When everyone can be Spielberg, don't be a studio. Be YouTube. AI video is turning every creator into a studio. The big winners will be the platforms, not the producers. That’s what iQIYI is doing right now. Listen to the full episode: podcasts.apple.com/us/podcas…

Counter Positioning Is Overrated. Early Movers Win More Often. My favorite strategy for is copying what works elsewhere. And getting there first. Shopee in SE Asia was a copy of Alibaba in China. Garena was a copy of Tencent gaming. (More on this in my free weekly email, link below.) That's a version of early mover. I call it fast follower. Investors, in contrast, get excited about counter positioning. I think it is clever but rare. Being early is far more powerful and far more common. My approach: 1. Tactics Are at the Bottom of My Moat Hierarchy. Digital lets you change prices, promotions and products in real time. These tactical moves never stop. They're important. But you can't win here long-term. 2. Early Mover Exploits a Temporary Supply Demand Gap. Demand shows up before supply does, so you grab customers before rivals arrive. This is even better in digital. You can test constantly, so being early also makes you smarter than later arrivals. 3. Counter Positioning Traps Incumbents, But Only for a While. Kodak saw digital cameras coming but delayed, fearing it would kill its film cash cow. Netflix put Blockbuster in the same bind. But this advantage is not a moat and fades fast. It's sexy. But overrated. ***TAKEAWAYS FOR MANAGERS: → Ask your team which of your current wins are short term tactics → Then identify 1 project that builds a lasting defense. I WRITE ABOUT DIGITAL AI STRATEGY EVERY WEEK IN MY FREE AGENTIC FRONTIER EMAIL. Sign up here and get my book Moats and Marathons free: jefftowson.com/free-email-si… QUESTION: Which incumbent today is stuck protecting a cash cow? What will it cost them?

JD’S 101HOME FIXES FURNITURE RETAIL’S BIGGEST WEAKNESS I once had lunch with Warren Buffett and visited his massive furniture store. (More on this in my free weekly email, link below.) So why would Warren Buffett be so invested in furniture retail? Because it is one of the few retail models with a real moat. But it also has a big weakness. And JD just opened 101HOME in Shanghai to fix that. FURNITURE RETAIL HAS A MOAT A bigger store means more sofas to try, larger warehouses and more delivery throughput. This means big fixed costs which smaller rivals struggle to match. This is how Buffett bought 80% of Nebraska Furniture Mart in 1983 on a handshake, without ever seeing the financials. He could see the moat. BUT THE BIG WEAKNESS IS LOW CUSTOMER FREQUENCY. Chinese families often spend 5-6 weeks working with in store designers on furnishing a new home. Then they buy and disappear for 2-3 years, leaving the retailer with no engagement and no data. JD's NEW 101HOME MODEL SELLS ENGAGEMENT, NOT JUST FURNITURE. It has coffee shops, a French bistro, a bookstore and a rotating Burgundy wine selection This gives people reasons to come back frequently. So furniture earns the margin, while cafes, art and wine create the interactions and the customer data. That's good strategy. ***TAKEAWAYS FOR MANAGERS: → Sort your products into three buckets (margin, engagement and data) → identify the weakest one. I WRITE ABOUT MOATS AND DIGITAL AI STRATEGY EVERY WEEK IN MY FREE AGENTIC FRONTIER EMAIL. Sign up here and get my book Moats and Marathons free: Https://jefftowson.com/free-… QUESTION: What is your engagement product, and how often does it bring customers back?

