AI adoption is surging, but investment in technology has stalled.
Recent findings from Ramp highlight a stagnation in AI spending, which remained unchanged during the latter part of summer, despite increasing token volumes. This trend indicates that competitive pricing and more knowledgeable customers may be limiting revenues for major players like OpenAI and Anthropic. According to Ramp, there was only a modest 1.5% growth in token expenditure from August to September, a stark contrast to a 42% increase seen in July.
As Anthropic prepares for its initial public offering (IPO), the data suggests that expectations of sustained rapid growth in AI firms might be misplaced.
“Currently, almost all AI expenditures come from venture-funded high-growth tech companies,” stated Ara Kharazian, lead economist at Ramp. “This trend may be sustainable as long as AI spending continues to increase, but any downturn could have widespread repercussions across the sector.”
Kharazian revealed that top-tier firms, known as the most “AI-pilled” companies, have already shown signs of reduced spending in August and September, with per-employee expenditures dropping just below the $10,000 monthly benchmark observed during peak summer performance. Even a slight decline among this elite group could significantly impact overall revenue despite increasing usage rates.
While median firms spend between $12 and $15 per employee monthly, and the top 10% allocate $600 to $1,000, this growth isn’t sufficient to offset declines at the higher spending tiers.
Kharazian indicated that current use cases do not suggest that firms outside the top tier will reach the spending levels seen by the top 1%. While there is demand for AI coding agents in sectors like finance and software, industries such as healthcare or automotive may not justify similar expenditure.
According to Kharazian, “The period of unprecedented growth could be coming to an end, and we may be seeing limitations in how much value can be extracted from the top 1% of firms.”
October data, however, revealed a resurgence, with token spending increasing by 8.9% from late September to late October. Despite this uptick, Kharazian warned of persistent challenges to sustained growth.
He emphasized that this slowdown is not due to an absence of demand—AI usage volumes are indeed rising—but rather the result of heightened competition in the market.
“Price reductions by OpenAI and Anthropic have made AI more affordable,” Kharazian explained. In September, OpenAI halved API prices for its GPT-6 Sol and GPT-6 Luna offerings. Anthropic followed suit, slashing prices on its Fable 5.1 and Opus 5.5 models—a move that previously resulted in an anticipated revenue decline of around $600 million.
Despite these enterprises providing discounts to encourage spending, Kharazian believes it has tilted the market in favor of buyers.
The strategy of subsidizing costly technology to retain customers, reminiscent of practices by leading tech firms, has seen limited success in the competitive AI landscape, where numerous alternatives exist. Kharazian noted that Meta and Nvidia are also introducing competitive offerings, often at lower costs or through open-source models, alongside new entrants like Jev, a tool rapidly gaining traction for its efficiency.
“For many AI applications, a simple Yes/No response suffices,” Kharazian stated. “By utilizing cost-effective models and transitioning to tools like Jev, businesses are finding ways to decrease their AI expenses.”
Improved operational efficiencies also contribute to the spending slowdown: as businesses become more adept at leveraging AI, they implement stricter controls over usage and budget allocations.
Notion co-founder Akshay Kothari recently shared data supporting Ramp’s findings, humorously summarizing: “tokens go brrr. costs go ahhh.”
Kharazian remarked, “Many assumptions regarding the trajectory of AI spending are proving incorrect. While usage may increase, it doesn’t necessarily equate to greater expenses. This trend benefits customers but raises concerns for the sustainability of revenue for AI companies.”
In a recent report, Bloomberg noted a considerable drop in Anthropic’s IPO valuation expectations, which fell from $2 trillion to an estimated $1.2 to $1.3 trillion. The company is also predicted to receive a high “junk” rating from credit agencies. Similarly, OpenAI has revised its projections, lowering its valuation from $1.4 trillion to $1.25 trillion and adjusting its fundraising goals downward by approximately $10 billion.
Kharazian believes that Ramp’s dataset may reflect prevalent investor anxieties. Nevertheless, there is a potential silver lining regarding future spending, albeit speculative.
A key component of the belief in a resurgence in spending is the anticipation that AI will create innovative use cases that could reignite growth. AI coding, which emerged as a commercially viable product just a year and a half ago, has played a crucial role in the ascent of companies like Anthropic and OpenAI.
Some industry experts speculate that personal AI agents could emerge as the next big growth drivers. Companies such as OpenAI, SpaceXAI, and Meta have launched their own personal agent products, alongside startups like Instinct. However, Kharazian remains skeptical about the economic productivity of these personal agents compared to coding agents, which are crucial for sustained growth.
“While projections indicate new products will emerge to stimulate growth, current observations and available offerings suggest that AI spending has begun to decelerate,” Kharazian concluded.
Editor’s Take
The stalling of AI investment growth raises critical questions for the industry. With major companies adjusting valuations and adopting aggressive pricing strategies, users may benefit from reduced costs but could also create long-term challenges for AI firms. As the sector evolves, businesses must navigate the competition carefully to ensure sustainable growth.
Source: techcrunch.com
