OpenAI chases coders — and hits infrastructure headwinds

OpenAI just raised the price of being a serious ChatGPT coder to $100 a month, and the pitch is simple: pay more, hit fewer limits, keep the coding agent running longer. VentureBeat and CNBC reported the new tier gives 5 times the Codex usage of Plus as OpenAI tries to win heavier developer work from Anthropic. (venturebeat.com, cnbc.com) Codex is OpenAI’s coding agent, which means software workers use it less like a chatbot and more like an extra pair of hands that can write, review, and ship code inside a terminal, an integrated development environment, or OpenAI’s own app. OpenAI’s help pages say Codex is bundled across Free, Go, Plus, Pro, Business, and Enterprise plans, but the new pricing is aimed at people who burn through usage during long sessions. (help.openai.com, openai.com) OpenAI’s own pricing pages show why this is a segmentation move, not a mass-market one. ChatGPT Plus remains the lower paid tier, while the new $100 Pro sits below a separate $200 Pro option that OpenAI says is for “heavy lifting” with even higher limits. (openai.com, help.openai.com) That pricing ladder tells you where the company sees demand: not just casual asking, but sustained work that ties up expensive computing power. OpenAI’s Codex pricing page says users who hit plan limits can buy extra credits, which is the software equivalent of an airline selling more legroom to frequent flyers. (developers.openai.com, venturebeat.com) At the same time OpenAI is selling more access to software, it is slowing down one of the hardest parts of the artificial intelligence business: buildings full of chips, power gear, and cooling equipment. Reuters, Bloomberg, and other outlets reported on April 9 that OpenAI paused its main United Kingdom data-centre project because of high energy costs and an unfavorable regulatory environment. (money.usnews.com, bloomberg.com, aol.com) The United Kingdom pause is tied to Stargate, the name OpenAI uses for its data-centre push. The Telegraph reported Sam Altman first unveiled the broader $500 billion Stargate investment program in January 2025, so the company is now trying to expand a giant infrastructure plan while pulling back from at least one flagship site. (telegraph.co.uk, bloomberg.com) Then came the personnel wobble. The Information reported that Peter Hoeschele, an OpenAI executive who helped launch the original Stargate team, left the company, and a separate report from the same outlet said three senior executives tied to the initiative had left or were preparing to leave. (theinformation.com, theinformation.com) That matters because software subscriptions can be changed with a pricing page, but data centres need land, permits, electricity contracts, construction partners, and political cover. DatacenterDynamics reported last month that OpenAI had already reorganized infrastructure leadership under new head of computing and infrastructure Sachin Katti before this latest round of departures. (datacenterdynamics.com, theinformation.com) There is also a legal flank opening in the United States. USA Today reported Florida Attorney General James Uthmeier said on April 9 that his office formally opened a probe into OpenAI and ChatGPT after the Florida State University shooting, raising public-safety concerns tied to the accused gunman’s alleged use of the chatbot. (usatoday.com) So in one 24-hour stretch, OpenAI did three very different things at once: it pushed harder into paid coding work, hit the brakes on a United Kingdom data-centre buildout, and picked up a new state-level investigation in Florida. The common thread is cost: compute is expensive, infrastructure is slower than software, and every extra power user makes the supply problem more visible. (venturebeat.com, money.usnews.com, usatoday.com)