxiety over an industry bubble. SoftBank's portfolio spans OpenAI, ByteDance, and Perplexity AI. However, with rising debt, investors worry these bold bets could backfire if the AI boom falters.

• South Korean Chipmakers: Samsung Electronics and SK Hynix are seeing rising margins on High-Bandwidth Memory (HBM) chips, essential for NVIDIA's GPUs. Bloomberg Economics projects a 35% surge in demand for Korean chips next year, nearly double the current rate. Despite this, some analysts remain optimistic, stating, "I don't see that much bubble. There is a lot of potential."

• TSMC's Barometer: TSMC, the chipmaker for NVIDIA, AMD, and Qualcomm, is often seen as an impartial barometer for demand. When NVIDIA posted its lowest monthly revenue growth in a year in November, it fueled fears that the AI rally was unhinged from fundamentals. Valuations of Asian markets suggest companies in the region are more protected from real earnings expectations compared to their U.S. counterparts.

• Foxconn's Perspective: Foxconn, a major supplier of AI services to Oracle and directly involved with OpenAI's infrastructure ambitions, views the situation differently. The Foxconn Chairman stated, "I look at the user model. From that angle, there is a larger model for our manufacture, for our daily uses. We have great potential for the use of that capability. We know that. It's not just for us, it's for all businesses. They need it. So the demands are there. How to fulfill it, in what way in terms of finance? They will figure out a way. As long as there is demands." He believes that if OpenAI can achieve $100 per user annually, it could generate around $80 billion a year, making monetization achievable.

Key Arguments and Perspectives

• Skepticism vs. Optimism: While some skeptics warn of an overdue correction, others remain confident in AI's long-term potential. One perspective is that "A bubble has merely gotten ahead of itself, not the likelihood of growth in the future. I'm not saying it is worth its price, but there is substance under the foam."

• SoftBank's Risky Bets: Panelists expressed concerns about SoftBank's strategy. Amir Anvarzadeh noted that SoftBank is a "poster child for any bubble" and that "when there is a bubble bursting somewhere, short SoftBank." He questioned if SoftBank has "put the house on the wrong bed? Which is OpenAI." Shuli Ren highlighted SoftBank's promise to invest $30 billion in OpenAI despite not having the funds, necessitating the sale of NVIDIA and T-Mobile stakes. She drew parallels to WeWork, citing an "asset-liability mismatch" and the potential for OpenAI to face a liquidity crisis.

• Oracle's Struggles: Shuli Ren pointed out that Oracle's stock has given up all gains made since its $300 billion deal with OpenAI, indicating investors are penalizing the company for its AI pivot. Amir Anvarzadeh noted Oracle's low gross profit margin (17%) on AI infrastructure leasing compared to its core ERP business (70%), calling it "an accident waiting to happen."

• Commoditization of LLMs: Amir Anvarzadeh argued that large language models are becoming a commoditized business, unlike the internet bubble where companies had distinct ambitions. In the current AI bubble, "everybody wants to come out with the same thing, general artificial intelligent machine."

• Hardware Resilience: Amir Anvarzadeh believes TSMC is the safest bet in the hardware space as "everybody is going to TSMC." He anticipates that while shares might dip if a bubble bursts, TSMC will ultimately be fine. Companies more deeply entrenched in specific camps, particularly those heavily tied to OpenAI, are seen as more vulnerable.

• AI Investment Scheme Hierarchy: Shuli Ren described the AI investment scheme as a pyramid, with "substring" (likely referring to foundational AI models or infrastructure providers) and providers like NVIDIA and TSMC at the top, who will be fine. At the bottom are LLM developers like OpenAI, Alibaba, and DeepSeek, where market competition is expected to limit individual player profits.

China's AI Ecosystem

• Alibaba's Outlook: Alibaba's CEO expressed optimism, not seeing an AI bubble forming within the next 2-3 years, believing that AI resources and supply will not meet demand. Alibaba's cloud division, housing its AI app, has seen significant downloads. However, Bloomberg Intelligence is more bearish, citing a lack of clear ROI in the cloud division's performance, despite lower overall costs in China.

