East/West AIpha · Edition 5 · Tuesday, August 25, 2026 · Day 1023

Automate Work. Amplify Life. East/West AI, for the people betting real capital and real careers on it.

Answer first. The AI race everyone scores by benchmarks is the race for the smartest model. The race that will tax the next decade of commerce is the race for the interface that spends the money. In February, China ran the largest live drill for it in history.

The number: 120 million consumer orders placed through 阿里巴巴 Alibaba’s AI app, 千问 Qwen, in six days. Not searched. Not compared. Completed. The AI found the shop, applied the discount, paid through Alipay, dispatched the rider (SCMP, Feb 12; Alibaba Cloud). 

The campaign was a subsidy, and it mostly did not work. Time in the app fell by half during the giveaway. Daily users dropped by about half from the peak once subsidies ended (Morgan Stanley and DataEye, via 36Kr). 

But the residue is real. Qwen’s daily base settled at several times the 7 million it started with. And 1.56 million people over sixty made their first online purchase through an AI (Alibaba Cloud). 

Meanwhile the West walked off the field. OpenAI launched Instant Checkout in September 2025 and pulled it back in March 2026. About a dozen of Shopify’s millions of merchants ever went live (OpenAI; Forbes, Mar 10). 

The prize is the front door. McKinsey sizes agent-orchestrated commerce at $3 trillion to $5 trillion globally by 2030 (McKinsey). 

If you read only this far: free money built the WeChat payments empire in 2015. In 2026, the same playbook, same holiday, same weapon, is binding not bank cards but the willingness to let a machine spend your money. The skeptics are right about the boba. They are wrong about the bet. Subscribe free at workoptional.ai.

A 62-year-old woman in a county town in Anhui says one sentence to her phone: “Order me a milk tea.” She had never bought anything online in her life.

She is one of 1.56 million people over sixty who made their first online purchase through an AI that week (Alibaba Cloud, Feb 12).  The tea was free. What Alibaba bought with it was not.

On February 6, eleven days before Chinese New Year, Alibaba launched a RMB 3 billion giveaway, about $431 million (Caixin, Feb 7).  The catch was the whole point. You could not claim the deal on Taobao or Alipay. You had to ask the AI, and the AI did everything: found the shop, applied the discount, paid through Alipay, dispatched the rider. One sentence, one transaction. Alibaba had wired Taobao, Alipay, Fliggy, and Amap into Qwen a month earlier (Alibaba Cloud, Jan 15). 

Ten million orders in the first nine hours (China Securities Journal, Feb 6). One hundred twenty million across six days (CNR, Feb 12). More than 130 million people ordered, roughly one in ten Chinese (QuestMobile, via Futunn).  Qwen jumped from number 10 to number 1 on China’s App Store and held it for six straight days. Tea shops hung temporarily closed signs as Hong Kong tea stocks rallied (STAR Market Daily, via Moomoo).  Alipay’s AI Pay cleared 120 million transactions in a week, a world first for agentic payment at scale (Business Wire). 

The strangest thing about that week: it has happened before.

Chinese New Year, 2014. 微信 WeChat launches digital red envelopes: free cash, sent through chat. Jack Ma watches users flood in and calls it the Pearl Harbor attack, perfectly planned and executed (Forbes, Feb 2015; People’s Daily, Feb 2014). 

He saw it coming. It did not save him.

The next year, Tencent sponsored the Spring Festival Gala and gave away RMB 500 million. Viewers shook their phones 11 billion times in one night (Forbes, Feb 2015).  To claim the cash, you bound your bank card (Quartz). Alipay’s share of third-party mobile payments fell from 82.6 percent in Q3 2014 to 55.0 percent by 2016 (CGAP and World Bank). 

Same holiday, same weapon, eleven years later. But the thing being bound this time is not your bank card. It is your willingness to let a machine spend your money.

The playbook has a failure mode. Baidu ran this exact move at the 2019 Gala: roughly RMB 900 million in cash, 104 million new users. Retention within days: 2 percent (Guojin Securities, Feb 2019).  Subsidies only compound when there is a network worth staying in; Alibaba’s bet is that its rails are the network.

And this time, the state is running point. The AI Plus guideline sets targets, not forecasts: smart terminals and agents above 70 percent penetration by 2027. Above 90 percent by 2030 (State Council, Aug 2025).  The sentiment gap shows it: 83 percent of Chinese respondents say AI’s benefits outweigh its drawbacks; in America, 39 percent (Stanford AI Index, 2025). 

While China stress-tested agentic commerce on 120 million orders, the West retreated. OpenAI launched Instant Checkout last September and pulled it back in March, routing transactions into merchants’ own sites (OpenAI, Mar 2026).  About a dozen of Shopify’s millions of merchants ever went live (Forbes, Mar 10).  Walmart found conversion inside ChatGPT ran three times lower than clicking out (Search Engine Land). 

The structural problem: the West has no superapp substrate, credit cards already work, and the rails are fragmented. OpenAI and Stripe’s Agentic Commerce Protocol now competes with Google’s Universal Commerce Protocol (Forrester). 

