Over the past few years, the market has been operating in a bubble of subsidized innovation. The shift from simple chatbots to autonomous AI agents that can code on their own, analyze thousands of pages of documentation and reason in the background has dramatically driven up inference costs. Providers have realized they can’t sustain profitability on rigid subscriptions.<\/p>
The clearest signal of this transformation was Anthropic’s decision: the creators of Claude removed the pools of free tokens from corporate subscriptions. The change began as early as November 2025, when Anthropic started renewing contracts under new terms, and in February 2026 a uniform model was introduced: a flat monthly fee of $20 per workstation (seat), with each token consumed by an employee billed separately at API rates.<\/p>
For Claude Opus 4.6, it’s $5 per million input tokens and $25 per million output tokens. The cost risk has been shifted onto businesses, forcing CFOs to monitor how much computing power employees use.<\/p>
A different, yet equally costly path for customers was chosen by OpenAI. Instead of pricing tokens, the company began pricing the operating environment in which the model performs tasks, for example, compiling code. As of March 31, 2026, developers pay $0.03 for each gigabyte of container capacity for every 20 minutes of a session. In the B2B sector, AI has become a virtual employee billed at an hourly rate.<\/p>
The effects of rising infrastructure costs have hit ordinary users just as hard. The classic $20 plan was quietly downgraded. Providers implemented request limits, causing users working on large documents to exhaust their quotas within a few hours. In February 2026, OpenAI introduced ads in the Free and Go plans.<\/p>
The first half of 2026 saw the birth of a new class of services, tailored to so-called super-consumers: freelancers, researchers and engineers.<\/p>
OpenAI<\/strong> rolled out a steep price hike: from $20 (Plus) straight to $200 (Pro). On April 9, 2026, the company filled the gap with a new $100-per-month tier, explicitly stating that it did so in response to Anthropic\u2019s offering. Both Pro tiers provide access to the same set of models, including the exclusive GPT-5.5 Pro. The difference comes down to limits: the $100 plan offers 5x the limits of Plus; the $200 plan offers 20x.<\/p> Anthropic<\/strong> maintained the structure that OpenAI copied. Claude Max plans at $100 or $200 per month offer, respectively, 5x or 20x higher usage limits than the $20 Pro plan.<\/p> Google<\/strong> has gone a step further, integrating the most powerful features into the Google AI Ultra plan at $249.99 per month. The bundle includes access to the Gemini 3.1 Pro Deep Think model, the Veo 3.1 video generator, and 30 TB of cloud storage. The gap between the Pro plan ($19.99) and Ultra ($249.99) is so large that Google is already working on an intermediate pricing tier, internally called “AI Ultra Lite,” most likely around $100<\/p> The way they handle limits reveals the customer retention strategies of the various players. When you hit the limit in OpenAI, the system forcibly switches you to a weaker model or forces you to upgrade. Google completely locks the user into its ecosystem: there\u2019s no option for a one-time purchase of additional queries\u2014the only path is to upgrade the entire Google One plan.<\/p> Price movements point to a new market logic. Language models themselves are becoming commoditized. The giants no longer make money solely from access to the algorithm, but from the surrounding infrastructure.<\/p> Google and Microsoft are monetizing vendor lock-in: AI models are so deeply integrated with office suites that, for enterprises, the cost of switching to a competitor is higher than accepting a price increase. In the free tier, pressure is mounting to introduce native ads: OpenAI has already been testing this model in the Free and Go plans since February 2026, and the company has acknowledged that the current pricing structure isn’t sustainable.<\/p> The golden age of cheap, subsidized AI is over. 2026 pricing makes it clear: the market will split between users on pared-back plans with a growing number of ads and those who will pay multiples of the original $20 threshold for full functionality.<\/p>","protected":false},"excerpt":{"rendered":" The first half of 2026 is the moment when Silicon Valley stops subsidizing access to artificial intelligence. The magical $20-per-month barrier is no longer a flat rate for everything, and the AI market has entered a phase in which digital intelligence is paid for like electricity usage.<\/p>\n","protected":false},"author":465,"featured_media":18358,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"rank_math_lock_modified_date":false,"footnotes":""},"categories":[888,832],"tags":[],"popular":[],"difficulty-level":[38],"ppma_author":[892],"class_list":["post-18423","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-business-2","category-editors-picks","difficulty-level-medium"],"acf":[],"authors":[{"term_id":892,"user_id":465,"is_guest":0,"slug":"kmironczuk","display_name":"Krzysztof Miro\u0144czuk","avatar_url":{"url":"https:\/\/haimagazine.com\/wp-content\/uploads\/2025\/10\/awatar-2.png","url2x":"https:\/\/haimagazine.com\/wp-content\/uploads\/2025\/10\/awatar-2.png"},"first_name":"Krzysztof","last_name":"Miro\u0144czuk","user_url":"","job_title":"","description":"Od lat zajmuj\u0119 si\u0119 nowymi technologiami w biznesie, edukacji i codziennym \u017cyciu. W centrum mojej uwagi pozostaje cz\u0142owiek \u2013 i to, by technologia wyr\u00f3wnywa\u0142a szanse, zamiast tworzy\u0107 bariery."}],"_links":{"self":[{"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/posts\/18423","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/users\/465"}],"replies":[{"embeddable":true,"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/comments?post=18423"}],"version-history":[{"count":1,"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/posts\/18423\/revisions"}],"predecessor-version":[{"id":18424,"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/posts\/18423\/revisions\/18424"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/media\/18358"}],"wp:attachment":[{"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/media?parent=18423"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/categories?post=18423"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/tags?post=18423"},{"taxonomy":"popular","embeddable":true,"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/popular?post=18423"},{"taxonomy":"difficulty-level","embeddable":true,"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/difficulty-level?post=18423"},{"taxonomy":"author","embeddable":true,"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/ppma_author?post=18423"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}AI as a public utility?<\/h4>