Defense Space Science

The Consensus Everyone Should Question

The single most confident prediction circulating in financial and policy circles holds that AI infrastructure spending will sustain high revenue growth and eventually produce the margin expansion that justifies current tech valuations. The narrative: AI revenue hit 25 billion dollars quarterly, it has now covered depreciation for two consecutive quarters, and the trajectory is clearly upward. Bank analysts and venture capitalists alike are converging on a story in which AI spending continues at its current pace and the economics eventually work out.

The Ford case offers the most direct challenge to that assumption. Ford built AI quality control tools, deployed them at manufacturing scale, found they fell short, and rehired 350 veteran engineers. That is a real-world experiment in AI replacing skilled human knowledge work — and the experiment partially failed. If that pattern repeats across industries, the enterprise AI adoption curves underlying current revenue projections may not materialize at the assumed rate.

Apple's price hikes add a second pressure point. A standard technology industry assumption holds that compute costs fall rapidly over time. If AI workloads have specific characteristics — high memory bandwidth requirements, particular chip architectures — that do not benefit from standard Moore's Law-type scaling, the margin expansion the market expects may arrive far later than models project. DeepSeek's trajectory is a third variable: the company that demonstrated frontier AI could be built cheaply is now spending like every other major AI lab after raising 7.4 billion dollars, suggesting its efficiency advantage may not have been durable.

Three concrete indicators are worth watching as potential signals that the consensus is wrong: enterprise AI contract renewal rates at the 18-month mark — are companies that deployed AI tools in early 2025 renewing at full value or renegotiating down; whether AI chip supply constraints cause another round of consumer price increases across product categories; and whether DeepSeek, after doubling its team, requires the same capital structure as its competitors. The revenue is real. The open question is whether the cost structure of delivering that revenue ever produces the margins the market is pricing in.