GENAI HAS NO NETWORK EFFECTS. NOW WHAT? My GenAI strategy for CEOs is basically 4 steps. (the full episode is on Apple Podcasts, link below.) STEP 1: START WITH PRODUCT AND PRODUCTIVITY EXPERIMENTS. Just starting playing with models in your ops. Not customer facing is lower risk. Also, see if you have product risks. Are they obsolete without AI? The first bottleneck here is usually data, not models. STEP 2: BUILD STRATEGIC INTELLIGENCE CAPABILITIES, NOT JUST TABLE STAKES. A bank that builds intelligence into credit approval can pull far ahead of rivals. So you need to identify where AI is a strategic capabilities. Don't just do the basics. that's table stakes. STEP 3: GENAI ALONE IS NOT A MOAT. LEVERAGE ONE YOU ALREADY HAVE. Alibaba is plugging GenAI into Taobao, which already has network effects. That's the best of both worlds. You mostly want to focus on traditional moats. ***TAKEAWAYS FOR MANAGERS: → Name the one intelligence capability that would hurt your competitors most in two years. → Map where your switching costs already live, and put GenAI there first. I GO DEEPER ON THIS IN EPISODE 272 OF THE TECH STRATEGY PODCAST. podcasts.apple.com/us/podcas… QUESTION: Which existing moat in your business could GenAI make stronger?

I call these droids. Right out of Star Wars.

4 COMPETITIVE FORTRESSES NOBODY DARES TO ATTACK A competitive fortress is a business that has 50-70% of a defined market. (More on this in my free weekly email, link below.) And has for a long time. I consider them the super predators of the business world. Nobody dares to attack them. Here are 3 of my 4 favorite competitive fortresses. 1. LOCAL MARKET DOMINANCE WITH NO RIVALS AND NO NEW ENTRANTS. The NBA is my example here. Because true basketball fans globally have no real alternative. Local dominance can mean a geography, a function (like NVIDIA in GPUs), or a customer niche (like the NBA). It comes from targeting a niche and having a combination of competitive advantages. 2. PLATFORMS ARE THE DIGITAL VERSION OF LOCAL DOMINANCE. Windows still has over 75% of desktop operating systems after 40 years. Android has similar dominance in mobile operating systems. This is mostly from network effects and switching costs. 3. LINKED BUSINESSES THAT FORCE A NEW ENTRANT TO COPY EVERYTHING. I like linked businesses. Such as Tencent, which links messaging, payments, media and gaming platforms. And Berkshire Hathaway's linkage between insurance float and investment. Linked businesses put rivals at a disadvantage because they must have multiple businesses to compete. ***TAKEAWAYS FOR MANAGERS: → Define one circumscribed market (a geography, function or customer niche) where you could realistically reach 50% share, and focus your resources there. → List your business units and ask your team which two could share data, bundle pricing or pool product development. I WRITE ABOUT MOATS AND DIGITAL STRATEGY EVERY WEEK IN MY FREE AGENTIC FRONTIER EMAIL. Sign up here and get my book Moats and Marathons free: jefftowson.com/free-email-si… #DigitalStrategy #CompetitiveAdvantage #Moats #Platforms #NetworkEf

CHINESE MEITUAN TRAINED A 1.6T PARAMETER MODEL WITHOUT NVIDIA So it was a food delivery app made China AI independent. Who saw that coming? For those not following, Meituan's Longcat recently made hx as the first AI model trained on domestic GPUs. Here's the basics. 1. LONGCAT 2.0 IS A FRONTIER MODEL TRAINED ONLY ON CHINESE CHIPS. It is a 1.6T parameter MoE model trained on 35T tokens. The training appears to have run on 50k Huawei Ascend 910C chips in 48 SuperPods. 2. MEITUAN HAD FOCUSED ON THIS QUESTION VERY EARLY ON. Its cofounder Wang Huiwen created a business focused on domestic training way back in February 2023. The team focused on getting frontier performance from domestic hardware at 7 nanometers while keeping inference fast. Meituan acquired this biz that June. 3. CHINA NOW HAS AN INDEPENDENT AI TECH STACK. The team is: -Huawei for infrastructure -SMIC for chips -CXMT for memory -Meituan for the open source model (for now). It works, but efficiency is the key question, especially the yield at the foundry. ***TAKEAWAYS FOR MANAGERS: → Add dependence on US frontier models to your risk register. Have an alternate model supplier. → Test one open source Chinese model on a real workflow. Thats's your backup. I GO DEEPER ON THIS TOPIC IN THIS PODCAST. podcasts.apple.com/us/podcas… #Meituan #ChinaTech #AIStrategy #longcat