• Consolidation in China: Bloomberg Intelligence anticipates the AI ecosystem in China will coalesce around Alibaba, Tencent, and ByteDance.

Step-by-Step Processes and Methodologies

The transcript doesn't detail specific step-by-step processes for AI development or investment. However, it implicitly outlines a cycle:

1. Massive Investment: Companies like SoftBank and potentially OpenAI are committing billions to AI infrastructure and development.
2. Chip Demand: This investment drives demand for advanced chips, particularly HBM chips, benefiting manufacturers like NVIDIA, Samsung, and SK Hynix.
3. Infrastructure Build-out: Companies like Oracle and SoftBank are involved in building data centers and AI hubs.
4. Product Development: Companies like OpenAI are developing AI models and services.
5. Monetization Challenge: The critical step of generating sufficient revenue from these AI products and services to justify the investment.
6. Market Scrutiny: Investors and analysts are closely watching revenue growth, profitability, and the sustainability of these investments, leading to concerns about a bubble.

Data, Research Findings, and Statistics

• NVIDIA Stake Sale: SoftBank sold its entire stake in NVIDIA for $5.8 billion.
• Korean Chip Demand: Bloomberg Economics sees demand for Korean chips surging by 35% next year.
• OpenAI Users: OpenAI has 800 million users, with 35 million paying subscribers.
• OpenAI Infrastructure Commitment: OpenAI is committing over $1.3 trillion in infrastructure over the next few years.
• Oracle Deal Size: The deal between Oracle and OpenAI is valued at $300 billion.
• Oracle Cloud Profit Margin: Oracle has a gross profit margin of only 17% on AI infrastructure leasing.
• Alibaba AI App Downloads: Alibaba's AI app has been downloaded over 10 million times after its relaunch.
• Alibaba Cloud Revenue: Alibaba's cloud division revenue rose by $132 million.

Notable Quotes and Significant Statements

• "The ambition's large, but the execution is disciplined." - Attributed to NVIDIA's statement on their approach to AI.
• "We are not going to find the revenue to justify the AI revolution, which I think it is impossible by 2030. The likelihood is we have a significant correction." - A skeptical perspective on AI revenue justification.
• "I don't see that much bubble. There is a lot of potential." - An optimistic view on the AI market.
• "A bubble has merely gotten ahead of itself, not the likelihood of growth in the future. I'm not saying it is worth its price, but there is substance under the foam." - A nuanced perspective on the AI market.
• "I look at the user model. From that angle, there is a larger model for our manufacture, for our daily uses. We have great potential for the use of that capability. We know that. It's not just for us, it's for all businesses. They need it. So the demands are there. How to fulfill it, in what way in terms of finance? They will figure out a way. As long as there is demands." - Foxconn Chairman on the demand for AI.
• "SoftBank is a poster child for any bubble, but OK. When there is a bubble burst somewhere, short SoftBank. That's the rule of thumb because they are always all over any theme that comes up." - Amir Anvarzadeh on SoftBank's investment strategy.
• "I think the AI investment scheme is a pyramid. The ones on the top of the pyramid is substring and they have providers like NVIDIA and TSMC. They will be fine. On the bottom layer is the large language model developers like OpenAI, Alibaba, DeepSeek. From a traditional viewpoint, a market that is more competitive, the individual players are not going to make as much profit." - Shuli Ren on the AI investment hierarchy.

Conclusion

The discussion highlights the complex and potentially precarious nature of the current AI boom. While the underlying technology holds immense promise, concerns about unsustainable "circular deals," the monetization of AI, and the potential for a market bubble are significant. Asian tech giants like TSMC and Samsung appear well-positioned due to their critical role in the hardware supply chain. However, companies heavily reliant on specific AI models or services, such as OpenAI, face greater scrutiny and potential risks. The long-term success of the AI revolution hinges on its ability to generate tangible revenue and demonstrate a sustainable business model beyond the current wave of speculative investment.