US Firms Lag AI Frontier, Making Pacing Debate Moot

A heated debate has emerged in Washington and Silicon Valley over whether artificial intelligence development should be deliberately slowed down to allow society to catch up. Proponents of this "pacing" approach argue that safety and alignment must keep pace with raw capability, while critics view it as a strategic error that would cede the technological lead to China. However, this argument may be based on a fundamental misunderstanding of how businesses actually adopt new technology.

According to Fortune, the core issue is that most companies are not ready to utilize the most advanced AI models available today. The prevailing assumption that every new model release immediately boosts the macroeconomy is flawed. In reality, corporate America is years behind the frontier, and the primary bottleneck is not the technology itself, but the internal infrastructure required to deploy it.

Corporate infrastructure lags behind capability

Since the rise of large language models, executives have scrambled to integrate AI into their operations. Yet, the structural realities of enterprise architecture, such as fragmented data silos and legacy systems, make true economic absorption a slow process. Many daily workflows in major corporations do not require the most powerful frontier models; they require simpler, more manageable tools. This means that pausing the development of the most advanced systems would not harm economic output because enterprises are still struggling to assimilate the capabilities already on the market.

Data readiness remains the primary barrier for most organizations. Surveys indicate that more than two-thirds of high-performing companies identify data quality as the main obstacle to AI implementation. Only a small fraction of firms describe their data as completely ready for AI, and fewer than a quarter have a defined data strategy. This suggests that the constraint is organizational and structural, not technical.

Historical patterns defy immediate productivity gains

The belief that AI will instantly transform the economy ignores historical precedents. General-purpose technologies typically take decades to generate broad-based productivity gains. Electricity took approximately 75 years to significantly lift economy-wide productivity, while computers required about 50 years. Even the internet and mobile devices, which were adopted more rapidly, took around 25 years to fully reorganize workflows. AI follows a similar curve, suggesting that the current hype overestimates the immediate economic impact.

Consulting firm McKinsey has highlighted that technical availability is fundamentally different from economic transformation. Their recent survey of the business community found that only six percent of companies reported a significant impact on earnings from AI adoption. This gap between capability and commercial realization underscores that the market is not yet at a stage where slowing down frontier research would cause a measurable economic loss.

Trust and adoption decide commercial success

The commercial fortunes of AI labs will likely be determined by trust and widespread adoption rather than by raw model capability alone. By focusing exclusively on cutting-edge performance, frontier labs have risked mismanaging public trust and messaging. A more balanced approach that prioritizes safety and alignment could stabilize the industry without sacrificing long-term growth. The debate over pacing, therefore, may be less about the speed of innovation and more about the readiness of the business ecosystem to absorb it.