TL;DR. Two data points from this week that change the AI line item in your 2027 budget. OpenAI CFO Sarah Friar told investors on August 14 that enterprise revenue has passed consumer at the company. Anthropic CFO Krishna Rao is running early IPO meetings and, per reporting on August 13, has not yet discussed valuation with investors. The commercial floor for enterprise AI just moved up. If your finance function is still buying AI on the “team plan, expense it, figure it out later” motion, this is the quarter to close that out.

What Actually Happened

Two stories, one signal.

OpenAI, August 14. On a call with investors, CFO Sarah Friar said enterprise revenue at OpenAI now exceeds consumer revenue, a crossover that happened faster than the company modeled. Annualized run rate is roughly $40 billion. Separately, OpenAI’s ads business inside ChatGPT is approaching a $1 billion run rate about six months after launch. CNBC covered the investor comments.

Anthropic, August 13. CFO Krishna Rao is leading early meetings with public-markets investors ahead of a potential IPO. No valuation on the table yet. Anthropic’s last private round closed in May at a $965 billion valuation. CNBC reported the meetings.

Why This Changes Your AI Line

When a vendor’s enterprise book crosses its consumer book, three things happen inside 12 months. All three are already visible in the pricing pages if you know where to look.

One. The floor moves up. The $20 per user “Team” tier stops being where new capability lands. New model releases, higher context windows, and the good tool-use loops go to Business, Enterprise, and API tiers first. The $20 tier holds its price and gradually gets less. This is not a conspiracy. It is what happens when the revenue mix flips.

Two. Contract math replaces credit-card math. Enterprise buyers want DPA-covered, audit-loggable, seat-based commits. Vendors want committed spend. Both sides are motivated to sign, which means annual commits with quarterly rebills replace month-to-month by end of year. Budget for that mechanically. Your controller does not want twelve invoices from three vendors on three different card statements.

Three. The ads product bleeds into enterprise UX. OpenAI is at a $1 billion ads run rate inside ChatGPT. The Enterprise SKU is where you buy that out. If your finance team is pasting real numbers into a consumer surface where a bidder can influence what shows up in the answer, you now have a defensible reason to move them to a paid tier that does not serve ads against your session. See the CFO LLM Privacy Guide for the redaction and enterprise-contract patterns that actually clear legal.

The Anthropic Angle

An IPO track for Anthropic tightens their commercial motion. Two consequences to plan for.

The Claude for Work Team and Enterprise SKUs stop being sold with “call us, we’ll figure it out” pricing. Public-comparable list prices firm up, and the sales team starts hitting quarterly numbers hard. If you have been negotiating a soft deal because Anthropic wanted the logo, get it in writing this quarter or next. Post-S-1 the ground shifts.

Model naming and lifecycle notices become more regular. Public-company vendors publish deprecation calendars. Bake a “model migration” line item into your 2027 IT budget. If your weekly finance workflows depend on a specific Sonnet version, know now what the retirement window looks like. Notes on picking a model for the recurring workflows live in the Claude vs ChatGPT vs Copilot CFO comparison.

What to Do In the Next Two Weeks

Skip the annual planning exercise. These are quick, and every one of them saves real dollars or reduces real exposure.

• Pull the current AI invoices. All of them. Team, Business, Plus, Pro, API, Copilot, the credit-card ones your CFO does not see. Total them. That is your baseline.
• Bucket the spend by workflow, not by vendor. “Weekly close prompts,” “board deck drafts,” “investor updates.” That view tells you which workflows are worth an enterprise commit and which should be canceled.
• For every workflow above $500 per month, ask the owner one question. “If this vendor changed pricing 25 percent next quarter, do we have a switching plan?” No plan means it is a single point of failure, not a workflow.
• Move any workflow that touches customer, employee, or financial data to an Enterprise or Team-with-commercial-data-protection SKU. This week. It is not a 2027 project.
• Add “AI vendor risk” as a standing line on the monthly finance risk report. One line, three bullets, five minutes. Board and audit committee will start asking.

The 2027 Budget Question

Assume total AI spend inside a mid-market finance function goes from roughly 0.4 percent of finance-function opex in 2025 to somewhere between 1.5 and 3.0 percent by end of 2027. That range depends on how much of the FP&A and controllership workflow lives in an AI-native pipeline by then. Higher end if a real portion of monthly close and board reporting has moved. Lower end if the team is still on the “expense it, figure it out later” motion.

Either way, it is now a real budget line. The 5-Prompt Weekly Financial Review only works if the seat is licensed and the data controls are cleared. Same with the AI-assisted weekly cadence. Same with the software rationalization pass you will run in Q4.

Where This Sits in the Broader Series

If you want the full reading order for the site, see The Pragmatic CFO Reading Order: 30 Posts in 6 Tracks. If you are hiring for a finance role that has to run this budget line credibly, start with the 2026 AI-Native CFO Job Description (Long Form).

Push Back on This

If you have a different read on the OpenAI or Anthropic financial disclosures, or you think the vendor floor is going the other way, tell us. Best counter-reads run on the Reader Contributions page. Email hello@thepragmaticcfo.com.

Sources

OpenAI CFO investor call reporting: CNBC, August 14, 2026.

Anthropic CFO IPO meetings reporting: CNBC, August 13, 2026.

Anthropic $965 billion valuation reporting: CNBC, May 28, 2026.