Dan Bin has fully exited Microsoft, doubling down on Google and Nvidia to bet on AI commercialization, arguing Microsoft's narrative-driven gains have played out while Google is better positioned for a slower payoff cycle.
Dan Bin Liquidates Microsoft: Private Fund Titan's "AI Faith Monetization"
On April 28, two days before SEC filings were disclosed, Dan Bin dropped a line on social media: "Too bad fund management has individual stock position limits. If the custodian doesn't agree, we still have to trim positions eventually. Individual investors can beat institutions because they can concentrate holdings more — if the direction is right!"
Few people paid much attention at the time. Two days later, Oriental Harbor's overseas fund first-quarter holdings were revealed: Microsoft was wiped out entirely from over $90 million, Google's position was raised to 38% (about $426 million), Nvidia was increased to 20%, and new positions were added in TSMC, Micron, and Circle. The fund's total size stood at $1.133 billion, with the top two holdings accounting for nearly 60% of the portfolio.
That line about "trimming positions if the custodian disagrees" looked like bragging on the surface, but it actually reflects a real dilemma once a private fund gets big — when scale grows, compliance forces diversification, yet Dan Bin's playbook is "betting on a single track," and the two simply don't mesh. The fact that he pushed it to 38% this time means the custodian loosened the reins, but the statement is already on record: if Google falls, "it's not that I didn't want to concentrate — the rules wouldn't let me."
Dan Bin's cut of Microsoft was timed with uncomfortable precision. On April 29, Microsoft reported earnings: Azure grew 39%, AI annualized revenue hit $37 billion, and capital expenditure reached $190 billion. The numbers weren't bad, but the stock didn't budge — meanwhile, Alphabet was up 7% after hours and Meta plunged 6% over the same period. "Not moving" is itself a sign of indifference.
Microsoft's problem isn't weak results; it's that the story is played out. Over the past two years, the capital markets have grown tired of hearing "largest OpenAI shareholder + Copilot disrupting office work + Azure AI infrastructure." The question now is: of that $37 billion in AI revenue, how much reflects real customer usage, and how much is CFOs reluctantly buying to hit "AI transformation" KPIs? Microsoft never breaks it down — and doesn't dare to, because breaking it down might expose the cracks.
The bigger issue is that Microsoft's AI spending is shifting from an "asset" to a "burden." Of the $190 billion in capex, $25 billion is purely the result of forced GPU and memory price increases. That means Microsoft's pricing power upstream is eroding. During the same period, Google also poured $180-190 billion in, but the market rewarded it — because Google Cloud's growth and margins are genuinely improving.
Dan Bin sees it clearly: Wall Street now rewards profits, not spending.
Dan Bin's heavy bet on Google isn't a bet that it wins — it's a bet that it dies slowly enough.
Google's earnings do look good: Cloud beat expectations, and the AI monetization loop across search ads, cloud, and YouTube is closing. But that's only the side Google wants you to see.
The other side: it's getting squeezed in the model layer by OpenAI, with Gemini's presence far behind GPT-4o; in search, Perplexity and ChatGPT Search are nibbling away bit by bit. Google's counter is to stuff AI into every search result — but that precisely reveals it's the besieged defender, not the attacker.
Dan Bin's bet is this: search ad cash flow can hold for another 3-5 years, long enough for Google to get its Cloud and YouTube AI monetization up and running. This is a defensive play of "buying time for space," not an offensive play of "technological leadership." What he didn't say: if Google's search market share shows a visibly visible decline over the next four quarters, that 38% position instantly becomes a hot potato.
Betting on "shovel sellers" also carries a timing trap.
Nvidia at 20%, plus TSMC and Micron — the logic is straightforward: hyperscalers' total capex is projected to hit $725 billion in 2026, and shovels always sell out before the gold.
But the 2026 capex frenzy largely reflects orders placed in 2024-2025. TSMC and Nvidia's results this year are a product of demand forecasts made two years ago. The current signal is that Meta, Google, and Microsoft are all absorbing component price hikes — upstream pricing power has peaked.
Dan Bin says "Nvidia's new high this round is no accident; it's fundamentals continuously delivering." That's only half right: what's delivering is past orders, not future demand. If, in the second half of 2026, the giants cut their 2027 guidance due to ROI pressure, Nvidia's valuation logic would reverse in an instant. Treating a cyclical industry like a utility — that works on the way up, but you're swimming naked on the way down.
The more hidden move is in A-shares.
Dan Bin's A-share portfolio is heavy in Zhongji Innolight, Shenghong Technology, and Eoptolink — three optical module companies. Optical modules are China's version of the "shovel sellers" for AI computing power.
Connect his positions together: U.S. stocks bet on Google and Nvidia (application layer and chip layer), while A-shares bet on optical modules (connectivity layer). This is a cross-market hedge — if U.S. AI commercialization beats expectations, Google and Nvidia profit; if capex keeps surging, optical modules profit; if both play out, he wins on both fronts.
But risks are also exposed on both ends: if hyperscalers cut spending, optical modules fall even harder than Nvidia (no CUDA moat); if Google search gets disrupted, the 38% position gets cut in half. "The custodian won't agree to trimming" now reads more like a preemptive excuse for extreme concentration risk — it's not that he doesn't want to be more concentrated; it's that he doesn't dare.
Dan Bin's core thesis is this: AI is moving from a "technology race" into a "commercial realization" phase, but the pace of realization will be slower and harsher than expected. He chose Google because it can best withstand slowness; he chose Nvidia and TSMC because in a "slow realization" cycle, shovels sell out before the gold; he dumped Microsoft because its narrative dividend has already been cashed in early. If, in the second half of 2026, any one of Meta, Google, or Microsoft cuts its 2027 guidance due to ROI pressure, this hand goes from "precise foresight" to "catching a falling knife at the top" — assuming the capex peak hasn't already arrived.
He said it himself: "The market no longer blindly chases AI themes; only companies that deliver business growth, AI implementation, and controllable profits will earn recognition."
Translation: I know the bubble is still there, but I'm betting I can get out before it bursts.