HUAWEI'S 500,000 CHIP ANSWER TO NVIDIA There is lot of talk about Huawei’s new Mate 90 smartphone, which uses their LogicFolding chip design. But I think everyone is watching the wrong part of Huawei's AI stack. The most important innovations are in the AI Data Centers 1. HUAWEI IS REBUILDING AI INFRASTRUCTURE AT SCALE. Huawei’s new Atlas 960E SuperPoD has 4,096 NPUs working as one giant integrated computer. And these can be combined into a SuperCluster with more than 500,000 NPUs. That’s what’s coming this year. They are matching Nvidia in large scale AI compute. 2. INTERCONNECT INNOVATIONS HAVE BEEN MAKING A BIG DIFFERENCE. Huawei lacks the most powerful NVIDIA GPUs, so it links many weaker chips as peers with its UnifiedBus. And in the 960E, they have used 5,500 new optical engines to replace more than 48,000 pluggable optical modules. This drops latency and doubles fault free operating time. 3. MEMORY IS THE STORY NOBODY TALKS ABOUT. The SuperPod system treats the memory on all chips as one shared pool of 1.1 petabytes. Agents create exploding context, so Huawei extends the KV cache to SSD layers and a single cluster reaches 64 petabytes. This whole AI architecture has been moving really fast for the past 1-2 years. ***TAKEAWAYS FOR MANAGERS: → Ask your infrastructure team whether latency and memory, not chips, are the real limits on your AI workloads. → Estimate how much context your planned agents will carry, then budget memory and storage before you scale them. I WRITE ABOUT AI AGENTS EVERY WEEK IN MY FREE AGENTIC FRONTIER EMAIL. Sign up here and get my book Moats and Marathons free: jefftowson.com/free-email-si… #Huawei #AIInfrastructure #AgenticAI #DataCenters #TechStrategy

TENCENT: 4 REASONS ENTERPRISE AI AGENTS FAIL TO SCALE I left WAIC 2026 thinking most companies are fixing the wrong problem. Tencent Cloud VP Wu Yunsheng argued that enterprises do not lack effective agents. What they lack is a system that integrates those agents into their operational workflows. THER ARE 4 PROBLEMS THAT BLOCK AGENTS AT SCALE. 1: Enterprise Agents Require Proprietary Enterprise Knowledge. You need trustworthy, accurate, and compliant context (i.e., data). That is the foundation for reliable Agent operations. 2: Enterprise Agents Require Production-Grade Stability 3: Enterprise Agents Require Organization-Level Governance 4: Enterprise Agents Need a Sustainable Flywheel You can read more at the full article (in comments). ***TAKEAWAYS FOR MANAGERS: → Audit one core workflow this week. Look for fragmented, outdated or poorly permissioned knowledge before you add another agent. → Decide which steps in that workflow must have fixed logic vs can be handed to autonomous reasoning. ——- I WRITE ABOUT ENTERPRISE AGENTS EVERY WEEK IN MY FREE AGENTIC FRONTIER EMAIL. Sign up here and get my book Moats and Marathons free: jefftowson.com/free-email-si… #AIAgents #EnterpriseAI #Tencent #DigitalStrategy #AgenticAI

