Dan Bin’s Dongfang Hongyuan Overseas Fund just filed its Q2 2026 13F. The headline number: $1.65 billion in US equities, up 46% from Q1. That’s not the story. The story is what they bought and what they dumped. Google remains the top holding at $370 million. But the real signal is the new money: Intel, SanDisk, AMD, Marvell, ARM, Broadcom, Lumentum. They added Micron. They cut NVIDIA, TSMC, Amazon, Meta. They exited Apple, Tesla, and leveraged ETFs entirely.
This isn’t a rotation. It’s a structural bet on the upstream of AI – the hardware that makes the GPU work. And I’ve seen this pattern before.
Context: The Fund’s DNA
Dan Bin is a macro-aware, tech-focused manager. His fund’s size and growth hint at institutional inflows – likely from the post-ETF approval wave. The Q2 filing shows a deliberate shift away from the obvious AI winners (NVIDIA, TSLA) toward the semiconductor supply chain, storage, and optical communication. This is a bet on the enabling layer, not the application layer. In crypto terms, it’s like selling the L1 tokens and buying the sequencer infrastructure.
The fund’s previous holdings were heavy on consumer-facing tech and AI leaders. Now they’re loading up on chipmakers, memory manufacturers, and optical networking. The message: the real value in AI is in the physical components, not the software or services.
Core: The Order Flow Analysis
Let’s break down the numbers. The fund increased its stake in Micron, added SanDisk (now part of Western Digital), and entered Lumentum. These are not high-growth darlings – they are cyclical, capital-intensive, and often undervalued. But they are essential.
Why optical communication? Lumentum makes photonic components for data centers. As AI compute scales, bandwidth bottlenecks shift from GPU memory to interconnects. Optical is the new L2 – it’s the rails that move data between chips. The fund is betting that the market underestimates this bottleneck.
Based on my experience auditing smart contracts in 2017, I learned that the most valuable positions are often in the components, not the final product. The real money in the ICO boom was in the auditing firms, not the tokens.
Similarly, the semiconductor supply chain is the “auditor” of the AI era. Every new GPU needs more memory, faster interconnects, and better cooling. The fund is positioning for that.
But there’s a catch. The fund also cut NVIDIA and TSMC. That’s the contrarian move.
Contrarian: Retail vs. Smart Money
Retail is piling into NVIDIA. FOMO is real. The fund is selling. Why?
One possibility: they see NVIDIA’s valuation as a trap. The narrative that “AI is the next internet” has already priced in perfection. But the semiconductor cycle is inherently mean-reverting. In 2022, I saw the Terra collapse – the same “this time is different” narrative applied to Luna. It wasn’t different.
Another possibility: the fund is hedging against a bubble in AI stocks. By buying the upstream suppliers, they get exposure to AI growth without the concentration risk. This is a delta-neutral strategy disguised as a long-only portfolio.
I’ve made similar trades. In 2020, during DeFi Summer, I ran a delta-neutral arbitrage on Compound and Uniswap. The yield was in the borrowing market, not the farming. The fund is doing the same: the real yield in AI is in the hardware, not the hype.
Code is law, but bugs are justice. Intel has had bugs – the Raptor Lake instability issues. But the market has already priced in the pain. The fund is buying the fear. That’s smart.
Greeks don’t lie – but they don’t predict the future either. The volatility in semiconductor options is screaming that the market is unprepared for a supply chain disruption. The fund’s move is a bet on that volatility.
Takeaway: The Plumbing Trade
So what’s the actionable insight? The fund is telling us that the next leg of the AI trade is not in the consumer-facing companies. It’s in the infrastructure that makes the data centers run.
But here’s the twist: this is also a warning. The fund’s size and the speed of the pivot suggest that institutional money is flowing into the same sectors. When everyone piles into the same trade, the edge disappears.
NFT floor is a feeling, not a number. The same applies to semiconductor stocks. The fundamentals are solid, but the sentiment can shift quickly. The fund is positioning for the long term, but the market may not cooperate.
My take: watch the optical communication names. Lumentum, Coherent, and others could be the next bottlenecks. But don’t buy the hype – buy the code.
The real bull market isn’t in AI directly. It’s in the tools that make AI possible. The fund understands that. Now the question is whether the rest of the market will catch up – or if this is another manufactured narrative.
I’ve seen both. In 2021, I tracked the Bored Ape Yacht Club wash trading. The same pattern: institutions buying the narrative, retail following, and then the rug. This time, the rug might be a hardware shortage.
Greeks don’t – but the order flow does. Follow the upstream, not the hype.