AI funding pace puts OpenAI-era spending habits in play

AI funding is on pace to rise about 300% year over year as companies reassess frontier-model spending and data control.

Reported private-market inflows into AI companies reached about $400 billion in 2025, then approached $300 billion in the first quarter of this year. That quarterly pace has intensified a familiar question for investors: whether model demand can justify the capital now moving into private AI companies.

Private capital accelerates

The funding surge has landed as public and private markets become more tightly linked for investors seeking exposure to AI growth. The harder test is not whether capital is available, but whether companies buying AI systems keep paying for the most expensive models at the levels investors have underwritten.

Much of the skepticism has focused on capital expenditure, especially the infrastructure required to train and serve large models. The newer pressure point is demand composition: if customers migrate from frontier models to lower-cost standard or lite systems, revenue growth can slow even while usage spreads.

Ramp data dents model pricing

Ara Kharazian, Ramp's lead economist, said spending patterns among the heaviest AI buyers are starting to shift. His data showed that the top 1% of businesses by AI spending spent $7,200 per employee per month in August, down 10% from a July peak of $8,000.

"AI spend declined among the top 1% of businesses spending on AI. In August, the top 1% of businesses spent $7.2K per employee per month, down 10% from a July peak ($8K). We're seeing more cracks in the AI thesis as a number of drivers show spend topping out."

Kharazian attributed the decline less to open Chinese models than to competition among model providers. He said price cuts and a rising share of spending on standard and lite models were pulling dollars away from frontier products, which carry higher costs for customers.

That matters for OpenAI, Anthropic and other large research labs that built early adoption around access to the newest frontier systems. If buyers decide good-enough models are enough for many workflows, the addressable market for premium model access becomes narrower than headline usage suggests.

Silvia claims finance-model edge

Silvia said its engineering team spent the past six months rebuilding its technology stack to own more of its AI capability in personal finance. The company said the project was designed to outperform frontier models on tax, mortgage and credit-card topics.

The claim is company-stated, and Silvia did not name a third-party benchmark auditor in the announcement. Still, the direction is consistent with a broader enterprise argument: companies want tighter control over proprietary data, intellectual property and product experience rather than routing core intelligence through a small set of outside labs.

For Silvia, the commercial logic is direct. If its in-house stack performs better on finance-specific tasks while costing less to run, the company can reduce dependence on external model pricing and tune outputs for investors using AI to manage personal finances.

Three paths for AI spend

If premium frontier-model demand holds, the funding boom has a cleaner macro mechanism: private capital keeps financing compute capacity, leading labs preserve pricing power, and AI infrastructure suppliers continue to benefit from high utilization. In that path, Silvia's vertical stack would be a differentiated product choice rather than a broad challenge to frontier labs.

If spending shifts toward lite and standard models, global capital allocation becomes harder to justify at current private-market pace because revenue per enterprise user may compress. OpenAI and Anthropic would face pressure to defend margins, while software companies and AI application developers could gain from lower input costs.

If more companies build proprietary AI stacks, the industry moves toward specialization rather than one dominant model layer. The open question is whether internal systems can consistently match frontier quality in regulated, high-stakes domains; that will determine whether Silvia's announcement is an isolated product claim or an early sign of a broader buyer shift.