The founder of Oriental Harbor recently found himself at the center of market controversy. Investors questioned the 35.4% drawdown of the Changjiang Oriental Harbor No. 2 fund in July, with only a 7% rebound in the first week of August, alleging that several stocks were sold at the bottom. More pointedly, they challenged that if he was heavily invested in AI, why hadn't the fund recovered its initial value after years of significant gains in AI assets?

On the evening of August 11, Dan Bin clarified that the fund remains fully invested. Before July's deep correction, it was heavily weighted in AI supply chain stocks without major reductions; the recent decline was a result of the broader AI sector downturn. The fund maintains positive overall returns, and investors uncomfortable with periodic performance can redeem at any time. In a subsequent live stream, he directly addressed what he called "misunderstandings" and "rumors."

His latest U.S. stock holdings, disclosed in a 13F filing, show the Oriental Harbor overseas fund's portfolio size at $1.65 billion (approximately RMB 111.3 billion) as of the end of the second quarter, up 45.6% from Q1. The holdings underwent a major overhaul, expanding from 12 to 13 stocks. The top ten holdings are: Alphabet Inc (GOOGL), Intel Corporation (INTC), NVIDIA Corporation (NVDA), SanDisk Corporation (WDC), Micron Technology Inc (MU), Advanced Micro Devices Inc (AMD), Marvell Technology Inc (MRVL), Taiwan Semiconductor Manufacturing Company Limited (TSM), Arm Holdings plc (ARM), and Broadcom Inc (AVGO). Seven new positions in AI upstream hardware, including Intel, SanDisk, AMD, Marvell, Arm, Broadcom, and Lumentum, were added in Q2, accounting for about 50% of new positions. Intel jumped to the second-largest holding at $258 million (16% of the portfolio), while SanDisk became the fourth-largest at $176 million (11%). Micron saw a 100% increase in holdings. On the sell side, Alphabet and NVIDIA were slightly reduced, TSMC was cut by 34%, and Amazon by 67%. Positions in Apple, Tesla, and Alphabet were fully liquidated, along with all leveraged ETF tools like the 2x Long Google ETF.

This shift deepened the divergence of opinion around Dan Bin. In a live stream, he offered a fresh interpretation of his U.S. holdings and directly addressed the most contentious issues in AI investing, including why he dared to "fire all remaining bullets" during the July selloff, whether the AI narrative has changed with the emergence of open-source models like DeepSeek and Kimi, whether rising capital expenditures by overseas tech giants risk repeating the internet bubble, the rationale behind Q2 portfolio changes, the moat of computing giants amid open-source challenges and U.S.-China competition, conditions for triggering a strategy shift, how external variables like chip restrictions and regulation affect decisions, and which AI sectors have the most potential over the next 5-10 years.

On the July selloff and buying the dip

Dan Bin views the July volatility within a ten-year AI super-cycle, arguing that "watching waves from the shore" and "watching the sea from a height" reveal completely different markets. He notes that in long-term bull markets, corrections are natural and often present excellent buying opportunities. Historical data, such as the VIX panic index exceeding 50, consistently signals good entry points. He emphasizes that without vision and conviction, investors risk missing an entire era. He argues that buying during downturns is a skill to master, not just a mental challenge, and suggests that even fearful investors should start with a small position.

On the impact of open-source models

Dan Bin applies the Jevons Paradox to explain the impact of open-source models. He argues that technological progress reduces costs, which in turn stimulates broader applications, rather than diminishing AI demand. He cites examples like NVIDIA's support for DeepSeek and the rapid growth of token usage amid falling costs. He believes that the development of Chinese large models actually promotes the AI era, though it may challenge companies like Anthropic and OpenAI that rely on high fees. He points to the capital expenditures of major U.S. companies, which exceed $700 billion, and the positive cash flow and demand confirmed by company earnings calls as evidence that the AI logic remains intact. He concludes that the competition between models keeps costs reasonable and drives progress.

On the risk of an AI bubble

Dan Bin dismisses the comparison to the internet bubble, noting that the dot-com era lasted a decade before peaking, while AI is still in its early stages. He observes that major tech companies are funding their capital expenditures through internal cash flow, not by issuing stock, and that the positive feedback loop from AI investments is already visible in the earnings of Amazon, Microsoft, and Alphabet. He highlights that AI is now driving innovation across industries like pharmaceuticals, finance, and software, using Palantir and Microsoft as examples. Unlike the internet bubble, many AI-supporting companies are highly profitable, with some trade at low forward P/E ratios. He argues that if the AI cycle is a decade-long or longer, valuation declines could represent opportunities to buy at attractive prices.

On the moat of computing giants and U.S.-China competition

Dan Bin believes in the principle of "the strong get stronger" and that NVIDIA is unlikely to be disrupted by GPU competitors in the short term. He argues that open-source models like DeepSeek and Kimi pose challenges, not fundamental threats, to leading companies. Regarding U.S.-China competition, he sees AI as a matter of national security, with China likely to support its own AI supply chain. He suggests that if Chinese companies successfully develop competitive GPUs, it could threaten U.S. firms, but he remains focused on leaders in the sector.

On portfolio changes and strategy

Dan Bin explains that his adjustments over the past year centered on the AI supply chain. He sold companies potentially disrupted by AI and replaced leveraged tools with positions in memory and Intel in the first half of the year. He acknowledges being late on memory investments and not as heavily weighted in NVIDIA or Alphabet as he would have liked. He uses a long-only strategy, adjusting based on supply-demand imbalances and valuation. He also employs a structural strategy of selling covered calls at high prices to stabilize returns, using the proceeds to buy on dips, such as with Micron and Apple. He believes that leading AI supply chain stocks in the A-share market are likely to reach new highs, comparing 2025 to 1994 for the internet era, a period of consolidation rather than a peak.

On conditions for a strategy shift

Dan Bin identifies company earnings and capital expenditure changes as key indicators. He would consider a significant shift if hardware companies like NVIDIA cut capital spending, or if Alphabet or Amazon stopped investing in AI. He notes that the current cycle is different from the internet era, as it is a battle among giants with massive investments, creating insurmountable barriers for smaller players. He is also considering adding back positions in Alphabet and Apple as the industry evolves. He emphasizes the need to be agile in observing these changes, as they could lead to rapid transformations in the sector.

On external variables like chip restrictions and regulation

Dan Bin observes that the AI era is characterized by a decoupling between the U.S. and China, unlike previous tech cycles. He sees opportunities in Chinese companies that can achieve self-sufficiency and form complete supply chains, as well as those that participate in global supply chains, such as the NVIDIA and Tesla ecosystems. He acknowledges that these external changes increase investment difficulty but believes there are still selectable opportunities for those willing to adopt a multi-dimensional perspective.

On promising AI sectors over the next 5-10 years

Dan Bin sees major opportunities in both hardware and software (applications) within the AI era. He is particularly impressed by the widespread adoption of autonomous driving in the U.S., citing Tesla's self-driving technology. He also sees potential in robotics, AI's expansion into physical spaces, and AI-powered healthcare, such as nanobots for medical diagnosis. He warns that the risk of missing the era is greater than the risk of volatility, urging investors not to overlook this opportunity.

On why he personally responds to criticism

Dan Bin explains that he uses his platform to document his work and life, not to seek attention. He addresses what he considers to be misunderstandings, such as a persistent rumor that Oriental Harbor charges a 5% management fee, which he clarifies is false, with actual fees ranging from 0.25% to 0.5%. He believes in speaking out to counter falsehoods, as remaining silent can allow them to persist. He states that his primary goal is to keep a record of his observations, and he welcomes others to follow along if they choose.