Leaked OpenAI financials show what the AI business is becoming underneath the product excitement

OpenAI’s revenue grew from $3.7B to $13.07B in a year. But sales and marketing grew even faster: from $1.11B to $5.73B. That moved selling costs from 30% of revenue to 44%.

For context:

• Salesforce spent 35% of revenue on sales and marketing in FY2026.

• Snowflake spent 44%

• Datadog was 28%.

• ServiceNow ~ 33%.

So OpenAI’s number is not absurd for high-growth enterprise software.

Then watch what every AI lab did next

In launching the OpenAI Partner Network earlier this month, OpenAI says the enterprise bottleneck is no longer model capability, but identifying use cases, redesigning workflows, integrating with systems, and driving adoption.

It is putting $150M into the partner ecosystem and aiming for 300,000 certified consultants by the end of 2026.

Anthropic is doing the same thing. Its Claude Partner Network commits $100M for 2026, including training, technical support, sales enablement, market development, co-marketing, and partner certifications.

Anthropic says partners help customers navigate deployment, compliance, and change management.

Cohere and Mistral built formal programs around consultancies, cloud providers, and systems integrators.

Hiring confirms the GTM direction

OpenAI and Anthropic are reportedly hiring heavily from Salesforce.

One report says Anthropic hired more than 45 former Salesforce employees since early 2026, while OpenAI hired nearly 40, concentrated in sales, marketing, GTM, and revenue roles.

That is not “AI replacing sales”; that is AI labs importing enterprise-sales DNA

So is selling AI different from selling enterprise software?

The benchmarks say no.

The foundation model was supposed to collapse the enterprise software stack.

Instead, the go-to-market stack is re-forming around the model.

AI has pull, but pull is not the same as enterprise adoption.

Customers do not just buy intelligence. They buy a changed operating model, with the risk carried by someone they trust.

Three lessons for alliance leaders:

1. Distribution is still the moat, just like it was - the labs are adopting familiar partner GTM playbooks

2. Own the last mile. Workflow redesign, compliance, and change management are exactly where labs admit they need help

3. Move early. With 300,000 certification slots and co-sell motions opening at once, the first movers set the terms.

Is AI distribution following the SaaS playbook, or writing a new one? What’s your take?