China's AI Edge: Practical Innovation Over Raw Power
Two AI strategies are colliding. The U.S. is winning the compute race. China is turning AI into revenue. Goldman Sachs' Kinger Lau argues the application-first approach gives Chinese firms a better shot at near-term monetization-and the market is still pricing it like an afterthought.
The U.S. Play: Build the Engines, Own the Future
American giants have gone all-in on hardware and infrastructure. Nvidia and AMD are supplying the chips and data centers that pushed the U.S. to roughly 75% of global AI compute and sent Nvidia's value north of $3.5T by late 2025.
That foundation enables breakthroughs in large language models and generative tools. But it comes at a cost: AI spending in the U.S. is projected to top $200B annually this year, and valuations sit at 35-50x forward earnings for the biggest names. Long-term upside? Yes. Near-term profits are the question.
China's Play: Ship Useful AI, Now
China is channeling AI into everyday workflows-healthcare, e-commerce, manufacturing. Baidu and SenseTime are rolling out diagnostics reading medical images with 95% accuracy and logistics systems cutting delivery times by up to 30%.
Policy is pushing it forward. The 2025 AI Action Plan directs over $50B into applications, and by mid-2025, 800M+ people were using AI-enhanced services daily-from Taobao recommendations to Huawei's industrial IoT. Monetization is direct: AI tutoring apps already capture about 40% of China's K-12 market via subscriptions.
Valuation Gap: Optionality Without Bubble Pricing
China's top tech firms trade at a fraction of U.S. multiples. The ten largest Chinese names (Tencent, Alibaba, BYD, and others) total around $2.5T in market cap versus roughly $25T for U.S. peers. Forward P/Es average 15-20x in China vs. 35-50x for the Magnificent Seven.
Weights tell the same story: tech is ~15% of the CSI 300 vs. nearly 40% of the S&P 500. With China's GDP growth forecast at 4.8% in 2026, earnings catch-up can drive a re-rating without needing heroic assumptions.
The 18-Month Lag That Helps
China trailed the U.S. on foundational models in 2023-2024. That lag is now an advantage as the "application boom" kicks in. AI startup funding hit $15B in H1 2025, up 60% year over year, targeting autonomous mobility and smart agriculture.
Examples are scaling: XPeng has deployed 100,000+ robotaxis in urban fleets, while AI-optimized irrigation is lifting crop yields by about 25%. The Hang Seng Tech Index is up ~25% year to date, and consensus expects aggregate earnings for top Chinese tech to rise ~12% in 2026 as AI-driven efficiencies flow through.
Global Expansion Is Underpriced
U.S. tech gets ~30% of revenue abroad. Chinese peers are ~15%-and closing the gap. Tencent's games now pull ~40% of its $10B quarterly revenue overseas. Alibaba Cloud is signing enterprise clients across Southeast Asia and Latin America, serving around 2M globally.
With tariffs easing in select corridors and digital infrastructure programs widening access, analysts see overseas sales doubling by 2028-an estimated $500B of upside that current multiples don't reflect.
Sentiment and Capital Are Turning
Public sentiment in China skews pro-AI: a 2025 survey from Stanford's HAI shows 78% positive versus 62% in the U.S. Source. That optimism-and policy consistency-matters to capital flows. Foreign inflows into A-shares topped $40B in Q3 alone.
The takeaway: compute wins headlines; applications win cash flow. China is leaning into the latter.
What Builders, Operators, and Investors Should Do Next
- Prioritize application-layer ROI: Healthcare imaging, logistics, education, industrial IoT-these categories already show measurable lift and clear pricing models.
- Track P&L impact, not hype: Look for subscriptions, usage-based pricing, and AI features that lower unit costs or lift conversion rates.
- Watch the re-rating window: Valuations (15-20x forward P/E) plus earnings recovery (~12% in 2026) create a cleaner risk/reward than crowded U.S. trades.
- Mind compute constraints: Scarcity encourages efficient models and vertical focus-often better for margins than brute-force scaling.
- Scale where compliance is easier: Application-first models can side-step some data scrutiny that foundational players face.
Practical Moves You Can Make This Quarter
- Investors: Build a basket tilted to application leaders (healthcare AI, logistics AI, education platforms) and suppliers enabling them. Size positions for policy and FX risk.
- Product and engineering teams: Ship one AI feature that trims costs 10-20% or boosts revenue on an existing line. Vertical beats horizontal right now.
- Operations: Pilot predictive maintenance, smart routing, or AI support agents with clear KPIs (AHT, defect rate, on-time delivery).
- Upskill fast: Align training with your role-data, engineering, product, finance-to cut cycle times from idea to ship. See curated options by role here: AI courses by job.
The Bottom Line
The U.S. built the engines. China is putting them to work. If earnings follow adoption-as they usually do-the market will have to catch up to the cash flows.
Your membership also unlocks:
