Author: danny

Most people who have been to Thailand have probably been to 7-Eleven. You've probably bought food there—shrimp wontons, fried chicken, sausages, a box of fried rice, eggs, peanuts... If you look closely at the packaging, you'll see two letters: CP.

It is the CP Group.

Most tourists will only see it once on the packaging; but Thais may see it a dozen times a day.

Thais face three things in their lives: life and death, taxes, and the CP Group.

But it's much more than just that bag of shrimp wontons. The 7-Eleven you're in is franchised by them; the phone card you casually buy might be from True, a company they control; the Makro supermarket downstairs is theirs; and the chicken you eat, from feed and breeding stock to slaughter and processing, has likely passed through their hands. In Thailand, you could spend an entire day within their network—eating their eggs in the morning, buying lunch at their convenience store, using their internet to make calls in the afternoon, and buying groceries at their market in the evening.

When a company infiltrates to this extent, it is no longer just a company in the Thai economy; it becomes a part of the economy itself.

And it all started with a bag of rapeseed.

1. A bag of rapeseed

The story begins more than a hundred years ago.

In the late 19th century, Chaoshan was a region with limited land and a large population, with Chenghai County, located by the sea, being particularly poor. There, a type of ocean-going sailing ship with a red-painted prow was called a "red-headed ship" by the Chaoshan people. These ships departed from Zhanglin Port, carrying people who could no longer survive to Southeast Asia. Most of the passengers carried only a few clothes and some travel money. Upon arriving in Siam, Malaya, and Singapore, they first worked as porters at the docks and then as laborers in rice shops. In Chaoshan dialect, this journey was called "going abroad."

That's how Xie Yichu arrived in Bangkok. In 1921, he and his brother Xie Shaofei rented a small shop on Yaowarat Road in Chinatown. The shop was called Zhengdazhuang and they sold vegetable seeds that were transported from their hometown of Shantou.

Why choose the "seed selling" business? Bangkok has a large Chinese population, and their tastes in vegetables are similar to those in Lingnan (Guangdong and Guangxi), requiring Chinese seeds. However, these seeds aren't available locally in Siam and have to be shipped in. More importantly, seeds are unpredictable: you can't tell their quality when you buy them; you only know if they were good or not, and whether you can make money, after they've been sown, sprouted, and grown into vegetables. A farmer who buys a packet of seeds is gambling on their entire harvest and their family's livelihood .

A farmer bought his seeds, sowed them, and that season the seedlings sprouted evenly, the vegetables grew well, and he sold a lot, making a lot of money. He came back the following season, and not only that, he also brought other villagers along. The season after that, it was a new batch of faces. In the next season, seeing how much they were making, others gathered around.

The name "Zhengda" (正大) has become a trademark among Chaozhou natives and Thai farmers, simply by selling packets of seeds. What Xie Yichu displays on his counter are rapeseed seeds; what he truly sells is the promise: "Trust me, and you'll make money." (Chaozhou pinyin: sìn uâi，ū zěng / gàn)

This sentence is worth far more than the seed.

II. A Chicken

The one who turned this seed plantation into a multinational corporation was Xie Yichu's youngest son, Xie Guomin.

Xie Yichu named his four sons Xie Zhengmin, Xie Damin, Xie Zhongmin, and Xie Guomin. The middle four characters, when read together, form "Zhengda China"—a Teochew man who did business in Southeast Asia, wrote his longing for his homeland into his sons' names. (By the way, most people probably first became familiar with the CP Group through the variety show "Zhengda Variety Show" 🤣) When the youngest son, Guomin, took over, CP had already expanded from selling seeds to selling animal feed.

Dhanin Chearavanont took the next step and completely changed the whole chicken farming business.

From hatching to market, a broiler chicken goes through several stages: breeding stock, feed, disease prevention, fattening, slaughter, and sales. For a farmer working alone, any one of these stages could be disastrous—a bird flu outbreak could wipe out their entire fortune overnight, or a drop in chicken prices could lead to the same financial ruin. Xie Guomin's solution is to control both ends: he handles the upstream breeding stock and feed, and the downstream slaughtering, processing, and sales. The most expensive, labor-intensive, and risky stage—fattening—is left to the farmers. The farmers provide the chicken coops and labor, while he supplies the chicks, feed, and technicians. Once the chickens are grown, he buys them back at a pre-agreed price.

In this deal, both sides got what they wanted. Farmers gained certainty—whether chicken prices rose or fell, payment would be made according to the contract, eliminating the need to gamble on market fluctuations; however, the burden of disease control, daily management, and the cost of purchasing chicks still fell on their own shoulders. Dhanin Chearavanont gained "leadership power" over the entire industry chain: what breed to raise, how many to raise, what feed to use, and what standards to use for slaughter—all were determined by CP Group. The profit margin per chicken was a minor matter; his control over the entire chain was the real issue. Even more effortlessly, this chain was built on the land, labor, and capital of tens of thousands of farmers—essentially expanding by leveraging someone else's resources, which was far faster than accumulating wealth penny by penny on their own.

