这些排名反映了企业的势头、增长速度以及收入增长加速情况。

它们展示的是当前正在快速发展的公司，而不是已经很大的公司。

阅读完整报道: https://t.co/CcdtuOFcb7

Brex Benchmark: Spring 2026’s top 25 fastest-growing software vendors

Source: https://www.brex.com/journal/brex-benchmark-may-2026?ref_code=os_tw_spring_benchmark_list Two patterns have shaped the fastest-growing vendor spend in spring 2026. First, as foundation models fragment and turn over weekly, significant value is accruing to whoever routes between them. And second, as vibe coding moves from hackathon to enterprise workflow, the platform that wins isn’t the one with the best output; it’s the one that owns where teams start. These trends rhyme: the winner sits in front of a fragmenting market and abstracts it for the user. The entry point is where the spend is going.

The top 25 trending software vendors (spring 2026)

These rankings capture momentum, scaling velocity, and dollar growth acceleration. They reflect the companies pulling away right now, not the ones already large. The data is realBrex cardandBrex bill payactivity from 35,000+ Brex customers, recency-weighted, not survey responses or stated intent.

## Two trends defining software this spring

Here are the two trends to keep watching in the summer 2026 AI landscape: labs are exploding and aggregators are solving that pain for AI-native startups, and the top three vibe coding players are splitting the market.

1. The ascent of fal.ai and the economics of an ecosystem

The top vendor on our list is fal.ai, and it’s a clean case study in what Ben Thompson callsaggregation theory: when supply gets commoditized and fragmented, value accrues to whoever owns the demand side and routes between options. Fal is evolving fast as their generative media platform matures into an ecosystem.

Nobody in generative media uses one model. The median production deployment runs through dozens of models: one for ideation, another for the first pass, another for editing, another for the final cut. New models drop every few weeks, and quality leadership changes hands constantly. Betting on one provider is slower than routing to the latest and greatest model.

fal.ai turned that into the fastest-growing business on Brex. One API, every model worth using, seamless switching. The user doesn’t pick a model. Rather, fal.ai does, or makes picking trivial. That’s the whole playbook: own the demand-side relationship, abstract the supply, capture margin in the middle.

OpenRouter (#2 on our list) is running the identical playbook for LLMs. Same fragmentation, same routing-layer solution. It’s the generative media ecosystem that production teams build AI solutions on top of. The #1 and #2 vendors on this list are both infrastructure plays. Not a coincidence.

What customers bundle with fal.ai

The bundle data shows two patterns, and both line up with what the aggregation theory framework predicts:

• **Aggregators stack on aggregators.**Replicate is fal.ai’s most common co-purchase: ~40% of Replicate customers also run fal.ai. No single router carries the full model universe yet, so teams hedge by running two or three together.
• **Heavy fal.ai users also buy directly from the labs.**Customers using fal.ai are far more likely than the average Brex customer to purchase straight from foundation model providers: 19x for Luma, 17x for Black Forest Labs, 14x for Stability, and 10x for Kling, Hedra, Ideogram, Magnific, and Krea.- This second pattern is the supplier pushback you’d expect. Labs window their newest models or offer aggressive pricing to keep a direct relationship and avoid pure commoditization. But the customers buying direct arealsorunning fal.ai underneath. The default has already shifted.

fal.ai closed 2025 and ripped through Q1 2026 pulling ahead by market share in generative model routing, ahead of Replicate and keeping pace with OpenRouter.

### 2. Vibe coding, Lovable, and why your first love matters

Vibe coding went from weekend hackathon to startup prototyping standard to enterprise workflow in 18 months, and our spend data tracks the curve almost perfectly. Three players have separated from the pack: Replit, Framer, and Lovable. Replit came in from the developer-first crowd through its cloud IDE lineage. Framer came in from design, converting its base into builders. Lovable came in pure-play AI-native, and from a standing start in January 2025, it grew the fastest of the three to match the other two on spend.

The three entry points tell you why Lovable is the one accelerating. Developers and designers already had tools they liked; AI-native was the entry point with no incumbent to displace. Lovable owned the green field, and our spend data shows the compounding. Teams aren’t just trying it, they’re scaling spend month over month as more prototyping and early product work moves onto the platform.

That’s the most valuable real estate in the category: the entry point. Whoever owns where teams start owns the relationship that compounds. The category consolidates around the platform that extends that entry point into the rest of the build cycle — staging, deploys, customer-facing apps running in one place. Lovable is growing fastest because more teams are starting there.

What heavy Lovable users also buy

The purchase bundle data points to a workflow that’s already forming around Lovable:

• **Teams hedge across vibe coding platforms.**Lovable users are 4x more likely to also run Replit. Three-way race or not, the operators are running multiple platforms in parallel. This is partly to compare output and partly because each tool has different sweet spots.
• **The full no-typing stack is emerging.**Lovable users are 5x more likely to buy Gamma (vibe-coded decks) and 5x more likely to buy Wispr (voice-to-text). The pattern: the people who stopped typing code stopped typing and clicking everything else too. They’re shipping faster than you because they’re not on a keyboard or mouse.
• **Production UI patterns are still part of the loop.**Mobbin users are 5x more likely than the average Brex customer to also buy Lovable. Mobbin is a library of 600,000+ screens from shipped apps, including checkout flows, onboarding sequences, and paywalls. The model writes the code, but the design decisions are still being outsourced.

The race is still three-way on paper. In the share of spend by customers on Brex, Lovable quickly caught up to lead the industry.

## The acceleration of abstraction

The trend among the trends is clear. fal.ai and OpenRouter sit between users and a fragmented supply of models. Lovable sits between users and the entire build cycle. The fastest-growing vendors on this list aren’t building AI — they’re abstracting it. That’s where the spending is going, and it’s accelerating.

Sumeet Marwaha is the Head of Data at Brex, supporting Brex in understanding how customers spend, adopt tools, and grow their businesses.

All analysis conducted for this report that uses Brex internal customer data is anonymized and aggregated for privacy. To learn more about how we use data in anonymized or aggregated form for these trend reports, email us at [email protected].