Goldman Sachs: DeepSeek's Peak-Hour Price Hike Signals a Return to Rationality in LLM Competition

Claire Weston
Published todayAbout 7 min read

DeepSeek doubled its peak-hour API prices, and Goldman Sachs reads the move not as weakening demand but as China's large-model industry shifting from a cash-burning price war toward cost-reflective pricing under tightening compute supply.

01

How much did DeepSeek actually raise prices?

Peak hours are set at 9 am–12 pm and 2 pm–6 pm; API prices during those windows double.
Non-peak pricing for V4 Pro and Flash stays unchanged — the hike targets only the busiest slots.
Goldman estimates the blended average lands at roughly $0.35 and $0.12 per million tokens. This means → the increase is structurally concentrated in peak hours, not a flat doubling of the total bill.
02

Why does Goldman call this a positive signal, not a warning?

Goldman states explicitly: the price hike does not reflect weakening demand. The opposite — strong demand and tightening compute gave DeepSeek the leverage to raise prices.
This reflects an industry-wide shift from "zero-margin or loss-making" aggressive pricing toward cost-reflective, rational pricing.
In plain terms = providers had been undercutting each other to the point of losing money on every API call. DeepSeek is the first to say "we stop subsidizing" — inference has a real cost, and the market is starting to acknowledge it.
03

Where is the competitive focus shifting?

Goldman notes rising industry concentration; competition is moving from "whose model scores highest" toward sustained product iteration and real commercial deployment.
Case in point — MiniMax. Its management argues its edge lies in lean organization, high infrastructure utilization, and fast iteration: when OpenClaw gained traction, it quickly launched MaxClaw and rolled out MiniMax Code in parallel.
This means → the winners going forward may not be the largest companies, but the ones that can turn every dollar of compute investment into sustained returns.
04

What to watch next?

The key validation point: whether other LLM providers follow DeepSeek's price increase.
If major players collectively raise prices, it confirms the industry pricing framework is being restructured — and the price-war era is truly over.
In plain terms = DeepSeek fired the first shot, but one company raising prices is not an industry turning point. The inflection is confirmed only when competitors follow suit, signaling the whole sector's competitive logic has changed.

Content is for reference only, not financial advice.

Goldman Sachs: DeepSeek's Peak-Hour Price Hike Signals a Return to Rationality in LLM Competition · nashnova