AI pricing and infra are becoming board‑level costs
OpenAI introduced a $100/month ChatGPT Pro plan aimed at heavy coding users while competitors like Anthropic press on infrastructure moves such as custom chips and rapid compute expansion, signalling more explicit AI metering and material backend cost differences. The trend toward metered AI pricing and big‑budget infrastructure races will force software buyers to treat AI as a measurable cost line alongside payments and other infra. (techzine.eu) (bleepingcomputer.com) (thenextweb.com) (news.futunn.com)
A year ago, most companies treated artificial intelligence like a software feature. This week, OpenAI started pricing one coding-heavy ChatGPT tier at $100 a month, which is the kind of number finance teams actually model. (community.openai.com) OpenAI said the new $100 Pro tier is meant for heavier Codex use, with 5 times more Codex usage than ChatGPT Plus and 20 times higher limits than Plus for demanding parallel workflows. That is not “premium” pricing in the old software sense; it is metered access wrapped in a subscription. (community.openai.com) Codex is OpenAI’s coding agent inside ChatGPT, and coding agents are expensive because they do long, multi-step work instead of one quick answer. OpenAI’s own 2024 launch post for the earlier $200 ChatGPT Pro plan said stronger models “take significantly more compute” to run. (openai.com) OpenAI’s current pricing pages now make that cost structure more explicit. The Codex pricing page says users on ChatGPT Plus and Pro who hit limits can buy extra credits, which turns “all you can eat” artificial intelligence into something closer to cloud storage overages. (developers.openai.com) Anthropic is moving on the other side of the same equation. On April 6, 2026, it said it signed a new agreement with Google and Broadcom for multiple gigawatts of next-generation Tensor Processing Unit capacity starting in 2027. (anthropic.com) A gigawatt is power-plant language, not software language. When an artificial intelligence company talks about multiple gigawatts of capacity, it is telling customers and investors that the bottleneck is no longer just model quality but raw electricity, chips, and data center buildout. (anthropic.com) Anthropic had already said in November 2025 that its expanded Google Cloud deal was worth tens of billions of dollars and would bring well over a gigawatt of capacity online in 2026. That gives a rough scale for what “keeping up” in frontier artificial intelligence now costs. (anthropic.com) The funding numbers are starting to match the hardware numbers. Anthropic said on February 12, 2026 that it raised $30 billion in Series G funding at a $380 billion post-money valuation, and it said the money would fund research, product development, and infrastructure expansion. (anthropic.com) OpenAI is also selling the infrastructure story directly. Bloomberg reported on April 9, 2026 that OpenAI told investors its early push to add computing resources gave it an advantage over Anthropic, and another report citing that investor note said OpenAI had 1.9 gigawatts of computing capacity available in 2025, triple the prior year. (bloomberg.com) (theedgesingapore.com) That changes the buying conversation inside ordinary companies. If one employee can burn through a $100 coding plan and then purchase extra credits, and a vendor’s margins depend on who has cheaper chips and more power, artificial intelligence stops looking like a flat software seat and starts looking like payments, cloud compute, or shipping. (community.openai.com) (developers.openai.com) (anthropic.com) The next fight is not just whose model writes better code. It is which company can turn giant capital spending on chips, power, and data centers into a price that a chief financial officer will approve for thousands of employees. (anthropic.com) (community.openai.com)