Let's say there's a farmer named Somchai on the outskirts of Chiang Mai. He wants to raise chickens, but he can't afford breeding stock, can't get enough feed, and can't afford the fluctuations in chicken prices.

CP Group then offered him a contract: CP Group would provide the chicks, feed, vaccines, and technicians; and CP Group would buy back all the chickens once they were fully grown. Somchai only needed to build the chicken coops. He then used the land inherited from his ancestors to apply for a bank loan.

From that day on, he was nominally an independent farmer, but in reality, he had become part of the CP Group's industrial chain. From then on, Somchai's chicken coop could only raise CP Group's breeds, feed them CP Group's feed, be sold to CP Group according to CP Group's standards, and sell the chickens only to CP Group; but if bird flu were to strike one day, the coop would be empty, the chickens would die, and he would be unable to repay his bank loans.

Thousands upon thousands of chickens form the foundation of CP Group's chain. CP Group doesn't need to build its own chicken coops, hire its own farmers, or bear the losses from bird flu affecting any particular shed—it only manages the two ends: supplying chicks and feed, and buying and selling chickens. A single contract clearly defines the people and risks involved. This is the most practical form of the "company + farmer" model, and also the smartest aspect of this business: expansion is achieved through the chicken coops and loans of thousands upon thousands of chickens, and the risk is also placed on their chicken coops and loans. This model outsources the most dispersed, difficult-to-manage, and labor-intensive aspects to farmers.

III. A Country

Chickens can be raised, harvested, and sold; the rest is about applying this logic in all directions.

Downstream, they turned chicken into dishes, resulting in the CP shrimp and CP chicken rice you see at 7-Eleven. To sell these products, they entered the retail sector: securing the exclusive franchise for 7-Eleven in Thailand, making convenience stores a ubiquitous infrastructure, and acquiring supermarkets and wholesalers—the supermarkets were called Lotus Supermarket in China (now called "CP Lotus"), "Easy" being Xie Yichu's name; the son used his father's name to brand the retail flagship. Further outwards, they entered the telecommunications industry with True; and the financial sector, becoming a major foreign shareholder of Ping An Insurance.

Having achieved this level of success, I believe no one would disagree that CP is a pillar enterprise in Thailand. It operates tens of thousands of 7-Eleven stores in Thailand, making it the largest 7-Eleven operator outside of Japan; it supports and employs tens of thousands of contract farmers and hundreds of thousands of people; a significant portion of the eggs, chicken, and pork on Thai tables originates from its breeding stock and feed. From morning till night, a Thai person's food, purchases, daily necessities, and internet access all repeatedly come from the same company.

At this size, it becomes inextricably linked to Thai politics, regardless of who comes to power. The prices it sets for feed and food directly impact inflation, farmers' incomes, and the livelihoods of hundreds of thousands. Its rise and fall are variables on the macroeconomic scale, influencing every government. Conversely, many of its businesses—convenience stores, telecommunications franchises, large-scale infrastructure projects—require government licenses and approvals to be implemented, making it equally dependent on the government. Both sides are indispensable to each other.

Thailand has experienced a series of coups over the decades, with civilian and military governments alternating in power. The CP's strategy has been to avoid siding with either side and maintain good relations with both. Benefits have flowed down through concessions and large-scale projects: 7-Eleven franchises, telecommunications licenses requiring state approval; the high-speed rail connecting three major airports, secured by a consortium led by the CP; the later merger of True and DTAC, splitting the mobile market into two entities, all approved; and the acquisition of Lotus retail stores, approved despite concerns about monopolies. Each of these deals wasn't won through market competition, but through negotiations with the state. And then there's another layer—the CP is the company that obtained China's first foreign investment license (number 0001) in 1979, and for decades has been one of the most important economic corridors between Thailand and China, maintaining close ties with Beijing.

Over the past century, from a single bag of rapeseed to where it is today, the path is clear: securing a crucial entry point where buyers can only trust you → vertical integration to control output → contracting out the most resource-intensive and costly stages to contracted farmers and leveraging their resources for expansion → integrating into daily life and becoming inextricably linked to the state. Each step amplifies the control exerted in the previous one. After these four steps, a single bag of rapeseed has miraculously grown into a force that holds immense power over the entire nation's industrial chain.

Here's the interesting part: the same path is being retraced step by step by the AI ​​computing power chain today.

