OpenAI's $1T IPO Just Got a Wall Street Showcase: ChatGPT for Financial Services Targets Junior Bankers With Morgan Stanley and Evercore Inside

ChatGPT for Financial Services is the first vertical extension of ChatGPT Work, and the product choices reflect the audience. The tool ships with built-in financial datasets from Daloopa (private), PitchBook, LSEG, Crunchbase (private) and others, and connects to existing bank subscriptions from Bloomberg, FactSet, S&P Capital IQ, MSCI, Moody's and Dow Jones Factiva via a single sign-on bridge. Roughly 50 connectors are wired through Model Context Protocol (MCP), including workflow systems like Intapp and Box, plus deal-management tooling like Datasite and private-markets data from Preqin.

Two design choices separate this product from generic ChatGPT. First, every figure the model surfaces links back to the underlying SEC filing, earnings transcript or news source with granular citations — the exact workflow gap that has kept Bloomberg Terminal's $30,000-per-seat, ~85% of Bloomberg LP revenue model alive for four decades. Second, admins can pre-load bank-branded PowerPoint and Excel templates, so a slide deck that used to take an analyst a full night was demoed live in roughly 10 minutes at launch.

The first read of Sep 10 was that OpenAI just took aim at the $30,000/seat Bloomberg Terminal. That framing is half-right. Bloomberg Terminal's edge has always been the chat-and-data combo in a single keystroke; ChatGPT for Financial Services now offers a comparable workflow — \"research this company, build the model, write the deck\" — for the price of a ChatGPT Enterprise contract. The product is gated to \"eligible institutions\" only, with pricing not publicly disclosed, but the directional pressure on per-seat data revenue is unambiguous.

The more important non-obvious finding is how the data incumbents themselves sorted. FactSet didn't fight the launch — it became one. A FactSet LinkedIn post confirmed it was \"selected as a launch partner\" and that the integration is \"powered by FactSet's MCP,\" allowing clients to plug their FactSet subscription straight into ChatGPT, Excel and Google Sheets. S&P Global, MSCI, Moody's and PitchBook followed the same template — embedding their licensed datasets inside the OpenAI environment rather than betting on a defensive moat. The losers are the platforms that stayed walled: Bloomberg Terminal's value depends on the chat being inseparable from the data feed, and that assumption just broke.

OpenAI's pitch is that bankers spend 100-hour weeks on pitch decks and comp tables that take minutes for the model. The Wall Street math underneath that pitch is already moving. Morgan Stanley cut roughly 2,500 jobs in March 2026 — about 3% of its 83,000-person workforce — across investment banking, trading and wealth management, with internal research published the month before forecasting that AI could eliminate 200,000 banking jobs industry-wide by 2030 alongside 30% efficiency gains. Goldman Sachs, with 47,400 employees at year-end 2025, has separately notified staff it plans to reduce its incoming investment-banking analyst class by roughly 40% globally starting in 2027. JPMorgan CEO Jamie Dimon disclosed on the bank's July 2026 earnings call that AI had already eliminated 30%–40% of headcount in specific units.

The product also lands inside OpenAI's IPO supply chain. Goldman Sachs and Morgan Stanley are the lead underwriters on the confidential S-1 OpenAI filed with the SEC on June 8, 2026, targeting a $1T-plus valuation on $40B+ of annualized revenue. The Sep 10 launch is effectively a reference deployment of the technology those very banks are taking to market — a credential OpenAI can point to in the S-1 narrative, and a soft commitment from its bankers to keep buying seats.

Near-term, the trade is on disclosure. FactSet, S&P Global and Moody's all report Q4 calendar 2026 results inside the next 60 days — those earnings calls will be the first public test of whether the OpenAI partnership shows up as incremental ASV (annualized subscription value) growth or as a competitive narrative they have to defend. The FactSet MCP integration was already live as of Sep 10; the question is whether the deal economics (revenue share vs. flat licensing) get disclosed in the Q1 FY2027 print in December.

Long-term, the structural question is whether OpenAI builds vertical AI tools fast enough to justify the $1T IPO multiple. CFO Sarah Friar told staff earlier this year that enterprise customers now account for roughly half of OpenAI's revenue — and that mix is what makes a financial-services vertical a $5B–$10B annual revenue line by 2028 if it lands with the top 50 global banks. The launch product is the prototype; the underwriters (Goldman Sachs and Morgan Stanley) are the test case; and the data partners (FactSet, S&P Global, MSCI) are the ones whose multiples re-rate if they get pulled inside the workflow instead of replaced.