CATL made 64.5 billion yuan in the first nine months, but its share price fell nearly 3% that day, while its Hong Kong-listed shares plunged more than 8% intraday.
The third-quarter results came out on the evening of October 20. Revenue for the first nine months was 442.1 billion yuan, up 56% year over year. Net profit attributable to shareholders was 64.53 billion yuan, up 31%.
The third quarter alone looked a bit rough. Revenue was 165.2 billion yuan, up 58%. Net profit was 21.2 billion yuan, up just 14%. That’s more than a threefold gap. However you look at the numbers, something doesn’t add up.
The more they sell, the thinner the profits. Last year, analysts were still just worried about this. This year, it’s right there in the financial statements. The third-quarter net margin was 12.9%, down from 15.3% in the second quarter and 17.8% a year earlier. Nearly five percentage points gone in a year. A pretty steep drop, and yet nobody’s saying a word.
Look closer, and it comes down to profit per watt-hour. CATL sold just over 260 gigawatt-hours of power batteries and energy storage in the third quarter. Brokerage estimates, using an integrated basis, put net profit at 0.082 yuan per watt-hour, down from 0.096 yuan the previous quarter. For a 60-kWh battery pack, that works out to just 4,900 yuan in net profit—and it’s set to fall further. That’s frighteningly thin. How many more years can this business hold up? I’m not so sure.
A lot of people think battery makers make money from technology. I’ve been thinking they’ve always made money from scale. And nothing scares scale like a price war. When the price war really arrives, with margins this thin, it’s hardly a surprise.
So instead, CATL is putting its money into other things. What things? Take a guess. Since the start of this year, it’s invested more than 10.5 billion yuan in power-computing coordination, and every deal has been decisive. In August, it injected 4.1 billion yuan into the parent company of Zhongheng Technology, indirectly acquiring a 17% stake in Zhongheng Electric. The company holds a 31% share of the domestic HVDC power supply market for AI computing centers, ranking first. Its customers include Alibaba, Tencent, ByteDance, and China’s three major telecom operators.
In May, it bought up to 38% of VNET Group for 6.4 billion yuan, becoming its largest shareholder. VNET has a presence in more than 30 cities, 50 data centers, 889 megawatts of wholesale capacity, and 49,000 retail cabinets.
In June, CATL’s ecosystem invested 5 billion yuan in DeepSeek’s 51-billion-yuan funding round, making it the second-largest external investor, behind only Tencent.
Put the three investments together and you can see the plan clearly. Electricity comes off the grid and needs energy storage to shave peaks. HVDC is needed to deliver it to the server rooms. Those rooms need somewhere to go, and the models need someone to run them. CATL has been buying up, one by one, several of the tollbooths along that route.
Changjiang Securities has estimated that electricity accounts for 60% to 70% of the operating costs of large language models. Buying data centers, power supplies, and models is, when you get down to it, buying access to electricity. It really is that straightforward.
The connection to crypto isn’t batteries—it’s electricity prices, which nobody can escape. Bitcoin mining was the first industry in the world to turn electricity into computing power and sell it for money. Miners switch on wherever electricity is cheap. As the network’s hashrate rises, difficulty adjusts automatically and earnings get diluted. It makes no difference whether you sell battery cells or integrated circuits. There’s no negotiating around it. This mechanism has been running for more than a decade, and the numbers have always balanced.
But this whole setup has always depended on cheap electricity. After China’s crackdown in 2021, miners moved to Texas, Kazakhstan, and Central Asia. What they were chasing was electricity prices. They went wherever power was cheapest. Now energy storage and data centers are starting to compete for the same kilowatt-hours—and they’re doing it by signing long-term contracts, securing allocations, and bringing supporting infrastructure. A gigawatt-scale data center has a flat load curve and runs 24/7 without interruption. Whoever signs the long-term contract first locks up those megawatts. The kind of load miners can switch off at a moment’s notice is a different proposition in the eyes of the grid. It has value, but not that much. There’s a gap.
Ethereum stopped using electricity in 2022 and switched to staking. It doesn’t burn a single kilowatt-hour. Bitcoin still burns electricity, and it has to keep doing so. That can’t be changed. The more miners there are, the higher the difficulty and the thinner the returns per terahash—exactly the same principle as CATL’s 0.082 yuan per watt-hour. CATL spreads its costs across production capacity; miners spread theirs across hashrate. Either way, the margins get thinner.
If you own mining rigs, or you’re calculating electricity costs, keep an eye on how capacity electricity prices are set next year. Once they’re set, the value of interruptible load will become clear, and so will miners’ ability to compete with data centers for power. That’s what really matters.
The public disclosures don’t spell out how these investments will be consolidated, or whether energy-storage prices will keep falling, even though every deal has been announced. My guess is we’ll see that 0.082 figure again in next year’s annual report.