And consumers flinch. In Visa’s 2026 consumer research, 58 percent are comfortable letting AI compare prices. Only 38 percent would let it complete a purchase. Sixty percent would not allow AI to spend without their approval. Trust clusters around banks (36 percent) and card networks (35 percent), not independent agents (28 percent) (Visa). 

The West is building smarter models and cleaner protocols. China is building habituated users. Only one of those compounds.

Now the part the hype skips.

The boba campaign mostly did not work. Daily time on Qwen fell from 6.3 minutes to as little as 3 during the giveaway (Morgan Stanley, via 36Kr).  People came for coupons, not conversation. Daily users ran from 7.1 million to a 73.5 million peak, then fell by roughly half once subsidies ended (QuestMobile, via Futunn; 36Kr).  Effective acquisition cost: about RMB 144 per daily active user, double the RMB 69 腾讯 Tencent paid for 元宝 Yuanbao. Total spend, including compute: an estimated RMB 6 billion (DataEye, via 36Kr).  Paid adoption is rented, not owned.

In the March quarter, sales and marketing spend jumped 47.6 percent to RMB 53.4 billion. Alibaba attributed it to quick commerce and Qwen user acquisition (Alibaba results).  In the June quarter, reported August 20: revenue up 9 percent, adjusted EBITA down 30 percent, net income down 75 percent. Free cash outflow hit RMB 44.7 billion as capex rose 75 percent. AI Cloud and Compute Services grew revenue 45 percent, adjusted EBITA up 133 percent. AI Labs and Applications, home of Qwen, lost RMB 13.9 billion in adjusted EBITA (SEC exhibit, Aug 20).  Bloomberg Intelligence estimates Alibaba redeployed over 90 percent of March-quarter China e-commerce profit into Qwen acquisition (May 2026). (MODERATE, single source)

But here is the residue. Qwen’s daily base still sits at several times its pre-campaign 7 million, and June monthly actives hit 167 million (DataEye, via 36Kr).  Alibaba says 250 million users have now had a first AI-driven shopping experience through Qwen (SEC exhibit).  Nobody un-learns delegation: in 2015, the skeptics also called red envelopes a gimmick, and the bank cards stayed bound.

And in August, Qwen began charging: office subscriptions at RMB 98 a month, quotas on the free tier (36Kr).  Management says consumer AI payments may take one to two years to monetize (earnings call transcript).  The monetization question is now open.

In 2014, Jack Ma watched free money take half his payments empire and called it Pearl Harbor. In 2026, his company spent $431 million on milk tea to make sure it never happens again.

The skeptics are right about the boba. They are wrong about the bet.

Two commandments from my book, The Ten Commandments of Investing, were written for a week like this. Commandment Four, Don’t Be a Dead Fish: “A dead fish will only go where the river flows.” This week there were two herds. One stampeded into the hype: 120 million orders, the future is here. The other stampeded into cynicism: the churn was brutal, agentic commerce is dead. The water between the herds is where the alpha swims. Both herds are right about the tea and wrong about the trend. Paid adoption is rented, but delegation, once learned, is never un-learned. So do the work: keep an independent mind, separate the subsidy from the substrate, and price the layer, not the promotion. That is Commandment Six, Be Patient, Be Bold: “the right time is when the price is right.” Patience pays off. When the right price presents itself, go bold.

For allocators: watch delegation metrics, not download spikes. Downloads are bought; completed agent transactions are earned. Three numbers decide the next two quarters: Qwen’s paid conversion, the AI Labs loss trajectory, and group free cash flow (SEC exhibit). In the West, the checkout vacuum favors whoever owns rails, not whoever rents chat traffic.

For operators: if an AI agent cannot reach your catalog, you are invisible to the fastest-growing storefront in China, and soon elsewhere. Make inventory agent-readable. Decide which agents may transact. Build the trust layer: a wrong answer costs goodwill, a wrong order costs a customer.

For builders: China’s edge is habituation on integrated rails. The West’s edge is trust and generality. Do not try to out-subsidize a superapp. Win where delegation fails: high-consideration, high-trust purchases. Build for a multi-protocol world, because the rails war is not close to settled (Forrester).

If you allocate into platforms, payments, commerce infrastructure, or Chinese consumer tech, or you are building an agent that touches a transaction, write to me. Tell me what you are building or allocating; I will share what is working for others. One email to [email protected].

If this was useful, subscribe free at workoptional.ai and forward it to one allocator still scoring the wrong race.

Here is my question for you. When the machine that spends your money is chosen, will you own the front door, or pay rent to whoever does? Reply and tell me. I read every one.

San

San Eng is the WSJ (#4) and USA Today best-selling author of The Ten Commandments of Investing. He is an institutional investor with $500 million-plus deployed across East and West. East/West AIpha exists for one reason: research you can trust, from an investor who actually builds.

Subscribe free at workoptional.ai.

Disclaimer. East/West AIpha is published by WorkOptional.ai for information and education only. Nothing in this edition is investment advice, a solicitation, or a recommendation to buy or sell any security. The author may hold positions in securities discussed. Items labeled MODERATE or LOW reflect single-source or company-reported data; private-market valuations and unaudited revenue figures may not reflect realizable value. Past performance does not guarantee future results. Consult a licensed financial adviser before acting on anything published here.