Tracking the enterprise revenue split between OpenAI and Anthropic remains an exercise in proxy forensics while both vendors guard their private ledgers ahead of anticipated initial public offerings. With audited books unavailable, transactional telemetry provides the only ground truth. That indicator is currently signaling an aggressive counteroffensive: OpenAI is clawing back market share in enterprise card spending across thousands of American companies.
Shifts in Corporate Expense Data
Data from Ramp, which monitors corporate card spending and bill pay across more than 70,000 businesses, shows Anthropic briefly seizing spend dominance in May at 41% versus OpenAI's 39%. By July, Anthropic stretched that margin to nearly 44% against OpenAI's 40%. While Ramp's cohort tilts toward venture-backed technology firms rather than legacy Global 2000 enterprises routed through American Express, the trendlines expose genuine enterprise sentiment.
By August, the momentum reversed. Ramp economist Ara Kharazian noted that OpenAI began outpacing Anthropic in sequential growth during the third quarter, turning what looked like durable Claude lock-in into an open dogfight.
"GPT-5.6 Sol is really good, increasingly the choice for developers," Kharazian posted on X about OpenAI's new growth. "Fable 5, meanwhile, disappointed both in adoption and real-world application given price + data retention requirements imposed by regulators."
Anthropic structured Fable as a specialized high-tier model, which explains its premium pricing. Yet adoption stalled over friction: Anthropic's disclosure that regulatory mandates impose a 30-day enterprise data retention window triggered pushback from compliance-conscious buyers, handing leverage straight back to OpenAI's API suite.
Spend Expansion and Account Churn
Aggregate B2B artificial intelligence budgets continue to climb. Ramp confirms that overall customer penetration is expanding, allowing both foundation model providers to post absolute top-line growth even as they poach accounts from each other.
Yet this rapid account churn undermines the foundational narrative sold to public-market investors. Enterprise AI spend was marketed as high-margin, sticky infrastructure revenue with SaaS-like durability. In practice, enterprise accounts treat foundation models as interchangeable commodities, migrating workloads with every incremental benchmark update. For CTOs and finance chiefs, this volatility is not a liability—it is leverage. Multi-model routing across both vendors has emerged as the baseline standard for risk management, keeping switching costs near zero and forcing both IPO candidates into continuous price competition per million tokens.