Opinion | America builds AI, China uses it. That gap may decide the future

The article argues that the US leads in AI development, with private AI investment topping US$109 billion in 2024, nearly 12 times China’s level, but China is ahead in large-scale implementation. China now has more than 600 million registered generative AI users and hundreds of deployed models across sectors including hospitals and logistics, while US firms often layer AI onto legacy systems. The piece frames AI adoption as an operational transformation issue, especially in trucking and supply chains, rather than a pure technology race.

Analysis

The market is still pricing AI as a model-training arms race, but the more important monetization gap is deployment velocity. That favors businesses that sell workflow orchestration, systems integration, and vertical software over pure compute or frontier-model exposure, because the value capture shifts from R&D spend to operational redesign. In the next 12-24 months, the incremental winners are likely to be “picks-and-shovels” vendors with embedded distribution into logistics, healthcare, and industrials, not the firms that merely announce AI features.

The second-order effect is pressure on U.S. incumbents with legacy operating systems and fragmented procurement. If AI is layered onto old processes, productivity gains stay trapped at the pilot stage, which means margin expansion can disappoint even when headline AI budgets rise. That creates a bifurcation: companies willing to rip and replace workflows should compound faster, while those treating AI as an add-on may see capex rise before revenue productivity follows.

For logistics specifically, the key signal is not AI adoption but network re-optimization. A fleet that dynamically routes, prices, and schedules can reduce empty miles and working capital tied up in inventory/transit; that benefits large-scale operators and software vendors that own the data exhaust. Healthcare is similar: providers with integrated records and centralized operations can extract leverage, while fragmented systems will mostly buy expensive point solutions.

The contrarian view is that China’s apparent lead in deployment may be easier to replicate than the article suggests, because U.S. enterprise AI cycles typically lag until standards, compliance, and integration templates mature. That means the current gap may narrow quickly once a few reference architectures prove ROI. The risk to the “China deploys faster” thesis is policy: export controls, data localization, and procurement restrictions can slow model access and chip availability, but they also push domestic U.S. buyers toward efficient software stacks rather than brute-force compute.

Key Decisions for Investors

• Long MSFT / short legacy enterprise software basket over 6-12 months: MSFT monetizes workflow integration and distribution; the short leg should be legacy vendors whose AI narratives are mostly feature-level. Target a 15-20% relative spread if enterprise AI spend shifts from pilots to platform rollouts.
• Long PANW and CRWD on a 3-9 month horizon: as firms redesign workflows around AI, identity, data, and access control become more valuable. Risk/reward is attractive because security spend scales with deployment intensity, not just model count.
• Long UBER and/or XPO on a 6-12 month horizon versus short high-cost, legacy freight intermediaries: AI-driven routing and utilization should compound network advantages and reduce empty-mile drag. Best entry is on any pullback tied to macro freight softness; upside is operational leverage plus multiple re-rating.
• Avoid chasing pure-play model-training beneficiaries unless they show clear enterprise deployment attach rates. If you need exposure, prefer a small call-spread in NVDA over common stock for 6-12 months: upside if capex stays elevated, but defined downside if the market shifts from training to software capture.
• Pair long VEEV or senior healthcare IT names against short fragmented provider exposure over 12 months: integrated data and workflow stacks should outperform as healthcare AI moves from experimentation to operational scheduling, documentation, and triage.