Look towards the place with tenfold growth.

Long time no see, I'm back.

This time, I was away for 15 days. Unlike the trip in April, this time I was more on a journey to address my own and the channel's confusions, chatting with friends along the way.

In Xiaolan, I met several senior entrepreneurs with Brother Hui. Some are already very well-known entrepreneurs in Guangdong. What surprised me was that instead of resting on their laurels, they chose to stay hands-on, seeking the 'second curve' of growth for themselves and their businesses.

Growth is perhaps the best solution to alleviate anxiety, if not the only one.

As consumption gradually returns to its essence, we've witnessed the rise of Mixue Ice Cream & Tea, which is still expanding rapidly with over 10,000 new stores annually. A good business model is about delivering the best products to users as quickly as possible through replicable methods. This, I believe, is the underlying logic behind Mixue becoming one of the 'three pillars' of new consumption.

Chatting with a founder of a kitchen appliance company, he and several other bosses expressed concerns when companies like Xiaomi entered the kitchen appliance sector. Xiaomi employs a 'combo strategy,' attacking from multiple dimensions—product, ecosystem, and sales. Users don’t just need a good range hood; they want it to integrate with other home devices, controllable remotely through phones or voice commands. This represents a new paradigm in the industry, and they are figuring out how to meet users' growing demand for smart solutions.

The premise of growth is entering a 'Big Market.'

When friends ask me about 'Mixue,' I always emphasize not to think of it as just a milk tea brand but to consider it from a 'population' perspective.

Yu Donglai, founder of Pang Dong Lai, once said, 'Don’t just give employees plain water; offer them beverages. Water quenches thirst, but drinks bring joy and satisfaction.' The same logic applies here. Most people might drink milk tea only once or twice a week, but their demand for cola, coffee, or tea could be daily. Viewed this way, these are entirely different markets.

It’s common knowledge that Xiaomi is already in a 'Big Market.' The 'Human-Car-Home Ecosystem' strategy has been repeatedly mentioned during my last trip, signaling that market divergence is fading, and a consensus on Xiaomi’s prospects is forming.

Looking around, where else can we find 'Big Markets' with tenfold growth potential?

In 2023, we witnessed Nvidia’s legendary journey from gaming GPUs to becoming the world’s most valuable company. While Nvidia rakes in nearly $100 billion in annual profits, other AI and large model startups are still struggling.

The reason? Nvidia remains the most solid foundation of the AI era, while other derivative products and business models are like tender shoots on tree branches—they may bloom or wither under harsh conditions, eventually turning into compost.

Investors find it hard to distinguish real opportunities from capital traps in AI. The future is unpredictable, but that doesn’t mean we can’t participate. Simple ethics and values can guide us to make the right decisions.

Truly good projects attract capital effortlessly.

This Spring Festival, DeepSeek emerged out of nowhere. The 'Six Tigers' in foundational large models have now become the 'Four Strong.' Whether it’s training, distillation, or inference, if the user experience is poor, the problem likely lies with the provider.

As investors, we needn’t delve into AI’s technical details. Just experience it firsthand for genuine feedback.

I love what a VP at 4Paradigm once said: 'We don’t teach users how to use AI; we focus on which business problems AI can solve better.' Our priority should be solving problems, then matching the right tools.

We often get intimidated by overvalued startups. In reality, valuations reflect market imagination about a company’s potential. The actual journey is fraught with unforeseen challenges.

Focus on business models, not just concepts.

Good business models can be explained in one sentence; bad ones require a 100-slide deck. There’s nothing new under the sun. Returning to basic principles, we choose our favorite restaurant based on proximity, ambiance, and taste—not because it uses a 'fried rice robot.'

The same goes for AI. Amidst endless new concepts, ask: Whose problems does it solve? Are they willing to pay?

Before DeepSeek, people were willing to foot the bill for billions in annual compute costs for foundational models. But with open-source models outperforming paid ones, both B2B and B2C models are being redefined. Investors are regaining rationality, focusing on where it truly adds value.

As large models shift from training to inference, compute costs will drop significantly. Foundational models become a game for tech giants, while downstream applications benefit. More open-source models from big firms form the 'large model supermarket' of the AI era.

DeepSeek is powerful but can’t solve all industry problems at once. Only deeply optimized, specialized models can address real-world scenarios, like deploying models in intranets.

AI applications are thriving not just in B2C but also B2B. After ChatGPT’s release, Palantir’s market cap surged nearly 20x, briefly exceeding $300 billion—more than Alibaba’s $270 billion at the time.

-----------

That’s about it for the main content. Let me share a few heartfelt words.

In May, I gained some attention and recognition. It’s gratifying to see my efforts acknowledged—nothing beats that.

But as I said at the start, I can’t rest on my laurels. Returning to my original purpose for writing, I must keep moving forward, creating valuable content, sharing insights, and offering fresh perspectives.

Happy weekend. June 20, Mars at Qinghe.

$PHANCY(06682.HK) $MIXUE GROUP(02097.HK) $XIAOMI-W(01810.HK)