Behind DeepSeek’s Usage Limits: Free AI Is Saying Goodbye to the Unlimited Model
DeepSeek has tightened limits on long conversations and daily usage in its app and web version for regular users, prompting dissatisfaction among heavy users. The article argues that computing and long-context costs are pushing AI platforms to redefine free services and shift toward models based on basic allowances, paid add-ons, or advertising subsidies.
Why read this
AI score 63
Starting with usage limits, paid tiers, and advertising models, the article explains how inference costs are driving free AI services away from unlimited use toward basic usage allowances.
Topics
Body · AI translation
By Zinc Scale | Author: Meng Huiyuan | Editor: Chen Dengxin
“I’ve already exchanged 150,000 to 200,000 Chinese characters with DeepSeek. I don’t want to lose this conversation.”
“Teacher D, don’t abandon me! As someone severely dependent on DeepSeek who spends more than ten hours a day on it, how am I supposed to live without Teacher D? I really want things to go back to the way they were.”
“What am I supposed to do? I have moderate depression and mania, and for various reasons I can’t see a doctor. I’ve always treated DeepSeek as a mental-health therapy app. Switching to the API isn’t out of the question, but that chat window of mine contains more than 2,000 messages…”
In the days since DeepSeek imposed usage limits, a tidal wave of public reaction has surged forth.
The first to be overwhelmed were those who treated “Teacher D” as a confidant, a counselor, or even the object of a romantic fantasy. They gave the AI a name, crafted a persona for it, and set its tone, then continued talking in a single chat window for months, with the persona, backstory, and development of the relationship all accumulating in the context. For these users, limiting conversations and terminating them after seven days of inactivity was not a product update, but a unilaterally announced “downgrade of the relationship.”
They repeatedly posted emotional pleas and flooded related comment sections, hoping the team behind DeepSeek would notice their needs. Yet what happened next was rather intriguing: a sizable number of people accustomed to enjoying “Teacher D’s” free services began actively begging DeepSeek to charge them. They said that as long as the price was reasonable, they would be willing to pay for services such as “resuming extra-long conversations,” “removing the seven-day inactivity restriction,” and “higher priority for computing resources.”
When DeepSeek first burst onto the scene, the entire internet cheered its generosity in offering the service for free. Now that it is proactively withdrawing those benefits, what is ending, in a sense, is not the “free” label itself, but the old premise that “capital subsidies can flatten computing costs indefinitely,” which once made “free and unlimited” usage possible.
Judging from user feedback, the changes brought about by DeepSeek’s usage limits boil down to three points: first, old conversations with no interaction for more than seven days can no longer be continued; second, the maximum length of a single input has been reduced to about 40,000 Chinese characters; and third, daily usage is capped. Once the quota is exhausted, users are told, “The server is busy and your usage limit has been reached,” and must wait for access to be restored.
Reports suggest that DeepSeek may have had little choice. According to OpenRouter, in the week ending November 15, DeepSeek V4.1 Flash 1017 processed 59.2 trillion tokens, ranking first worldwide, while the preceding model in the same series handled 15.8 trillion tokens per week two months earlier. Incident records on its official status page also show that DeepSeek has experienced multiple service failures, most involving degraded performance. Specific symptoms have included slower responses, high generation latency, and occasional request failures, although users still had some chance of submitting conversations. In more serious cases, parts of the website or API were even disrupted.
At the technical level, long conversations continually accumulate KV cache—the intermediate data kept in GPU HBM memory so the model can maintain context and avoid redundant computation—which means greater resource consumption. The longer a conversation persists and the longer its context becomes, the larger the cache grows and the more scarce HBM capacity it occupies. Long-idle conversations no longer generate new interactions yet still effectively “hold on to GPU memory,” a classic misallocation of resources when computing capacity is tight. For DeepSeek, reclaiming the cache occupied by these idle conversations allows the freed HBM capacity to be reassigned to active users who are currently using the service and continuously generating inference requests, easing queues and slowdowns during traffic surges.
But when this technical logic is applied to actual people—the current restrictions mainly target the consumer-facing DeepSeek app and website—it becomes another matter entirely.
For DeepSeek’s heaviest users, represented by those engaged in “human-AI romance,” their usage intensity naturally approaches the quota ceiling, making the impact undoubtedly more acute: long conversations, high frequency, and deep context all run directly into the sharp edge of the limits. Thus, a rather absurd chorus of “tearful pleas” and “petitions” began spreading across social platforms. Some shared screenshots of conversations that had been cut off and begged the company to withdraw the usage limits and other new rules; some advertised everywhere and simply started businesses charging people to help them connect to the API; while others posted frequently online to declare their willingness to pay for DeepSeek’s services.
Over the past two years, users’ expectation of “free AI” has not been merely “free to use,” but “free to use however much they want.” That expectation was shaped by the previous battle of capital subsidies, which welded the idea of “unlimited usage” to the concept of “free.” But subsidies have never been the norm for a business model; they are merely temporary fuel during the user-acquisition stage. Once the user base crosses a critical threshold and usage begins doubling week by week, platforms must choose between “continuing to burn money in exchange for growth” and “making the unit economics at least marginally viable.”
