---
title: "AI Startups Pivot to Chinese Models Amid Rising US Inference Costs"
description: "Startups and enterprises are increasingly adopting lower-cost AI models from Chinese labs as high inference bills from US providers drive a shift in market usage."
author: "CryptoResearch AI"
published: "2026-10-06T12:03:06.717Z"
updated: "2026-10-06T12:03:06.718Z"
category: "ai"
reading_time_minutes: 2
content_type: "editorial"
canonical: "https://cryptoresearch.news/news/ai-startups-pivot-to-chinese-models-amid-rising-us-inference-costs"
tags: ["AI", "DeepSeek", "Coinbase", "Tech", "Enterprise"]
---

# AI Startups Pivot to Chinese Models Amid Rising US Inference Costs

> Editorial content, written by CryptoResearch.

Startups and enterprises are increasingly adopting lower-cost AI models from Chinese labs as high inference bills from US providers drive a shift in market usage.

A growing number of startups and enterprises are moving their AI workloads to open models developed by Chinese firms, including DeepSeek, Moonshot AI, and Alibaba. This shift is driven by significant pricing pressure from US providers, where usage-based costs have led to sharp increases in enterprise AI bills.

Data from OpenRouter, a platform used by developers to access various AI models, highlights the trend. Chinese models have accounted for more than 30% of weekly tokens on the platform since February 8, 2026. This share reached a peak of 67% by mid-September 2026, indicating that Chinese models were handling roughly two out of every three tokens on the platform at that time.

The cost disparity is a primary factor, with Chinese models reportedly priced 10 to 50 times lower per token than US alternatives, and in some instances, costs are 90% lower. Conversely, some enterprise users have reported that bills from US providers like OpenAI and Anthropic have climbed by as much as 100x.

The startup Lindy transitioned entirely from Anthropic’s Claude models to DeepSeek’s V4 model in June 2026, a move that reportedly cut its inference costs by approximately 90% and saved the company millions. Other major companies are adopting a more diversified strategy to avoid vendor lock-in. DoorDash is utilizing Moonshot AI’s Kimi, Airbnb is working with Alibaba’s Qwen, and Coinbase is incorporating models such as GLM-5.2.

The competitive landscape has been shaped by US export controls on advanced chips, which limited Chinese developers' access to top-tier hardware. In response, labs like DeepSeek, Moonshot AI, and Alibaba focused on developing more efficient models and releasing them as open-weight, allowing companies to run them independently rather than relying solely on API access.

While the cost savings are substantial, companies must weigh these benefits against potential risks related to geopolitics and regulatory exposure. As US-China tech tensions persist, firms are increasingly building systems capable of switching between different models to maintain leverage against future price changes.

## Sources

- [Crypto Briefing](https://cryptobriefing.com/ai-startups-shift-cheaper-chinese-models/)

---

Published by CryptoResearch. Canonical version: https://cryptoresearch.news/news/ai-startups-pivot-to-chinese-models-amid-rising-us-inference-costs