As DeepSeek makes the temporary 75% discount on its highest-performing AI model, V4 Pro, permanent, the era of a full-scale AI price war has begun, allowing anyone to utilize powerful artificial intelligence at an overwhelmingly low cost.
Imagine this: a top-tier, 3-star Michelin restaurant that you only visit on special occasions suddenly hangs a huge banner in front of its store. “Starting today, we are offering a permanent 75% discount on all our premium course meals!” This means you can enjoy top-grade steak and caviar, which used to cost hundreds of dollars, every day at the price of a local diner. Such an unprecedented event, which might force a restaurant to close its doors due to unbearable deficits, is actually happening right now in the world’s fiercest and smartest artificial intelligence (AI) market.
Recently, the Chinese artificial intelligence startup DeepSeek announced that the 75% discount on its flagship AI model, ‘V4 Pro’, will be cemented as a ‘permanent regular price’ rather than a limited-time event DeepSeek permanently reduces the price of its flagship V4 model by 75 percent - Engadget. This is a massive bombshell that is shaking the foundation of the industry. What on earth is happening in the invisible AI market? And how is this seemingly complex tech news related to our daily lives, where we use countless smartphone apps every day? From now on, we will unravel this immense wave of change in the easiest language possible.
Why It Matters
When we converse with an AI assistant on a smartphone app or summarize a foreign language document in one second, the developer who created that app is quietly paying an ‘intelligence usage fee’ behind the scenes to massive AI companies (like Google, OpenAI, Anthropic, etc.). If this intelligence usage fee is expensive, developers have to ask you for a costly monthly subscription fee to maintain the service, or if it is a free app, they have to cram it with ads that cover half the screen. In other words, the cost of borrowing AI’s smart brain is the ‘raw material cost’ of all the innovative digital services that permeate our lives.
| The reason DeepSeek’s latest announcement has sent massive shockwaves through the global IT industry is that the price tag for the ‘premium intelligence’ they offer has literally broken through the floor and gone underground. DeepSeek’s permanent discount is a bold move that completely freezes the existing price at a staggering one-quarter of the original level [China’s DeepSeek to make permanent 75% price cut on flagship V4‑Pro AI model | Reuters](https://www.reuters.com/world/china/chinas-deepseek-make-permanent-75-price-cut-flagship-v4pro-ai-model-2026-05-23/). When compared to the top-tier models of other global Big Tech companies, this difference becomes dizzyingly stark. Compared to ‘Claude Opus 4.7’, currently considered one of the smartest AIs in the industry, DeepSeek’s V4 Pro is about 34 times cheaper in terms of the cost of reading your questions or data (input), and a whopping 86 times cheaper in terms of generating text (output) DeepSeek V4 Guide 2026: Pro, Flash, Pricing & Benchmarks. |
Let’s use an easy analogy. Imagine taking a comfortable taxi ride from New York to Washington D.C. If using a competitor’s AI is like riding in a ‘luxury premium taxi’ where the meter constantly ticks up, using DeepSeek’s equally comfortable and fast AI is like paying a mere ‘local bus fare’ for a high-speed journey. When the cost of borrowing intelligence becomes this extremely cheap, even broke college founders and small startups can freely integrate top-tier AI into their brilliant ideas without worrying about expenses.
The Explainer
Let’s dig a little deeper into three core technological changes embedded in DeepSeek’s unprecedented announcement, comparing them to the ‘coffee’ and ‘cooking recipes’ we encounter every day.
1. Token: The ‘Word Puzzle Piece’ for Conversing with AI
| In the AI market, service fees are calculated using a unit called a ‘Token’ (the basic data unit for AI reading and writing text). You can think of it as fragments of letters or words used when we speak to AI or when AI responds to us. Due to this permanent discount, the cost charged when DeepSeek V4 Pro generates an answer (output token) has dropped to a mere $0.87 per 1 million tokens DeepSeek made its 75% discount permanent. The AI price war just escalated.. This means it costs less than 1 dollar for the AI to write a massive amount of text equivalent to several thick English novels on your behalf. Even the price of the API (Application Programming Interface, a pathway that connects programs) used by app developers to feed information into the AI has dropped below 50 cents per 1 million tokens [DeepSeek V4-Pro 75% Price Cut 2026 | Complete Cost Analysis …](https://promptcost.org/en/blog/deepseek-v4-pro-price-cut-analysis-2026/). |
2. Cache-hit Cost Reduced to One-Tenth: “Special Treatment for Regular Customers”
In a previous update, DeepSeek introduced another amazing system to save costs. They drastically lowered the ‘input Cache-hit’ cost to one-tenth of its original level DeepSeek Slashed V4-Pro Pricing by 75%: Why Cache Costs ….
