Eight of twelve public B2B software companies now draw 40% or more of their revenue from outside the US — and most got there after years of doing almost nothing internationally. That is the core finding from SaaStr's Jason Lemkin, who pulled international revenue splits from a dozen public SaaS filings in a September 5, 2026 post: 53% at Figma, 49% at Cloudflare and HubSpot, 46% at MongoDB, 36% at Twilio, 26% at Snowflake, 20% at Okta.

The trigger for the analysis: Replit just opened its first international office, in London. Replit was founded in 2016, has more than 50 million users, and is on track for $1 billion in run-rate revenue by year-end. Its $400M round at a $9B valuation was earmarked in part for expansion into Europe, Asia, and the Middle East. Ten years in, a billion in run-rate — and the first office outside the US opens now.

"It's the normal pattern, of maybe waiting … just a bit too long," Lemkin wrote.

What do the growth numbers show?

The pattern repeats across the cohort: international is growing faster than US revenue at most of these companies.

• Cloudflare (49% international): US revenue grew 41% last quarter, EMEA 30%, APAC 32%
• Okta: US revenue grew 12.6% against 9.1% non-US — the exception
• Klaviyo: international up 33% versus 22% in the US
• HubSpot: international up 23% versus 17% domestic
• Snowflake: non-US up 39% versus 34% US
• Figma: international up 48% against 46% total growth

MongoDB has sat at 45%-46% international for over three years. Datadog has held at 28%-30%.

Does putting people on the ground actually move the mix?

Klaviyo offers the clearest measured case. The company added two regional hubs — Dublin and Singapore — not a global rollout. Last quarter its US revenue was $215.8M of $370.6M total. A year earlier the international share was 39.5%; it is now 42%. International revenue grew 33% against 22% in the US, moving the mix nearly three points in four quarters.

That matters because this number is flat for most companies on the list. Lemkin's read: two hubs cost a fraction of a full rollout and moved a metric that MongoDB has printed identically three years running.

Why are Snowflake and Okta stuck at 26% and 20%?

The source draws a sharp distinction between product motions. For self-serve products, underinvestment means leaving free demand unconverted. For enterprise, sales-led products, it is more literal: revenue only appears in a country once you staff it. Snowflake and Okta are not at 26% and 20% because international demand is weak — they are there because the deals require people on the ground and the people were not hired. Snowflake's non-US revenue is now growing faster than its US revenue, which Lemkin notes is what happens once hiring starts.

Okta CEO Todd McKinnon confirmed the diagnosis on X on September 6, 2026: "We should have gone much harder much earlier. One of the many mistakes I've made. Like many things it's just focus and priority!"

When should a founder actually lean in?

Lemkin's threshold is 5%, not 20%. Once a single country hits 5% of revenue with zero local investment, that market has already proven the three hard things: people there can find you, the product works for them, and they will pay through a buying process never designed for them. "5% with no investment is the strongest signal you will ever get about a market," he wrote. "It's demand that survived your neglect."

Leaning in at 5% does not mean a lease and twenty headcount. The prescribed moves:

• Local currency and local payment methods, not just a translated pricing page
• Support hours covering that market's business day
• Pricing tiers set for that market
• One person on the ground in that timezone, whatever their title
• Data residency if you sell to anyone regulated

Lemkin ties the list back to the filing data: this is the difference between Klaviyo moving three points in a year and MongoDB printing the same number three years running.

Figma illustrates the same gap from the product side. 85% of its users are outside the US while 53% of revenue is. Its Bengaluru office, India data hosting and governance, and Brazil localization are all recent — but the gap was visible in the data for years before anyone acted. The broader implication for founders: with half the cohort above 40% and international growth outpacing domestic at most of them, the question is less whether international becomes 40% of revenue than how many years of already-proven demand you leave unconverted before staffing it.