First Audited Look at Inference-Economics: MiniMax Hit 24.6% Margin, Peer Lost 75% of OpenRouter Volume
AccBalanced · x · 2026-09-28
@RMladek launches a deep thread on the economics of inference companies — providers that rent GPUs, run an inference stack, and sell tokens. The category has zero US IPOs, but MiniMax and Zhipu (Z ai) listed in Hong Kong in January, giving audited numbers for the first time.
Key facts:
- Token demand is exploding exponentially, yet price per token keeps falling while GPU costs rise, squeezing margins from both sides.
- One of the two went from losing money on every token to a 24.6% gross margin in 18 months.
- The other lost 75% of its OpenRouter volume in ten weeks.
The thread dissects MiniMax's financials line by line, offering a rare data-backed look at how inference providers can become profitable.
More from Venture
- Yu Bo: Founders chasing productivity miss the bigger opportunity — killing time — oran_ge · 2026-09-28
- US creative industries lost 200k+ jobs in four years, worst stretch outside recessions — korymath · 2026-09-28
- VC LPs' top fear: an efficiency step-change that stops the AI music — JosephJacks_ · 2026-09-28
- If only Google could buy a popular Muse alternative for a measly $10B — willcb · 2026-09-28
- Dave Blundin on AI investing, founders, valuations and next opportunities — PeterDiamandis · 2026-09-28
- LAUNCHBridge uses AI to set up an LLC, website and payments in 72 hours — PeterDiamandis · 2026-09-28