Relocation Is a Means, Not an End | Generated by AI
This isn’t a technical question, so I’ll drop the format and just be direct with you.
The honest answer: probably not as a primary goal, given everything in your situation. Here’s the reasoning.
The constraints you’ve already documented argue against it. You have a 900K mortgage, two daughters, a wife with her own career, and parents nearby. A US move isn’t “you relocate” — it’s uprooting a four-person family system with elderly-parent proximity that Chinese families typically don’t sacrifice lightly. Your own playbook says startup viability comes when mortgage drops to ~500K. A US move resets that clock, doesn’t advance it. The visa path alone (H-1B lottery odds are brutal, ~25-30% per attempt; O-1 is realistic for you but needs deliberate portfolio-building) could eat 2-3 years before you’re even working there legally.
The thing you actually want doesn’t require the US anymore. Your idols are instructive here. Karpathy is American, sure. But Liang Wenfeng built DeepSeek entirely in China and is now arguably more influential than most US lab researchers. The frontier of open-weight models, MoE architecture, and efficient training is being pushed hard from China right now. You’re already training GPT-2 from scratch on MI300X, consuming 2B tokens/year, ranked top 6% in AI assistant usage. The geographic premium the US held in 2018 has compressed. What’s scarce now is demonstrated capability, and that’s location-independent — your GitHub, your notes, your trained checkpoints travel anywhere.
Where the US still genuinely wins: compensation (a strong AI engineer clears $300-500K+ at frontier labs vs. a fraction of that in Guangzhou), proximity to the actual research frontier and the people making it, and equity upside if you land somewhere early. If your real goal is “work at a frontier lab” (Anthropic, OpenAI, a top startup), being in the US is still a meaningful advantage despite remote work. That’s a legitimate reason. But notice it’s a different goal than “build my own AI-native startup” — and your stated long-term goal is the latter.
The reframe I’d push: “Go to the US” is a means, not an end. What’s the actual end? If it’s learn at the frontier → you can do that remotely now, faster than a visa would let you. If it’s maximize income to kill the mortgage and fund the startup → run the math; a remote contract with a US/global company often beats relocation when you account for cost-of-living, family disruption, and the visa lag. If it’s build credibility and optionality → ship in public, get the GitHub traction, present at the level where O-1 becomes trivial, and then the door opens on your terms rather than a lottery’s.
A concrete alternative path that fits your existing playbook: spend the next 2-3 years building such an undeniable public AI-engineering portfolio that you have leverage — remote offers, O-1 eligibility, or startup traction. Then decide from strength. That dominates “apply for grad school / H-1B now and hope,” because it keeps every option open including the US one, while not resetting your financial and family clocks.
The one scenario where I’d say go sooner: if you and your wife both want the move for life reasons beyond career (kids’ education, long-term emigration, lifestyle), and the family resistance is specifically about startup risk not leaving China. Those are different objections. Worth separating which one you’re actually facing.
What’s driving the question — is it income, frontier access, emigration for the family, or escaping the bank-contractor ceiling? The right answer forks hard on that.