WHY SHEIN CAN’T STAY A FASHION SPECIALIST I have followed Shein for +5 years. Its recent prospectus was interesting reading. 1. SHEIN’S ULTRA FAST FASHION MODEL IS STILL A REAL STRENGTH. SHEIN tests new styles in batches of 100 to 200 units and can ramp up production in about five days. It uploads 4,700 new items a day using 370 in house designers and 7,500 contract manufacturers. Impressive. 2. THE FINANCIALS SHOW MARGIN COMPRESSION. Revenue went from $32 billion in 2023 to $42 billion in 2025. But growth has fallen from 100% to roughly 8-10%. Operating margins are about 4%, with fulfillment eating 40-45% of revenue And marketing rising from 10 to 15%. 3. A SPECIALTY ECOMMERCE PLAYER MUST AVOID THE GIANTS, AND SHEIN CANNOT. Fashion is ground zero, with Amazon, Alibaba, JD and Pinduoduo all moving toward it. SHEIN needs to get bigger to compete long term. That means going beyond specialty ecommerce. ***TAKEAWAYS FOR MANAGERS: → Check this week whether your core business sits on the strategic path of a much larger player. If it does, plan your move before they arrive. → Track your fulfillment and marketing costs as a share of revenue over the past three years. Treat a steady rise as an early warning and act on it now. ——— I GO DEEPER ON THIS IN EPISODE 292 OF THE TECH STRATEGY PODCAST. Listen to that podcast at: podcasts.apple.com/us/podcas… #Podcast #Shein #Ecommerce #DigitalStrategy #TechStrategy

DEEP Robotics is great. I like industrial grade quadrupeds as a use case.

NEVER FIGHT FAIR. ALWAYS HAVE A MOAT. Years ago, I opened a couple of restaurants. And it was brutal. (see podcast episode, link below.) It was an operational nightmare. There was endless competition and price wars. I'm never doing that again. I am now 100% against fair fights. I always want a moat. And I've spent a decade writing 7 books about how to do that. Here are 3 take-aways: 1. OPERATING PERFORMANCE MATTERS. BUT MOATS MATTER MORE. Operating performance is about people, systems and technology. It's about speed, effectiveness and hours spent. It's like bike racing. How fast you pedal really matters. However, a moat is like having a better machine. They're riding bikes and you have a car. In my competition hierarchy, operating performance is important. But it's at the bottom of the hierarchy. Moats (barriers to entry and competitive advantages) are at the top. 2. MOAT DO NOT GUARANTEE GROWTH OR PROFIT. Being the only hospital in a small town means you are very well defended. You have no competitors. But you are not going to grow much. So competitive dominance doesn't necessarily mean growth. Similarly, dominance also doesn't guarantee mean profits. Didi and Uber are monopolies / duopolies in ridesharing. Yet they can barely reach operating profit. 3. THE REAL VALUE OF A MOAT IS CONTROL OF YOUR OWN DESTINY. When a rival coffee shop cuts prices 20% or boosts its marketing spend 20%, Starbucks does not respond. Why not? It doesn't have to. It has a moat based on brand, high traffic locations, and economies of scale. Plus the loyalty programs have switching costs. Moats mean you can ignore rivals (somewhat). And you can chart your own course. Your own destiny. Rather than having your future mostly determined by competitor moves. That's invaluable. ***TAKEAWAYS FOR MANAGERS: → Sort your competitors into three groups this week: bigger, same size and smaller. Mark every place where you are on the wrong side of a competitive advantage. → Find one segment where rivals are few or weak. Point your next growth push there instead of at your smartest competitor. For more, please listen to my podcast on this topic: podcasts.apple.com/us/podcas…

