GPU IPO Hype vs. Startup Cash Flow Realities

When this news first popped up, I thought domestic GPUs had already won. But scrolling down: Biren's HK stock jumped 80% on day one, market cap hit HK$84.6 billion; H1 2025 revenue was RMB 58.9 million against a loss of RMB 1.6 billion. MetaX surged over 568% at its STAR Market opening, breaking through RMB 300 billion in market cap—reportedly, winning one lot could net you RMB 300k. The capital side is buzzing, but the business hasn't started balancing the books yet.

For startups, this direction is worth watching, but don't mistake applause from the financing market for orders from the operating market. I manage a team of thirty; my biggest fear is that after buying cards, the entire burden of software stacks, drivers, operator adaptation, and ops responsibility falls back on us. No matter how high the book valuation, it can't pay salaries at month-end; no matter how loud the IPO bell rings, it can't cover electricity bills in the server room.

My advice is straightforward: Don't treat domestic GPUs as a religion; put them into your cash flow statement first. Buying cards is primarily an operational decision. Depreciation, electricity costs, maintenance, failure compensation, customer payment cycles—if any of these aren't calculated clearly, they'll turn into holes at month-end. Only when customer payments can cover these costs should you talk about replacement; if they can't, even a RMB 300 billion market cap won't save your cash flow.