To easily understand a ‘cache-hit’, imagine a sensible barista at your local favorite cafe. There is a customer who stops by every morning on their way to work and orders, “Iced Americano with an extra shot and lots of ice.” One day, just by hearing the ‘ding’ of the door opening and seeing the customer’s face, the barista starts making that customized coffee right away. Time and energy are enormously reduced because the brain activity (computation) of listening to the new order and nodding in understanding is skipped every single time. AI is exactly the same. Once a developer has the AI thoroughly read a company manual spanning hundreds of pages or a thick law book, the next time a question is asked based on the same data, the AI instantly pulls out the memory (cache) it has already stored in its mind. DeepSeek slashed the fee charged for “reusing already known information” by an astonishing 90%. Thanks to this, companies can operate smart chatbots with very low maintenance costs without having to expensively input vast amounts of data over and over again.
3. Weights Released under MIT License: “Free Distribution of a $10 Million Secret Recipe”
The reason DeepSeek’s moves come across as both true terror and joy to the industry is that they don’t stop at merely selling ‘finished dishes (AI services)’ cheaply. They have released the ‘Weights’, the core brain of their heavily funded highest-performing model, V4 Pro, to the world under the MIT License. This has made self-hosting possible, where anyone can install the AI directly on their own computer or company server DeepSeek V4 Guide 2026: Pro, Flash, Pricing & Benchmarks.
In terms of cooking, AI ‘weights’ are like the ‘ultimate special sauce recipe miraculously perfected after thousands of trials and errors, spending millions of dollars on ingredients’. Massive electricity bills and expensive computer chip costs are consumed to create this recipe. However, DeepSeek distributed this precious trade secret under the unprecedented condition of the MIT License, which essentially means, “Whether you make money commercially or modify the recipe to suit your taste, use it completely free of charge.” It is no different from Coca-Cola sharing its secret formula with the entire world for free. Small startups and underfunded labs without massive capital can now download the world’s top-tier AI brain entirely for free and easily customize it into their own bespoke services.
Where We Stand
| This unbelievable bombshell discount was originally intended as a ‘limited-time promotion’ scheduled to run only briefly until May 31, 2026 [DeepSeek to make a 75% permanent discount on V4 Pro AI model | Seeking Alpha](https://seekingalpha.com/news/4596488-deepseek-make-75-permanent-discount-v4-pro-ai-model). Upon seeing this announcement, many industry experts asserted, “No matter how fierce DeepSeek’s momentum is, if they offer top-tier AI at such a dirt-cheap price, they won’t be able to handle the server maintenance costs and will quietly raise the price soon.” |
| However, defying all expectations, around May 22, DeepSeek drove a wedge through its official website and announcements that it would solidify this 75% discount not as a temporary event but as a ‘permanent regular price’ DeepSeek가 V4 Pro 가격 할인을 영구화함 — 개발자가 알아야 할 핵심 …. Through various official communication channels of the company, it was clearly reaffirmed that this is not a simple marketing bait product [DeepSeek Makes 75% V4-Pro API Cut Permanent | AI Weekly](https://aiweekly.co/alerts/deepseek-makes-75-v4-pro-api-cut-permanent). |
What surprised the market even more was the timing of this decision. Just a little over a month after the V4 model, DeepSeek’s new generation that the whole world had been waiting for so long, was released to the world, this permanent discount was abruptly executed DeepSeek V4 Pro tops global bang-for-buck ranking after 75% …. To use a car company analogy, it’s like announcing, “We’ll sell it for a quarter of the price for life,” just a month after ambitiously launching a top-of-the-line new sedan model. Hearing this news, the community of millions of app developers worldwide was swept up in an absolutely enthusiastic festival atmosphere.