IV. Chips are like new rapeseed.

Let's look at the entry point first. Chips are like the seeds carrying "profit and hope"—the FLOPs on the specification sheet don't equal the effective computing power after you actually build the cluster and run the model. Interconnect efficiency, utilization, stability and yield after scaling up—all of these can only be known after the machines are set up and run through a cycle, just like seeds need to sprout. The sources are equally narrow: NVIDIA designs, TSMC manufactures, HBM is one of only three, and there's only ASML for lithography machines. You can't cultivate them locally; you can only get them from these few companies.

However, chips have an additional layer of control over the supply compared to rapeseed. Farmers can use CP seeds this year and switch to another supplier if they don't like them next year; but if they buy NVIDIA cards, the developers have millions of lines of code, the entire operator library, and the whole toolchain built on CUDA, making it impossible to switch. CP not only monopolizes the reliability of the seeds, but also the very tool used in the field. CP has never been able to secure this double lock-in on the seed side—back then, seeds were low-profit and easily replaceable, while CUDA is neither.

The "company plus farmer" model fits perfectly. NVIDIA supplies not only GPUs, but also reference architectures and CUDA, sometimes even investing capital directly—this is like providing the seeds, the feed, and the technicians, plus credit. Neocloud and those sovereign and regional data centers provide capital and handle operations and maintenance—like farmers providing the chicken coop and labor. The long-term computing power buyback contracts signed between mega-factories and model labs simply mean buying back at the agreed-upon price. It's all about using other people's balance sheets to expand capacity.

Back then, the scarce permits that Chinese computing power (CP) sought were state-issued licenses and large-scale projects. Today, those permits are for electricity, land, grid interfaces, and chip quotas—all still held by the state. Whoever enters a region rich in electricity and capital first gains a first-mover advantage, not through market competition, but through negotiations with the government. CP used these licenses to expand its industrial chain across Thailand; today, computing power players are deploying their clusters across various countries through sovereign AI agreements. At the very top of this chain, from beginning to end, lies the state.

Seeing this, doesn't it look a bit like AI copying the CP (couple) system exactly as it is? However, there are two places here that are exactly the opposite.

V. Two reversed places

First, things that depreciate in value are in the wrong place.

In the livestock farming industry chain, the cheap and easily consumed items are the inputs (chickens, rapeseed), while the valuable and durable items are the assets (land, chicken coops). Farmers who hold assets are less likely to see their capital depreciate.

In the AI ​​field, the situation is reversed: the most expensive GPUs are the ones that depreciate the fastest, lasting only two or three years (the exact number is still debated within the industry, but no one believes they can last for decades like factory buildings). Once the next generation is released, the previous generation immediately drops in value. And holding onto this rapidly depreciating asset is Neocloud, which acts as a data farmer.

Even if Xie Guomin's farmers lose everything, they still have a piece of land left; Neocloud's server racks, after two or three years, are left with a bunch of old cards that can only be sold at a discount. The farmers are in the same location, but the things they have are worlds apart.

Secondly, the quality of downstream demand is reversed.

Why is CP's recycling stable? Because the downstream is supported by essential needs like playing games, which require daily consumption, repurchase, and continuous operation. The output that the entire chain needs to produce is significantly less than the amount that urban residents buy with real money. Behind the contracts is a visible and attainable cash flow.

The situation with AI is more complicated. It's not that the demand doesn't exist—ChatGPT, Claude, and Copilot are already collecting substantial subscription fees and corporate funding, so saying "the demand for AI hasn't been proven" is going too far. The real unresolved issue is a step back: will the profits these companies generate be enough to cover the capex capital expenditure that's been poured into computing power globally today? The demand downstream of computing power providers clearly exceeds the output of this chain; the demand downstream of AI is real, but whether it's enough to support the investment in this chain is still unknown.

Secondly, there's a concept of an "internal circulation" within the computing power industry chain: NVIDIA invests money in companies that buy its cards, and these companies use this money, plus debt, to build clusters and buy more cards. Their revenue depends on procurement commitments from the labs, and a portion of the money for lab expansion comes from the same upstream suppliers and investors. The money circulates within this circle, adding a demand and a revenue entry to the books. This circle is real; it inflates the apparent demand. However, real money is also flowing in from outside the circle—cost reduction for enterprises, private deployments, and subscriptions for consumers. It's currently a race: can the rate at which real revenue from outside the circle increases keep up with the depreciation rate of GPUs every two to three years? If real money hasn't filled the gap before the old cards become scrap metal, the long-term underwriting agreement is just an empty promise; if real money fills the gap quickly enough, the chain will stand on its own.

The same framework acts as a stabilizer for content providers (CPs) – distributing the risk to those who can withstand it, while real demand is supported underneath; but for AI, it acts as an amplifier – once the real money from outside the industry underperforms depreciation, assets, debts, and contracts will all shrink.

VI. Another possibility

However, we must also consider this: the above reasoning rests on a crucial premise: the demand for AI cannot keep up with the production capacity already invested.