Computing Costs Cannot Be Flattened, and the Entire Industry Is Paying the Price
Clearly, the word “free” is being redefined in the context of AI. In the past, it meant “unrestricted use without payment.” Now, it is closer to “use within a basic allowance without payment, with usage beyond that available on demand for a fee or after waiting in a queue.”
Take users of emotional companionship and role-playing scenarios, for example. They often maintain dozens of chat windows simultaneously, loading each with tens of thousands of Chinese characters of settings and plot, while a single conversation may run to hundreds or even thousands of turns. Some regenerate responses more than a thousand times to keep the character persona consistent. On any platform that charges by the token, this usage pattern would result in a substantial bill. Under a free allowance, it becomes a sunk cost borne solely by the platform, which has neither the capacity nor the obligation to permanently underwrite usage at this intensity.
The shakeout in response to industry changes has already begun. According to TMTPost, citing IT Juzi data, more than 10 of a sample of roughly 200 pure-play API startups had ceased operations or pivoted in Q1 2026. Another commentary article said that, during the same period, seed-stage financing in China’s AI sector fell 11% year over year, while more than 300 small and midsize AI companies shut down because their cash flow dried up.
After all, the rise in costs is real. SemiAnalysis data shows that the price of a one-year H100 contract rose from a low of $1.70 per hour in October 2025 to $2.35 in March 2026, an increase of nearly 40% in five months. By early October, the transaction price had reached $2.80, for a cumulative increase of about 65%. Leading vendors, with long-term contracts and their own computing capacity, are significantly less exposed to spot-market fluctuations. The companies with truly nowhere to pass on the costs are pure-play API providers with no traffic ecosystem, no stockpile of computing capacity, and no ability to offer private deployments.
The real lesson of the usage-cap controversy may be that the entire industry needs to recalibrate the narrative of “free.” Previously, platforms traded free access for scale, while users traded scale for a better experience; both sides got what they needed, and everyone was happy. But faced with the reality that subsidies cannot flatten computing costs, the strategic shift for the “DeepSeeks” of the world will inevitably be from “scale first” to “retention and unit economics first.”
Take Doubao, for example. Lightweight features such as basic chat, writing, and image generation remain free, while computing-intensive scenarios such as software development, data analysis, and professional design have become paid value-added services. This creates a tiered structure in which lightweight features attract users and heavyweight features generate revenue. In essence, revenue from paid, high-value scenarios is used to cover the computing costs of free, mass-market use. This is the classic cross-subsidy model that the internet industry has employed for years. In the AI era, however, the imbalance may be even more extreme, because the computing consumption of heavy users does not grow linearly but expands exponentially.
DeepSeek’s throttling policy comes at the cost of a short-term erosion of user trust. In the long run, however, if usage caps can deliver more stable service quality and clearer paid tiers, this public-relations storm may prove to be a necessary shakeout—provided that DeepSeek clearly explains the rules governing the caps, defines the boundaries of the free allowance, and gives genuinely high-intensity users a predictable path to paid access. The key question is not whether to impose caps, but what will keep users after the caps are introduced.
OpenAI, meanwhile, is using advertising to give its free tier a source of revenue. According to the company, by August 31, 2026, less than 200 days after launch, ChatGPT Ads had reached an annualized revenue run rate of $1 billion. Ads appear only on the free tier and the lower-priced Go tier, while paid subscription tiers remain ad-free. Yet the advertising route presents obvious challenges as well. Inserting ads into AI conversations is far more difficult than doing so in search results or content feeds, because users expect coherent, immersive interactions, and any interruption may be seen as contaminating the experience. OpenAI has the confidence to pursue this path because its user base is large enough and advertisers are sufficiently willing to pay a premium for “AI-native traffic.” Whether this model can be replicated by small and midsize platforms remains unknown.
Returning to the question users care about most: Is the era of free AI coming to an end? What is ending is not “free,” but “unlimited.” Doubao’s selective monetization, DeepSeek’s usage caps, and OpenAI’s advertising all point to the same structural reality: inference costs grow linearly or even exponentially with usage, while the pace at which computing supply can ramp up falls far short of the rate at which demand is expanding.
This race between costs and demand is pushing the entire industry into a more sober—and more brutal—phase.
Previously, free access was almost regarded as the default price of admission for AI products. Tech giants used it to cultivate user habits, startups used it to build fundraising narratives, and users grew accustomed to AI services that were “unlimited and barrier-free.” But once inference costs move from assumptions in a PowerPoint presentation to tangible expenses on financial statements—once every conversation, every generated image, and every code completion consumes computing power that costs real money—free access ceases to be a business model and becomes a subsidy endurance race.
Players with limited endurance are destined to leave the field first.
AI Trends