70% OF SHOPPERS NOW ASK AI FIRST. TENCENT NOTICED.   I left the Tencent forum at WAIC 2026 rethinking my view of advertising. (More in my free weekly email, link below.)   AI-first marketing (and advertising) is just a sweeping change. It’s stunning. And it’s going to affect everyone selling online.   I had 3 take-aways: 1. AI IS REBUILDING THE USER DECISION PATH. China now has 515M GenAI users and 70% of users already use GenAI when making a purchase. So this is not a theoretical change. It’s real. And happening at scale. Tencent's Xu Wei says the “search and click” model is shifting to a “question, generate, decision” model. 2. CREATIVITY HAS BEEN THE BOTTLENECK IN SCALING ADVERTISING. THAT’S ENDING. Creativity is rare, human made and slow to produce. But consumers burn through content faster every year. And ad lifespans keep getting shorter. This creates a real problem for brands, which must constantly create and refresh in an inefficient way. Plus, they have little data on what content to make. 3. AI AGENTS SOLVE THE CREATIVITY AT SCALE PROBLEM. Rooms of ad execs are being replaced by 1 or 2 people managing an AI system. Take a look at the photo with this post. That’s a good summary. At Tencent, they are pairing Miaosi (a one stop AIGC creative platform) with Marketing AIM+ for automated and intelligent ad placement. Those are two products to look into. ***TAKEAWAYS FOR MANAGERS:   → Audit how your customers use GenAI before they buy, then check whether your brand shows up in those answers. This is gonna be unpleasant.   → Shift one campaign budget this week from creative headcount to computing power, data plus a feedback loop. Just try one.   ----- I WRITE ABOUT AI ADVERTISING AND AGENTS EVERY WEEK IN MY FREE AGENTIC FRONTIER EMAIL.   Sign up here for more. And to get my book Moats and Marathons free:   jefftowson.com/free-email-si…

UBTech’s white robots dance like White guys That’s real attention to detail

SHEIN MUST EVOLVE TO BEAT THE ECOMMERCE GIANTS In this podcast, I evaluate Shein and the question: Can a specialized cross border fashion player remain viable against massive ecommerce rivals? My points: 1. Shein’s Low cost flexible manufacturing (in China) has been a short-term strength Their on demand production model provides enormous variety, which consumers love in apparel. It also gives them maintaining tight control over manufacturing costs. But this can be copied over time. 2. Shein’s thin operating margins make it vulnerable to cash rich giants. international shipping rates and shifting cross border tariffs can also hit this. 3. Specialty ecommerce companies rarely hold durable competitive advantages against marketplace giants. without unique infrastructure, network effects or switching costs, specialty players struggle as big ecommerce platforms replicate their product offering. MY TAKE: Overall, phase 1 of SHEIN has been very successful. But phase 2 will require a bigger business model. They need to evolve. To learn more, listen to our Tech Strategy Podcast here: jefftowson.com/2026/08/shein… ——- About Us: We help businesses launch AI agents fast. Get our free Agentic Frontier email here: agentorgnow.com #Shein #Ecommerce #TechStrategy

WHY JD BUILT 101HOME: THE CASE FOR EXPERIENTIAL FURNITURE RETAIL JD just launched a new flagship furniture format in Shanghai. And I like furniture retail as a business model (note: so does Warren Buffett). I did a deep dive. Here are the takeaways: 1. Home furniture is inherently experiential. Customers don’t buy major pieces without testing them in person. Nobody buys a sofa without sitting on it first. This insulates the category from purely digital competitors (somewhat). 2. Curated physical spaces can make furniture buying an immersive experience. 101HOME has 33 fully furnished dream homes designed by prominent creators. They have designed it as an immersive space. For discovery. For enjoyment. 3. Engagement services solve the low visit frequency problem that plagues furniture retail. The store has art installations, Burgundy wines by the glass, art, pottery, and cafe spaces. Why? To give shoppers reasons to return every month rather than disappearing after a single purchase.  TAKE-AWAYS FOR MANAGERS: -immersive offline environments can be good for tactile products that mobile screens cannot convey. -high frequency products and services (cafes, art, and home accessories) can pull shoppers into low frequency furniture retail spaces. It’s an interesting retail format. Smart. You can learn more here: jefftowson.com/membership_co… ——- About Us: We help businesses launch AI agents fast. Get our free Agentic Frontier email here: agentorgnow.com #JD #Retail #101home