Want smarter insights in your inbox? Sign up for our weekly newsletters to get only what matters to enterprise AI, data, and security leaders. Subscribe Now
Today, many eyes are on OpenAI CEO and co-founder Sam Altman‘s ongoing public feud with Elon Musk on the latter’s social network, X.
But Altman’s recent statements regarding the ongoing rollout of his company’s latest and greatest large language model (LLM), GPT-5, are probably more important to customers and enterprise decision-makers.
After an admittedly “bumpy” debut of GPT-5 last week that saw some users clamoring for restored access to deprecated older LLMs in ChatGPT such as GPT-4o and o3 — OpenAI granted the former — Altman is now pivoting towards ensuring OpenAI’s underlying infrastructure and usage limits are a good fit for the company and its 700 million active weekly ChatGPT users.
The company’s latest updates include a more detailed compute allocation plan and the introduction of additional third-party connectors for ChatGPT Plus and Pro plans.
AI Scaling Hits Its Limits
Power caps, rising token costs, and inference delays are reshaping enterprise AI. Join our exclusive salon to discover how top teams are:
- Turning energy into a strategic advantage
- Architecting efficient inference for real throughput gains
- Unlocking competitive ROI with sustainable AI systems
Secure your spot to stay ahead: https://bit.ly/4mwGngO
Managing GPT-5 demand and usage limits
In a post on X last night, Altman outlined how OpenAI will prioritize computing resources over the next several months.
He said the company’s first priority is ensuring that current paying ChatGPT users receive more total usage than they had before GPT-5’s release, though he did not provide specific figures for the increase.
However, Altman previously posted on X that OpenAI was “trying” a 3,000 messages-per-week usage limit when using the GPT-5 “thinking” mode, more reasoning power and time spent reasoning on harder problems, for ChatGPT Plus subscribers (the $20 per month plan).
Interestingly, one report from an AI app creator on X said that OpenAI told him the usage limits for GPT-5 plus thinking on the ChatGPT Team plan ($30 per user per month) is much lower than that of ChatGPT Plus users, only 200 “Thinking” messages per week when selected manually by the user.
OpenAI’s availability of GPT-5 through its application programming interface (API) for third-party developers is also being tweaked.
Altman also stated in his X post that OpenAI would “prioritize API demand up to the currently allocated capacity and commitments we’ve made to customers.”
In other words, existing API users and those already in contract will get the first dibs on GPT-5 access through OpenAI’s API, others may have to wait longer.
Altman also clarified “we can support about an additional ~30% new API growth from where we are today with this capacity,” meaning they can take on more API users, but not too many.
While OpenAI hasn’t definitively shared how many users of its API there are in some time, the company did say it has “5 million” businesses paying for access to ChatGPT.
Altman also said OpenAI plans to roughly double its compute fleet over the next five months. He did not specify the current size or type of infrastructure involved, but indicated the expansion should ease capacity constraints and improve performance for both ChatGPT and API users.
I’ve reached out to OpenAI to ask for more specifics on the above numbers — 30% API growth up from what? doubling the compute fleet up from what? — and will update when I hear back.
New options for ChatGPT Plus and Pro users to search across Microsoft Teams and more…
Also last night, OpenAI updated its ChatGPT release notes online to allow subscribers of ChatGPT Plus ($20 per month) to connect the application to search for files and projects across their third-party accounts on Box, Canva, Dropbox, HubSpot, Notion, Microsoft SharePoint, and Microsoft Teams.
And just a few moments ago, OpenAI again updated the service to allow connections for Gmail, Google Calendar, and Google Contacts to Pro users first, followed by Plus, Team, Enterprise, and Edu plans.
For example, ChatGPT users can search their Gmail for all emails matching a certain query, Dropbox account or Notion workspace during a conversation without toggling over into those separate apps.
In addition, subscribers to the ChatGPT Pro tier ($200 per month) may now link their accounts to Microsoft Teams and GitHub connectors and search those third-party applications.
These join OpenAI’s previous connectors to Gmail, Google Drive and Google Calendar, among other apps.
The individual user/account holder first needs to manually connect these external accounts to ChatGPT.
To do so, they’ll need to:
- Click on their account name in the lower left corner of the web interface
- Click “Settings” from the pop up menu and then…
- Click “Connectors from the left sidebar menu. This should pull up a gallery view of available external apps and icons. Screenshots below.
Unfortunately, these connectors are not available for Pro and Plus subscribers in Europe, Switzerland, and the United Kingdom.
The new connectors are currently in beta and disabled by default for Enterprise and Education plans, though administrators can enable them in settings.
Balancing supply and demand
By combining capacity planning with new productivity integrations, OpenAI is positioning GPT-5 not only as a more powerful AI model but also as part of a more connected workspace.
The staged approach to compute allocation reflects the company’s effort to serve existing customers first while scaling up for future demand.
As the compute expansion comes online, paying users stand to benefit first from both higher availability and more ways to integrate ChatGPT into their daily workflows.
But first, OpenAI needs to stabilize its release and ensure GPT-5 is working smoothly for all the users who want it.