$BTC $ETH $TAO
#比特币挖矿 #储能 #ComputingPower
The third-quarter results came out on the evening of October 20. Revenue for the first nine months was 442.1 billion yuan, up 56% year over year. Net profit attributable to shareholders was 64.53 billion yuan, up 31%.
The third quarter alone looked a bit rough. Revenue was 165.2 billion yuan, up 58%. Net profit was 21.2 billion yuan, up just 14%. That’s more than a threefold gap. However you look at the numbers, something doesn’t add up.
The more they sell, the thinner the profits. Last year, analysts were still just worried about this. This year, it’s right there in the financial statements. The third-quarter net margin was 12.9%, down from 15.3% in the second quarter and 17.8% a year earlier. Nearly five percentage points gone in a year. A pretty steep drop, and yet nobody’s saying a word.
Look closer, and it comes down to profit per watt-hour. CATL sold just over 260 gigawatt-hours of power batteries and energy storage in the third quarter. Brokerage estimates, using an integrated basis, put net profit at 0.082 yuan per watt-hour, down from 0.096 yuan the previous quarter. For a 60-kWh battery pack, that works out to just 4,900 yuan in net profit—and it’s set to fall further. That’s frighteningly thin. How many more years can this business hold up? I’m not so sure.
A lot of people think battery makers make money from technology. I’ve been thinking they’ve always made money from scale. And nothing scares scale like a price war. When the price war really arrives, with margins this thin, it’s hardly a surprise.
So instead, CATL is putting its money into other things. What things? Take a guess. Since the start of this year, it’s invested more than 10.5 billion yuan in power-computing coordination, and every deal has been decisive. In August, it injected 4.1 billion yuan into the parent company of Zhongheng Technology, indirectly acquiring a 17% stake in Zhongheng Electric. The company holds a 31% share of the domestic HVDC power supply market for AI computing centers, ranking first. Its customers include Alibaba, Tencent, ByteDance, and China’s three major telecom operators.
In May, it bought up to 38% of VNET Group for 6.4 billion yuan, becoming its largest shareholder. VNET has a presence in more than 30 cities, 50 data centers, 889 megawatts of wholesale capacity, and 49,000 retail cabinets.
In June, CATL’s ecosystem invested 5 billion yuan in DeepSeek’s 51-billion-yuan funding round, making it the second-largest external investor, behind only Tencent.
Put the three investments together and you can see the plan clearly. Electricity comes off the grid and needs energy storage to shave peaks. HVDC is needed to deliver it to the server rooms. Those rooms need somewhere to go, and the models need someone to run them. CATL has been buying up, one by one, several of the tollbooths along that route.
Changjiang Securities has estimated that electricity accounts for 60% to 70% of the operating costs of large language models. Buying data centers, power supplies, and models is, when you get down to it, buying access to electricity. It really is that straightforward.
The connection to crypto isn’t batteries—it’s electricity prices, which nobody can escape. Bitcoin mining was the first industry in the world to turn electricity into computing power and sell it for money. Miners switch on wherever electricity is cheap. As the network’s hashrate rises, difficulty adjusts automatically and earnings get diluted. It makes no difference whether you sell battery cells or integrated circuits. There’s no negotiating around it. This mechanism has been running for more than a decade, and the numbers have always balanced.
But this whole setup has always depended on cheap electricity. After China’s crackdown in 2021, miners moved to Texas, Kazakhstan, and Central Asia. What they were chasing was electricity prices. They went wherever power was cheapest. Now energy storage and data centers are starting to compete for the same kilowatt-hours—and they’re doing it by signing long-term contracts, securing allocations, and bringing supporting infrastructure. A gigawatt-scale data center has a flat load curve and runs 24/7 without interruption. Whoever signs the long-term contract first locks up those megawatts. The kind of load miners can switch off at a moment’s notice is a different proposition in the eyes of the grid. It has value, but not that much. There’s a gap.
Ethereum stopped using electricity in 2022 and switched to staking. It doesn’t burn a single kilowatt-hour. Bitcoin still burns electricity, and it has to keep doing so. That can’t be changed. The more miners there are, the higher the difficulty and the thinner the returns per terahash—exactly the same principle as CATL’s 0.082 yuan per watt-hour. CATL spreads its costs across production capacity; miners spread theirs across hashrate. Either way, the margins get thinner.
If you own mining rigs, or you’re calculating electricity costs, keep an eye on how capacity electricity prices are set next year. Once they’re set, the value of interruptible load will become clear, and so will miners’ ability to compete with data centers for power. That’s what really matters.
The public disclosures don’t spell out how these investments will be consolidated, or whether energy-storage prices will keep falling, even though every deal has been announced. My guess is we’ll see that 0.082 figure again in next year’s annual report.
$BTC $ETH $TAO
#比特币挖矿 #储能 #ComputingPower