1 00:00:02,720 --> 00:00:07,200 Speaker 1: Bloomberg Audio Studios, podcasts, radio news. 2 00:00:08,360 --> 00:00:11,320 Speaker 2: Earlier this year, our colleague James Mager took a trip 3 00:00:11,360 --> 00:00:13,560 Speaker 2: to the middle of the desert in the remote region 4 00:00:13,640 --> 00:00:16,560 Speaker 2: of Shinjong, China. 5 00:00:16,760 --> 00:00:22,919 Speaker 3: Shinjung is the western third of China. Basically it borders Mongolia, Kazakhstan, 6 00:00:23,239 --> 00:00:27,520 Speaker 3: Uzebekistan and where we went on this trip is situated 7 00:00:27,520 --> 00:00:30,880 Speaker 3: in a river valley. It's got these lovely snow capped 8 00:00:30,920 --> 00:00:35,280 Speaker 3: mountains overlooking the town. Really was a stunningly beautiful place, 9 00:00:35,760 --> 00:00:37,080 Speaker 3: but very hard to get to. 10 00:00:37,960 --> 00:00:40,680 Speaker 2: James was there on a lead generated by Bloomberg's data 11 00:00:40,720 --> 00:00:44,479 Speaker 2: journalism team that a handful of Chinese companies were building 12 00:00:44,520 --> 00:00:47,879 Speaker 2: some forty data centers across the region and planned to 13 00:00:47,960 --> 00:00:51,960 Speaker 2: power them with tens of thousands of Nvidia chips. Chips 14 00:00:52,040 --> 00:00:55,000 Speaker 2: that the US banned from being exported to China in 15 00:00:55,040 --> 00:00:58,160 Speaker 2: twenty twenty two is as well. China has criticized the 16 00:00:58,280 --> 00:01:02,400 Speaker 2: US moves to expand restrictions one is access to sympiconductor technology, 17 00:01:02,520 --> 00:01:05,640 Speaker 2: saying that they will harm supply chains and the world economy. 18 00:01:05,920 --> 00:01:10,520 Speaker 1: They want to build a AI industry that can go 19 00:01:10,640 --> 00:01:12,120 Speaker 1: head to head with the US. 20 00:01:13,040 --> 00:01:16,440 Speaker 2: Bloomberg data reporter Andy Linn led the team's investigation. 21 00:01:17,040 --> 00:01:21,680 Speaker 1: In modern warfares, AI is playing a larger role, so 22 00:01:22,160 --> 00:01:26,440 Speaker 1: the US is worried about China developing the AI capacity 23 00:01:26,880 --> 00:01:30,600 Speaker 1: with developing their high end military must be. 24 00:01:30,640 --> 00:01:36,920 Speaker 2: In Since the Biding Administration's initial restrictions on advanced AI chips, 25 00:01:37,200 --> 00:01:40,520 Speaker 2: the US has tightened export controls, putting them at the 26 00:01:40,600 --> 00:01:45,080 Speaker 2: center of tensions with Beijing. Just last week, Bloomberg learned 27 00:01:45,080 --> 00:01:48,200 Speaker 2: of plans by the Trump administration to further restrict the 28 00:01:48,240 --> 00:01:52,160 Speaker 2: shipments of AI chips to Thailand and Malaysia, part of 29 00:01:52,200 --> 00:01:55,840 Speaker 2: an effort to prevent them from being potentially smuggled into China. 30 00:01:56,440 --> 00:01:59,880 Speaker 1: So by building these data centers and build up the 31 00:02:00,040 --> 00:02:05,360 Speaker 1: computing power, they aim to build a local, domestic EI 32 00:02:05,400 --> 00:02:08,560 Speaker 1: industry that can go ahead to have with opening AI 33 00:02:08,680 --> 00:02:11,280 Speaker 1: Alphabet Amazon Meta in the US. 34 00:02:15,200 --> 00:02:18,840 Speaker 2: This is the Big Take Asia from Bloomberg News. I'm Wanha. 35 00:02:19,040 --> 00:02:21,520 Speaker 2: Every week we take you inside some of the world's 36 00:02:21,520 --> 00:02:25,280 Speaker 2: biggest and most powerful economies and the markets, tycoons, and 37 00:02:25,360 --> 00:02:29,600 Speaker 2: businesses that drive this ever shifting region. Today on the show, 38 00:02:30,000 --> 00:02:40,919 Speaker 2: China's AI ambition rises from the desert. China is fast 39 00:02:40,960 --> 00:02:44,520 Speaker 2: becoming a world leader in artificial intelligence it's one of 40 00:02:44,520 --> 00:02:49,000 Speaker 2: the top consumers of semiconductors, and as Deepseek's latest model shows, 41 00:02:49,280 --> 00:02:53,360 Speaker 2: Chinese companies are hungry to create top tier AI technology 42 00:02:53,639 --> 00:02:58,799 Speaker 2: to rival international competitors. But Andy says the US government's 43 00:02:58,880 --> 00:03:02,519 Speaker 2: export controls have created a real problem for China by 44 00:03:02,520 --> 00:03:06,040 Speaker 2: limiting its access to Nvidia's priced semiconductors. 