Of course, this premise may not hold water; it simply offers another perspective on the issue.

Railways, power grids, and fiber optics all followed the same path. Production capacity was built before demand, and the first wave of investors were buried by the bubble. But the rails, cables, and fiber optics survived, and demand eventually caught up little by little. The dark fiber optic cables laid in the 1990s were in surplus for many years before being fully utilized. If AI inference eventually becomes a daily necessity like electricity and bandwidth, those offtake contracts that are considered speculative today will become stable cash flow in the future; the rapid depreciation of GPUs will no longer be a problem because they will run under high load every day, and the depreciation will be amortized. This path exists, and the probability is not small.

So we're not going to gamble on which path will work. What we want to talk about is something else: regardless of who wins the race for demand in the end, the risks and cash flow that each position in this chain bears are different from the very beginning, and this distribution is fixed the day the industry chain is built.

VII. Who is standing in which position?

The power in an industry chain is never about scale itself, but rather about how risks and cash flow are distributed.

Many people believe that power in the supply chain comes from scale. In reality, scale is merely a result. What truly determines power is how risk and cash flow are distributed.

Whoever receives the cash flow bears the depreciation; whoever bears the demand fluctuations bears the financing pressure.

This is where the power of the industrial chain grows.

So, who is playing the role of the CP (couple)?

The most valuable position for content providers (CPs) has never been at the seed end. While they did start with seeds, they truly consolidated their power through demand entry points like 7-Eleven, Lotus's, and Makro, by connecting the entire supply chain. And throughout this journey, they never carried the heaviest assets themselves. NVIDIA occupies a different position—a seed provider locked in by CUDA, nearly monopolistic, and taking the lion's share of profits. This position is incredibly valuable, far beyond what those who supplied seeds to CP could ever dream of, but it's entirely different from the position CPs ultimately secured. Those who truly occupy the CP's position are AI platforms and model labs that simultaneously control demand entry points, user relationships, and cash flow—they connect to the 7-Eleven end, while NVIDIA is the seed end.

The remaining position—where the farmer pays the money, handles maintenance, holds the fastest-depreciating asset in the entire chain, and lives off a single contract—is occupied by neocloud and sovereign computing power.

Today's computing power buyers are far more aggressive than the Thai farmers of yesteryear. To secure a monopolistic supply of resources, they've mortgaged their still-developing GPUs—pledged to private lending firms, various structured financing options, leveraging several times their initial investment—to buy even more cards. They're holding the fastest-depreciating assets in tech history, burdened with debt that's unlikely to be forgiven. Caught between monopolistic seed companies and labs that haven't even figured out their own financials, they appear as glamorous computing power upstarts, but in reality, they're using their own balance sheets to bail out the entire chain. So when something goes wrong at the terminal, they face a triple whammy: asset devaluation, debt maturities, and contract cancellations, with no buffer in between.

To put it bluntly, Neocloud is the Somchai of the AI ​​era. Somchai gambled with a chicken coop and a bank loan, while Neocloud gambled with cabinets of GPUs that would depreciate in two or three years and leverage several times over. But their fallback plans are vastly different: even if Somchai's coop is empty, the land beneath his feet is still there; once Neocloud's cards become obsolete, all that's left is a pile of electronic waste, burdened with an insurmountable debt. Both are at the bottom of the chain, but the AI ​​version is stripped bare.

The real value of CP Group's story has little to do with whether it deals in chickens or seeds. Its most valuable asset is that it has found the most comfortable position in the entire industry chain: controlling the demand entry point and integration power, while leaving the heaviest and most depreciating assets on other people's accounts; holding onto cash flow and avoiding the lowest level of operational risks.

The AI ​​computing power chain is also searching for this position today. Ultimately, the winner won't be who has the most GPUs, but who can put the fastest-depreciating assets on someone else's balance sheet while simultaneously securing the most stable demand. Power in the industry chain is never about scale itself, but about how risk and cash flow are distributed.

According to this standard, the conclusion actually has nothing to do with whether AI wins or loses the demand race. Those who occupy the entry point and demand are replicating the power that CP Group has grasped—if demand wins, they reap the rewards; if demand loses, they suffer the least. Those who pay to bear the depreciation of assets and live off contracts are replicating the risks that CP Group never lets its own farmers touch—if demand wins, they get a share; if demand loses, they are the first to be eliminated.

Which path leads to success, who reaps the rewards, and who suffers the losses—that was already predetermined the day the industrial chain was established.

A hundred years ago, Somchai mortgaged his land and chicken coop. Today, the new Somchai mortgages GPUs and leverage. Both believe they are riding a wave of the times.

The only difference is:

Somchai represents the chicken and eggs that Thais eat every day.

What exactly is the demand behind today's conjecture? No one can give the same definitive answer yet.