On May 31, 2026, the promotional discount on DeepSeek V4 Pro's API expires. But DeepSeek has announced it won't be rolling prices back. The 75% discount is being locked in as the permanent list price. For a service that was paying 1 million won (about $730) a month in API charges, the same usage now costs 250,000 won (about $180). This isn't a hypothetical. It lands on the first June invoice.

On Hacker News, the developer forum, the announcement drew 244 points and 145 comments. What's striking is that a good share of the comments weren't celebrating the discount itself — they were from people who had already started redesigning their own service architectures around this price. The baseline for AI API costs has, in effect, dropped another notch.

The Promotion Ended, but the Price Didn't Go Up

From the start, the discount on DeepSeek V4 Pro's API pricing was announced as a temporary measure. The expiration was spelled out clearly: 15:59 UTC on May 31, 2026. Most users assumed prices would revert to normal after that point.

In mid-May, DeepSeek announced something different through its official documentation and social channels. Even after the promotion ended, it said, the quarter-price rate would be formally fixed as the permanent list price.

Put that number next to the other major models and its position comes into focus. As of May 2026, OpenAI's GPT-4o costs $5 per million input tokens and $15 per million output tokens. Anthropic's Claude Sonnet line runs about $3 for input and $15 for output. Google's Gemini 1.5 Pro offers a free tier up to a certain usage level and charges beyond it. In a head-to-head comparison, DeepSeek V4 Pro's permanent fixed price sits markedly below all of them.

On performance, DeepSeek V4 Pro posts numbers that compete with the top models from OpenAI and Anthropic on public benchmarks for coding, math, and complex reasoning. The price came down at the same moment the model landed near the top of the performance charts.

Behind DeepSeek's ability to hold this price is a difference in architecture. When DeepSeek R1 launched in early 2025, there were claims that its training costs were a fraction — tens of times lower — than OpenAI's. The exact figures were never independently verified. Even so, the strategy of building price competitiveness on a low-inference-cost structure has stayed consistent. With Google, Amazon, and Meta all steadily cutting the prices of their own AI services, DeepSeek's latest move reads as a bid to be the first to nail down the floor price in a competitive field.

Draw in API users with low prices, get them to build their services on the assumption of your platform, and you create a structure where they can't easily walk away even if prices rise later. That's why fixing the price is more than a simple extension of a promotion.

The Concerns Behind the Low Price

There are reasons this change is hard to greet with unqualified enthusiasm.

DeepSeek is a company headquartered in China. There isn't enough transparency about which servers data passes through via the API, or how it's handled once it gets there. Some public institutions in Europe and some companies in the United States have already restricted DeepSeek's use internally or issued guidance to steer clear of it. How this API can be applied under Korea's Personal Information Protection Act, and under the data-handling rules of the finance and healthcare sectors, is a question that calls for separate legal review.

The pattern of capturing a market with low prices and then changing the terms later has recurred across the cloud industry. AWS attracted startups early on with low-price policies and free tiers, then adjusted prices in stages once dependence on the platform had set in. When a service that leans heavily on the cloud meets a price increase, switching costs make it hard to move. There's currently no guarantee DeepSeek won't walk the same road.

Service availability needs checking too. In 2025, as DeepSeek's service grew explosively, there were stretches when the API became unstable under surging traffic. Adopting it without directly testing real latency and availability in a production environment is a risk in its own right.

These concerns don't carry equal weight in every situation. The size of the risk shifts with the nature of a service's data and its regulatory environment. Skipping the review just because concerns exist, and adopting without verification just because the price is low, each generate costs in their own different way.

It's Time to Pull Out the Invoice

The people for whom this price change actually matters are those already wiring AI APIs into real work. If you're a solo operator in Korea, a solo PM, or a middle manager driving AI adoption, there are things to check right now.

Start by working out which APIs you're using, and at what volume. More often than you'd expect, this gets lumped together as a “ChatGPT subscription” or an “AI tool cost.” Without knowing, line by line, which model you're using for which purpose, it's hard to even explore alternatives. Pulling out the last three months of invoices and breaking them down item by item is the starting point. For example, if you're handling both customer-email summaries and social-media draft posts with the same model, you can first calculate the cost difference of peeling off the repetitive, less complex work — drafting, say — onto a cheaper model.

Not every task needs the highest-performance model. Teams have reported cutting costs by 30 to 60 percent by handling repetitive work — simple classification, keyword extraction, draft summaries — with cheaper models, and reserving the expensive ones only for complex reasoning or the final, customer-facing output. With prices down, now is also the moment to redesign that structure.

To decide whether to adopt DeepSeek, you first need to sort out the nature of the data you handle. Internal workflow automation, content generation based on public information, and processing that involves no personal data all carry relatively low risk. If you're handling customer personal data directly, are contractually bound to guarantee where data is processed, or operate in a heavily regulated industry like finance or healthcare, legal review has to come first.

When smartphones first went mainstream, the teams that explored the market and designed services in the first two or three years after the App Store opened ended up well positioned in the competition that followed. The same thing played out when cloud services became a real option for startups. The people who got ahead in these moments had something in common: they had built up hands-on experience experimenting with exactly how the new environment connected to their own way of working. Simply knowing how to use a technology, and judging which technology to connect — when and how — are entirely different capabilities.

The fact that AI API costs are now locked at a quarter is a signal that the window for making that judgment has opened again. If you're working without a grasp of your own tooling cost structure, that blind spot is quietly draining out of your invoice every month.