Ramp's economist: 35.5% of US businesses are using AI — four times higher than the census estimate — and vibe coding platforms are growing faster than model companies

Summary

Ramp's in-house economist Arjun Ramani has a simple but pointed argument: the US Census Bureau's 8% AI adoption figure is wrong, and the gap between official statistics and ground-level reality is now wide enough to matter for policy and investment decisions.

Ramp's latest spending report puts 35.5% of US businesses as active AI adopters — defined by actual transaction data, not survey responses. That's roughly four times the Census estimate. The methodological difference is significant. The Census asks a single respondent per company whether they use AI to produce goods and services — often whoever picks up the phone. Ramp tracks actual contracts and payments to AI vendors across its platform, capturing line-item API credits even when billed through hyperscalers like Azure.

Sector breakdown

Adoption varies sharply by industry. Tech sits at 65% — high but, Ramani notes, still leaving a surprising slice of the sector uncommitted. Restaurants are at 18%, almost entirely for back-office tasks rather than customer-facing operations, which means 82% of restaurants haven't touched it yet. Adoption rates continue to rise month-over-month, with no sign of a plateau.

Model company rankings

OpenAI is the clear leader: more than a quarter of businesses on Ramp's platform hold an active OpenAI contract, driven by both ChatGPT subscriptions and API integrations. Anthropic is a distant second but still growing quickly. xAI has grown fast despite only operating for a few months — Grok 3 drove a meaningful adoption spike — though it remains a small player in absolute terms. DeepSeek had a sharp spike in January that faded quickly the following month, partly because many enterprise implementations routed through hyperscalers or local-server deployments for security reasons rather than paying DeepSeek directly.

Businesses routinely multihome: using OpenAI one month, switching to Anthropic the next, often running both simultaneously. Ramani frames this as a learning curve rather than settled vendor selection.

Vibe coding outpacing model companies

The sharpest growth signal in the latest data isn't the model providers. Coding platforms — Cursor, Lovable, and their peers — are growing faster than the underlying model companies, posting growth rates in the 40s month-over-month on Ramp's data. Canva also ranked among the fastest-growing vendors, though driven by broad AI feature integration rather than coding specifically.

Tariffs and recession signals

Ramp's transaction data hasn't yet shown an aggregate slowdown in business spend, but Ramani is cautious about reading that as a clean bill of health. Thirty interviews with retail, construction, and manufacturing businesses last week found widespread policy uncertainty and a dominant wait-and-see posture — several companies told Ramp they were pausing all capital expenditures until tariff policy settles. The data lag means the full effect may take another couple of months to surface.

The leading indicator Ramani watches most closely is digital advertising spend. It's immediate, discretionary, and a cleaner signal of business confidence than CEO sentiment surveys. Last month, 55% of large businesses reduced or held flat their ad spend year-over-year. Small businesses and startups are still increasing it, but the large-company trend is worth watching. Polymarket currently prices a US recession at 51%.

The reallocation question

Ramani doesn't take a strong position on where AI-driven efficiency savings ultimately flow, but the question being raised is concrete: if an e-commerce business drops content production costs from $50,000 to $5,000 a month, does that freed budget expand customer acquisition, enter new markets, or simply compress overall spend? The same logic applies to professional services — a CPA firm engagement that costs $20,000 a month might cost $5,000 once AI reduces headcount requirements. Whether those savings recirculate into growth investment or narrow business budgets more broadly is the open macro question Ramp's data is not yet positioned to answer.