Salesforce is a $42 billion a year business. It took 27 years to get there. Founded in 1999, it was the first cloud company to hit $1 billion in revenue, the first to hit $10 billion, and the first to pass $40 billion. It has 150,000+ customers across every industry in every country. It basically invented the category of cloud software.
No B2B company has ever built something that big, that durable, with that much compounding.
And then Anthropic, a company that did not exist four years ago, passed Salesforce’s year-23 revenue run-rate. In its fifth year.
This is not remotely a shot at Salesforce. No one in classic B2B software has ever done better. Ever.
But this is a chart that should recalibrate how every B2B founder thinks about what is possible.
The Three Curves
Here is what the revenue trajectories look like for three of the most important enterprise technology companies in the world:
- Salesforce (founded 1999): Year 5: $100 million. Year 10: $1 billion. Year 15: $4 billion. Year 19: $10.5 billion. Year 21: $17 billion. Year 25: $38 billion. Year 27: $42 billion.
- OpenAI (founded 2015): Year 7: essentially zero commercial revenue. Year 8: $2 billion. Year 9: $6 billion. Year 10: $25 billion.
- Anthropic (founded 2021): Year 3: $87 million. Year 4: $1 billion. Year 5: $30 billion.
Plot those on a chart and the visual is striking. The Salesforce line is a beautiful, steady compounding curve that rises over nearly three decades. It is what great B2B execution looks like when everything goes right for a very long time.
The OpenAI line is flat at zero for seven years, then shoots almost vertically upward once ChatGPT arrives and the API business takes off. Seven years of pure research, then three years of the fastest revenue ramp in technology history.
The Anthropic line is the most extreme. It barely registers for the first three years, then goes vertical so fast that on a 27-year x-axis, its entire revenue history looks like a straight line pointed at the sky. From $87 million to $30 billion in roughly 15 months. Salesforce went from $87 million to $30 billion in about 22 years.
What Made Salesforce’s Trajectory So Remarkable
It is easy to look at the AI numbers and forget what Salesforce actually accomplished. Salesforce did not have usage-based pricing that scaled with compute. It did not have cloud marketplace distribution. It did not have an existing $500 billion infrastructure layer to plug into. It built everything from scratch.
Marc Benioff started with a CRM product, a four-person team, and an apartment in San Francisco. The company had to convince enterprises that software could run in a browser. That data could live on someone else’s servers. That you could pay monthly instead of writing a seven-figure check for an on-premise license. Every single one of those ideas was considered insane in 1999.
Salesforce spent its first decade building a direct sales force, expanding from CRM into customer service, then marketing, then commerce, then analytics, then platform. Every new billion in revenue required expanding into an adjacent category, building or buying a new product, and hiring thousands of salespeople to sell it.
That is 27 years of relentless execution. And the result is one of the five or six most valuable enterprise software companies ever created.
The AI companies are not doing something better than Salesforce did. They are doing something structurally different.
Why the AI Revenue Curves Are So Much Steeper
There are four structural reasons the AI frontier labs are compressing the Salesforce-era timeline by 80% or more. They all matter for how you think about building a B2B company in 2026.
1. Revenue scales with compute consumption, not seats.
Salesforce’s pricing model is per-user, per-month. If a company has 10,000 employees, Salesforce can sell them 10,000 seats. That is a natural ceiling. Growth requires either more customers or more products.
Anthropic and OpenAI price by API consumption. A single enterprise customer can go from $50,000 a year to $5 million a year just by expanding the number of workflows running through the model. There is no seat ceiling. Revenue scales with usage, and usage scales with how many business processes the customer moves onto the model. Anthropic now has 1,000+ customers spending more than $1 million annually, and that number doubled in under two months. That kind of expansion velocity does not happen with seat-based pricing.
2. Cloud marketplace distribution replaces decades of sales team buildout.
Salesforce spent 15 years building one of the largest enterprise sales forces in the world. Thousands of account executives, SDRs, solution engineers, and channel partners. That infrastructure is what let them land Fortune 500 accounts and expand within them over time. It was effective. It was also expensive and slow.
Anthropic ships through AWS Bedrock, Google Cloud Vertex AI, and Microsoft Azure Foundry. It is available on every major cloud platform. When an enterprise already spends $50 million a year on AWS, adding Claude is a procurement decision that can close in weeks. The entire cloud marketplace infrastructure that Salesforce helped prove out over 20 years now exists as a distribution channel that new companies can plug into from day one.
3. The product is horizontal from the start.
Salesforce began in CRM. It took years to expand into Service Cloud, then Marketing Cloud, then Commerce Cloud, then Analytics. Each expansion required new product development, new go-to-market teams, and sometimes multi-billion dollar acquisitions like ExactTarget, MuleSoft, Tableau, and Slack.
AI models are horizontal on day one. Claude handles customer support, code generation, legal review, financial analysis, content production, data extraction, and dozens of other use cases without building a separate product for each one. The model is the platform. There is no “land in one department, expand to the next” motion. The product is useful everywhere from the first API call.
4. Developer adoption creates a pull-through engine that Salesforce never had at this speed.
Claude Code launched in May 2025. By February 2026, it was at $2.5 billion in annualized run-rate. Developers adopt AI coding tools, build them into production systems, and the usage compounds. That developer-led pull-through motion is similar to what made AWS and Stripe grow quickly, but with AI coding tools the adoption curve is steeper because the productivity gain is so immediately obvious.
Salesforce had developer adoption through Force.com and later Heroku, but those were platform plays that took years to build critical mass. AI coding tools went from zero to a multi-billion dollar market in under a year.
The Profitability Question (It’s a Big One)
There is a very important caveat to all of this. Salesforce is profitable. Very profitable. It generated over $6 billion in free cash flow last fiscal year. It is buying back stock. It pays a dividend. The company is a cash generation machine.
Neither OpenAI nor Anthropic is profitable. OpenAI is burning approximately $17 billion in cash this year and does not project positive free cash flow until 2029. Anthropic has raised over $67 billion in total funding. The run-rate revenue is real, but so are the compute costs required to serve it.
The investors behind both companies are making a specific bet: that compute costs per unit of intelligence continue to fall, that revenue compounds faster than burn, and that whoever owns the AI infrastructure layer in 2029 will generate returns that make the interim losses irrelevant.
That bet might be right. It might not be. But the revenue scale is not in question. Whether the revenue translates to Salesforce-level profitability over time is the open question.
Salesforce proved that cloud software could be an incredible business over three decades. The AI companies have proven they can build revenue faster than anyone thought possible. What they have not yet proven is that the margins follow.
The Game Really Has Changed. Challenge Yourself to Go Much, Much Bigger.
If you are building a B2B company right now, these three curves tell you something important about the range of outcomes that exist.
The Salesforce curve is epic, and still a viable and attractive path. Build a focused product, sell it well, expand into adjacencies, compound for 20 years. That model produced one of the greatest businesses in technology history. There is nothing wrong with it.
But the AI curves show what happens when the structural conditions change. When pricing scales with consumption instead of seats. When distribution is built into the cloud platforms your customers already use. When the product is horizontal from day one.
You do not need to be building a frontier AI model to benefit from these structural shifts. Application companies building on top of Claude and GPT can use the same pricing models, the same distribution channels, and the same horizontal positioning. The infrastructure layer is compressing faster than anyone expected. The application layer should compress too.
Salesforce took a generation to reach $42 billion.
The first AI company to match that number will probably do it in under 10 years.
The playbook Benioff wrote still works. It just runs a lot faster now.
