Goldman Sachs Raises China AI Model Market Forecast by 30% to $13 Billion
0xBroomberg
Goldman Sachs lifted its ARR forecast for China's AI model market from $10 billion to $13 billion, citing aggressive price cuts, technical breakthroughs, and faster enterprise adoption — a signal that China's AI race has shifted from capability to cost-efficiency.
Where does the $13 billion forecast come from?
In an August 4 report, Goldman raised its annualized recurring revenue (ARR) forecast for China's AI model market by 30%, from $10 billion to $13 billion.
Three drivers behind the upgrade: vendors slashing prices, sustained technical breakthroughs, and accelerating enterprise adoption.
This means → Goldman sees the market in a virtuous cycle — lower prices pull in more users, which lifts total revenue even as unit costs fall.
Which companies did Goldman single out?
Goldman simultaneously raised year-end ARR forecasts for two Hong Kong-listed companies: Zhipu AI (智谱AI) to $2.5 billion and MiniMax to $1 billion.
The report stated: "We expect competition among China's AI models for the best cost-performance ratio to intensify further."
In plain terms = Goldman is not betting on who has the best technology — it is betting on who can deliver adequate capability at the lowest price. Cost-efficiency is the scoreboard.
Two vendors moved on the same day — what happened?
MiniMax released its H3 model last Friday with open weights, supporting multimodal processing across text, image, video, audio, and music, priced at just 30–50% of prevailing market rates.
On the same day, DeepSeek opened API access to its V4 Flash model, which matches Zhipu AI's flagship GLM-5.2 in front-end coding capability.
This means → two vendors played their cards simultaneously — one on price, the other on capability. That is exactly the cost-efficiency arms race Goldman's report describes, playing out in real time.
Where do these models actually rank globally?
According to Arena AI's latest leaderboard, Zhipu AI's GLM-5.2 ranks seventh globally in front-end coding, with DeepSeek V4 Flash close behind at eighth.
In plain terms = Chinese AI models have broken into the global top ten on a hard benchmark — they are no longer trailing the field.
This reflects a dual push: technical capability and aggressive pricing advancing in parallel — not "cheap but weak," but "cheap and competitive."
Can the $13 billion target be met — and what decides it?
Goldman's forecast rests on one key assumption: enterprise adoption keeps accelerating. If corporate wait-and-see sentiment returns, the number could come in lower.
The near-term signal is positive: vendors are voluntarily pushing prices to 30–50% of mainstream levels, lowering the barrier for enterprises to experiment.
This means → whether $13 billion materializes by year-end depends less on further model breakthroughs and more on whether enterprises convert AI from "let's try it" into an everyday tool.
Content is for reference only, not financial advice.