The AI Pacing Debate Misses the Real Business Bottleneck

The intensifying argument over whether to slow frontier artificial intelligence development is built on a false premise, according to a growing body of evidence from corporate boardrooms and economic history. The debate over «pacing» — the deliberate throttling of cutting-edge model development until safety and alignment can catch up — has divided Washington and Silicon Valley into two camps. Skeptics call it unilateral disarmament in the competition with China. Supporters call it the only responsible path for a technology whose own creators warn of catastrophic risk.

Both sides, however, have fallen into what analysts describe as the «Compute-to-GDP Fallacy»: the mistaken belief that every incremental leap in model performance immediately translates into macroeconomic output. Every prior general-purpose technology took decades to diffuse into measurable productivity gains. Electricity required 75 years to lift productivity economy-wide. Computers took 50 years. The internet and mobile devices demanded 25. Artificial intelligence is following the same curve at an accelerated pace, but the laws of economic history have not been suspended.

The practical reality inside Corporate America is stark. More than two-thirds of high-performing companies identify data as the primary barrier to implementing AI, a figure that has remained stubborn even as models have leaped forward. Only 7% describe their data as «completely ready» for AI. Fewer than a quarter have a data strategy at all, and 63% either lack AI-suitable data management or are unsure whether they have it. Fragmented data silos, legacy enterprise resource planning systems, strict compliance regimes, and basic data hygiene problems make true economic absorption an inherently slow process.

McKinsey research reinforces the gap between technical capability and business results. Only 6% of companies report a «significant» impact from AI and modest earnings attribution. Most are concentrating on high-reward, low-risk automation tasks that models one or two generations old can already solve. As one former Wall Street chief executive put it, these systems will run in parallel with legacy systems for years to confirm they operate correctly and that no regulatory risk is unknowingly absorbed.

That gap has produced a parallel shift in the economics of silicon. Older-generation chips, initially cast aside in the scramble for cutting-edge accelerators, are finding a second life as workhorses for the practical inference tasks that dominate enterprise demand. Miro Dimitrov, founder and chief executive of Growth Protocol, said at the Yale CEO Caucus that deploying neuro-symbolic architectures allowed his enterprise reasoning platform to cut inference costs by roughly 80-fold in live client deployments, largely by shifting workloads off expensive GPUs and onto everyday enterprise CPUs.

Corporate AI adoption is best understood in three phases, distinguished by how much work a company can responsibly hand over, gated by data readiness and earned trust. The first phase, assistance, consists of off-the-shelf copilots that ride atop platforms such as Salesforce, connecting data across existing applications and enabling employees to work faster with minimal re-architecting. Payback arrives quickly and risk stays modest because a human still performs much of the work.

The second phase, orchestration, covers agentic workflows that demand real investment — structuring proprietary data and connecting far-flung data lakes never meant to meet — with a human in the loop approving each consequential step. The third phase, autonomy, brings end-to-end agentic operations across seamlessly interconnected systems with human oversight reduced to exception handling.

The pacing skeptics raise legitimate questions. Is pacing real, or a marketing gambit by frontier labs and cybersecurity companies polishing their financials ahead of public offerings? Would pacing cede the U.S. lead to China, or would Beijing reciprocate in its own manner? Those questions deserve serious answers. But they miss the central point: enterprises need time simply to assimilate the capabilities already on the table.

Pacing would therefore neither harm economic output nor choke off the labs' commercial revenues. The labs' fortunes will be decided by trust and adoption, not raw capability. Racing ahead of alignment could cost far more than slowing down. The argument that matters is not how fast the frontier moves, but how quickly the broader economy can absorb what already exists.

3Views

Author

Delaney Sawyer

Society Reporter

Delaney Sawyer covers public affairs, politics, business, culture and daily news for Core Memo. The role focuses on verification, context, and clear explanations for readers.

All articles by this author