OpenAI shifts raise vendor risk
OpenAI announced a permanent London office and expansion, while the company has recently shelved products like Sora and pushed faster, cheaper endpoints such as GPT‑5.3 Instant Mini — moves that highlight how quickly vendor roadmaps can change. Those shifts are being framed as a reminder that enterprise AI strategies should manage dependency on single vendors rather than assuming product stability ((reuters.com), (futurumgroup.com), (futurumgroup.com)).
OpenAI said on April 13 it signed its first permanent London office, adding space for more than 500 staff in a city where it now employs about 200. (cnbc.com) The new site is an 88,500-square-foot office in King’s Cross, and OpenAI has said London will become its largest research hub outside the United States. Reuters, carried by RTÉ, reported the office is due to open in 2027. (cnbc.com, rte.ie) The expansion comes days after OpenAI paused a large United Kingdom data-center project, while continuing to add people and office capacity in Britain. The Independent reported the company plans to more than double its British workforce from current levels. (independent.co.uk, cnbc.com) At the same time, OpenAI has been changing its product lineup quickly. Futurum reported on April 12 that OpenAI will shut down the Sora web and app products on April 26, 2026, with the Sora application programming interface scheduled to end on September 24, 2026. (futurumgroup.com) OpenAI has also been pushing lower-latency, lower-cost models. OpenAI’s developer documentation says GPT-5.3 Chat now points to the GPT-5.3 Instant snapshot used in ChatGPT, with a 128,000-token context window and pricing of $1.75 per million input tokens and $14 per million output tokens. (developers.openai.com) In a consumer-facing post last month, OpenAI said GPT-5.3 Instant is an update to ChatGPT’s most-used model and described it as more accurate and more useful for web search. Futurum wrote on April 12 that OpenAI also introduced GPT-5.3 Instant Mini in ChatGPT alongside new Pro plan options. (openai.com, futurumgroup.com) Those moves leave enterprise buyers dealing with two kinds of change at once: expansion in places where OpenAI wants to hire, and turnover in the models and products customers may have built around. Futurum’s April survey write-up said 67% of organizations already run generative artificial intelligence in production and 75% expect to increase budgets, making migrations and rewrites more expensive when roadmaps shift. (futurumgroup.com, futurumgroup.com) OpenAI’s pitch is that faster and cheaper models widen adoption, while its London hiring shows it is still investing heavily in research and sales capacity. The counterpoint from analysts is that companies using any single model provider need backup plans for discontinued products, changed pricing, and renamed or redirected endpoints. (openai.com, futurumgroup.com, cnbc.com) For now, the clearest signal is that OpenAI is growing in London even as parts of its catalog are being retired or rerouted. Customers that treated one vendor’s roadmap as fixed now have fresh dates, prices, and shutdown notices to plan around. (rte.ie, developers.openai.com, futurumgroup.com)