What’s Next
| With this single massive stone thrown by DeepSeek, the fierce ‘AI Price War’ among the world’s leading AI companies has entered an entirely new brinkmanship phase DeepSeek makes 75% price cut permanent on V4-Pro as AI price war enters new phase. Influential global financial media such as Bloomberg and Seeking Alpha evaluate that the unilateral decision made by this young Chinese startup has dragged the service pricing baseline of the entire global AI market completely to rock bottom [DeepSeek makes 75% discount permanent on V4-Pro | Let’s Data Science](https://letsdatascience.com/news/deepseek-makes-75-discount-permanent-on-v4-pro-5e78d825). Competitor Big Tech companies that have been charging expensive fees until now are caught in a dilemma: they must either prove why their expensive services are worth that much with far more overwhelming performance, or bite the bullet and drastically lower their service prices following DeepSeek. |
Then, as many question, what magic did DeepSeek use to be able to operate the company stably while slashing prices like this? Providing AI services that answer the questions of hundreds of millions of people around the world requires an unimaginably large number of expensive computer chips (semiconductors) running non-stop. In this regard, some analysts suggest that DeepSeek has drastically lowered hardware costs because it has been able to smoothly secure a massive supply of ‘Ascend 950’, the latest AI chip developed by China’s Huawei. Of course, DeepSeek is neither confirming nor denying whether this massive price cut is truly due to the stable supply increase of Huawei chips, holding back from an official confirmation DeepSeek makes 75% price cut permanent on V4-Pro as AI price war enters new phase.
Whatever the real secret to cost reduction hidden behind it may be, from the perspective of general consumers using the services, there is nothing left but to cheer. If the wholesale price for developers borrowing smart intelligence plummets, soon top-tier AI surpassing human capabilities will very naturally, and in a nearly ‘free’ form, deeply permeate almost everywhere in our daily lives—the smartphone apps we touch every day, highly immersive video games, customized educational software for children, and complex business tools for companies. The sci-fi movie-like future where an invisible army of smart personal assistants that perfectly understand my mind assists my life from the moment I wake up until I fall asleep has approached us at least a few years faster thanks to this price war.
AI’s Take
As with powerful technological revolutions of the past, cutting-edge artificial intelligence is also rapidly descending from the position of an ‘expensive luxury enjoyed only by a few’. This radical price war triggered by DeepSeek clearly demonstrates that AI is evolving beyond a simple novelty toy into a true ‘public infrastructure’ that anyone can borrow cheaply and naturally, just like electricity comes on and water pours out when you turn on the faucet.
The fact that the cost of using intelligence is becoming so cheap that it converges to zero means that the barrier to imagination is collapsing. Countless solo creators and small-to-medium-sized businesses who had great ideas but hesitated to take on challenges due to the fear of hundreds of thousands of dollars in server maintenance costs can now unhesitatingly mount world-class AI brains onto their services.
In the mud-slinging price competition of fiercely battling companies, it is ultimately us ordinary people who reap the biggest benefits. We are already looking forward to the amazing and heartwarming everyday innovations that will pour out in a world where the technological entry barrier has been lowered.
References
- DeepSeek permanently reduces the price of its flagship V4 model by 75 percent - Engadget
-
[China’s DeepSeek to make permanent 75% price cut on flagship V4‑Pro AI model Reuters](https://www.reuters.com/world/china/chinas-deepseek-make-permanent-75-price-cut-flagship-v4pro-ai-model-2026-05-23/) - DeepSeek makes 75% price cut permanent on V4-Pro as AI price war enters new phase
-
[DeepSeek to make a 75% permanent discount on V4 Pro AI model Seeking Alpha](https://seekingalpha.com/news/4596488-deepseek-make-75-permanent-discount-v4-pro-ai-model) - DeepSeek made its 75% discount permanent. The AI price war just escalated.
- DeepSeek V4 Guide 2026: Pro, Flash, Pricing & Benchmarks
-
[DeepSeek V4-Pro 75% Price Cut 2026 Complete Cost Analysis …](https://promptcost.org/en/blog/deepseek-v4-pro-price-cut-analysis-2026/) - DeepSeek V4 Pro tops global bang-for-buck ranking after 75% …
- DeepSeek Slashed V4-Pro Pricing by 75%: Why Cache Costs …
- DeepSeek가 V4 Pro 가격 할인을 영구화함 — 개발자가 알아야 할 핵심 …
-
[DeepSeek Makes 75% V4-Pro API Cut Permanent AI Weekly](https://aiweekly.co/alerts/deepseek-makes-75-v4-pro-api-cut-permanent) -
[DeepSeek makes 75% discount permanent on V4-Pro Let’s Data Science](https://letsdatascience.com/news/deepseek-makes-75-discount-permanent-on-v4-pro-5e78d825)
- 25%
- 50%
- 75%
- Reduced by half (1/2)
- Reduced to one-tenth (1/10) of the level
- Converted to free
- Private proprietary license
- MIT License
- Educational restricted license