US enterprises cut AI costs; OpenAI and Anthropic growth may slow as customers shift to efficiency
The AMW Read
Novelty 2: the cost-optimization trend is emerging but not yet dominant in coverage; Significance 3: a structural shift in enterprise AI spending patterns could reshape the competitive dynamics across foundation models, hyperscaler infrastructure, and capital markets.
US enterprises cut AI costs; OpenAI and Anthropic growth may slow as customers shift to efficiency
US enterprises and startups are aggressively reducing AI spending and prioritizing efficiency, a shift that could slow the explosive revenue growth of OpenAI and Anthropic. Uber introduced tiered usage limits for some AI tools after burning through its annual AI budget in four months. Lindy, an AI startup, switched all traffic from Anthropic’s Claude to DeepSeek’s cheaper open-weight model, with CEO Flo Crivello saying the cost curve collapsed to the floor immediately. Microsoft launched a low-cost AI model family, while Amazon plans to compete with frontier models next year using in-house silicon, and Google highlighted Gemini 3.5 Flash at half the price of rivals.
The shift signals the early stage of a capital-compression arc in the foundation-model layer. After two years of enterprises maximizing token consumption regardless of cost, CIOs are now demanding ROI proof before committing to large deployments. This directly pressures the premium-pricing power of OpenAI and Anthropic just as both prepare for IPOs. Analyst Gil Luria of DA Davidson noted that current growth rates may be the highest ever, creating urgency to go public before a temporary growth slowdown hits. The hyperscaler distribution moat—Microsoft, Amazon, Google controlling both infrastructure and models—gives them the leverage to undercut independent labs on price, reinforcing a winner-take-most dynamic at the infrastructure layer.
The article explicitly validates two open debates in the substrate: whether enterprise AI spend will sustain frontier-lab ARR trajectories, and whether hyperscalers will use their vertical stack to compress margins of independent model providers. DA Davidson’s Luria argues that market timing for an IPO has never been better because a growth deceleration is structurally baked in. However, consultant Jeff Henry of Highspring reminds that the toothpaste is out of the tube: AI adoption is irreversible, and the current belt-tightening precedes a wave of mid-market deployments that could reaccelerate demand. The next 12 months will reveal whether the capital-compression arc is cyclical or structural.
#FoundationModels #EnterpriseAI #HyperscalerMoat #CostOptimization #OpenAI #Anthropic #IPO
