A narrow concentration of corporate spenders is driving artificial intelligence software revenues, with 1% of firms accounting for 80% of total business spending on OpenAI and Anthropic, according to corporate card provider Ramp. Spending growth across commercial users has virtually flattened during August and September, leaving the broader expansion heavily reliant on a small group of enterprise clients.
High Concentration and Falling Prices Stifle Growth
Per-employee spending among the top 1% of AI-consuming businesses reached $9,568 in September. While this represents an 11.7% increase from August, it remains 3.1% below the peak level recorded in July. Ramp lead economist Ara Kharazian attributed the flatlining revenue growth primarily to falling prices for top-tier AI models, even as total token usage volume expanded. Paid AI adoption across Ramp business clients reached 55.9% in September.
However, economist Brad DeLong noted that Ramp's customer base skews heavily toward technology companies and omits free AI tool usage. Reporting on the findings, CNBC reporter Kate Rooney highlighted that roughly 80% of these top spending firms are venture-backed. If elevated interest rates or a quiet initial public offering market restrict venture funding, the resulting spending cuts could transmit shockwaves from model creators to cloud hosting providers and semiconductor manufacturers. Market participants have raised broader concerns regarding technology sector dependency; for instance, Fidelity Macro Director Jurrien Timmer warned of AI bubble risks across financial markets.
S&P 500 Market Concentration Echoes Dot-Com Era
Asset management firm Apollo confirmed that enterprise AI spending expansion has plateaued. Concurrently, Citadel CEO Ken Griffin issued a graphic processing unit (GPU) oversupply warning to Bloomberg, stating that excess chip inventory could force hardware prices lower.
Investment bank Goldman Sachs estimates that AI capital expenditures account for approximately half (50%) of all S&P 500 earnings growth, prompting the bank to steer risk-averse clients toward defensive sectors such as healthcare and consumer staples. Concentration risks are also visible in major benchmark indices. Financial commentary publication The Kobeissi Letter reported that the 15 largest equities now comprise a record 46.7% of the S&P 500, compared to 27.2% at the height of the dot-com bubble in 2000. As investors evaluate hardware exposure, commentators like venture capitalist Fred Wilson argue energy represents a simpler bet on AI than semiconductor equities.
Key Takeaways
- 1% of enterprise clients generate 80% of corporate revenue for OpenAI and Anthropic.
- Top spenders averaged $9,568 per employee in September, down 3.1% from July's peak.
- Roughly 80% of the top AI spenders are venture-backed businesses.
- The top 15 companies account for 46.7% of the S&P 500, up from 27.2% during the 2000 dot-com peak.
Why It Matters
The extreme concentration of AI spending among venture-backed startups presents significant risk for the broader technology and financial sectors. If venture funding cools due to persistent macroeconomic pressure or delayed public listings, model developers and infrastructure providers could face immediate revenue contractions. Furthermore, with half of S&P 500 earnings growth tied to AI CapEx, any prolonged spending slowdown threatens equity market valuations beyond the tech sector.