45 00:03:07,280 --> 00:03:12,919 Speaker 1: Nvidia is the industrial standard for any AI training operations, 46 00:03:12,960 --> 00:03:16,120 Speaker 1: so for example, Open Ai, the most famous AI companies 47 00:03:16,120 --> 00:03:19,560 Speaker 1: in the world, use Nvidia chips for training the chat, 48 00:03:19,600 --> 00:03:25,359 Speaker 1: GPT models and whatever. You want to build a sophisticated 49 00:03:25,680 --> 00:03:29,000 Speaker 1: AI model, you need to use a lot of Nvidia chips. 50 00:03:29,480 --> 00:03:32,600 Speaker 1: So MVA chip is sought after across all the world 51 00:03:32,760 --> 00:03:36,720 Speaker 1: for all the AI development because every country wants their 52 00:03:36,840 --> 00:03:38,760 Speaker 1: own AI. 53 00:03:38,080 --> 00:03:42,200 Speaker 2: Industry, and he says China's ultimate goal is to have 54 00:03:42,240 --> 00:03:45,800 Speaker 2: its own source of advanced AI chips, ones that are 55 00:03:45,800 --> 00:03:50,680 Speaker 2: on par with Nvidia's offerings. For now, though, in semiconductors 56 00:03:50,760 --> 00:03:54,040 Speaker 2: are the best product on the market and Chinese companies 57 00:03:54,080 --> 00:03:55,840 Speaker 2: are eager to access them. 58 00:03:56,440 --> 00:04:01,480 Speaker 1: Since the US introduced these export controls, I heard sources 59 00:04:01,520 --> 00:04:04,440 Speaker 1: in China saying they are having a hard time getting 60 00:04:05,240 --> 00:04:10,200 Speaker 1: large volumes of Nvidia chips for their operations. I think 61 00:04:10,280 --> 00:04:14,960 Speaker 1: this slow China's access to these chips, but not stop 62 00:04:15,080 --> 00:04:17,120 Speaker 1: them from getting them outright. 63 00:04:18,520 --> 00:04:22,039 Speaker 2: The US government has no official consensus on how many 64 00:04:22,080 --> 00:04:26,279 Speaker 2: restricted Nvidia chips are currently in China. Two senior Biding 65 00:04:26,320 --> 00:04:30,200 Speaker 2: officials estimated that China had around twenty five thousand chips, 66 00:04:30,800 --> 00:04:33,080 Speaker 2: but most of the people Bloomberg spoke to say there 67 00:04:33,120 --> 00:04:36,479 Speaker 2: isn't an agreed upon estimate. That's why when Andy found 68 00:04:36,480 --> 00:04:40,080 Speaker 2: documents laying out the volume of Nvidia chips that companies 69 00:04:40,200 --> 00:04:42,800 Speaker 2: hope to obtain for these data centers, he was. 70 00:04:42,760 --> 00:04:48,240 Speaker 1: Surprised because most of the people I talked to would 71 00:04:48,320 --> 00:04:52,440 Speaker 1: say that they assume some developments of these data standard 72 00:04:52,560 --> 00:04:58,320 Speaker 1: using Nvidia chips is ongoing, but not like declaring to 73 00:04:58,960 --> 00:05:01,880 Speaker 1: the world that they're going to use band Nvidia chips 74 00:05:01,880 --> 00:05:03,120 Speaker 1: in this large volume. 75 00:05:03,520 --> 00:05:06,400 Speaker 2: In the fourth quarter of last year, the Chinese government 76 00:05:06,440 --> 00:05:09,799 Speaker 2: approved a total of thirty nine data center investment projects 77 00:05:09,839 --> 00:05:13,480 Speaker 2: in Shinjan and the neighboring Chinghai province. They've built out 78 00:05:13,560 --> 00:05:16,000 Speaker 2: plans to use more than one hundred and fifteen thousand 79 00:05:16,200 --> 00:05:17,320 Speaker 2: Nvidia processors. 80 00:05:17,760 --> 00:05:20,599 Speaker 1: If wanted to compare this scale to what the US 81 00:05:20,640 --> 00:05:24,320 Speaker 1: have usually the tech companies there will claim if you 82 00:05:24,440 --> 00:05:28,360 Speaker 1: have more than one undred thousand Nvidia chips in one place, 83 00:05:28,839 --> 00:05:32,719 Speaker 1: then you'ing in a place to develop a state of 84 00:05:32,800 --> 00:05:38,400 Speaker 1: the art models which can possibly compete with Gemini Chat JBT. 85 00:05:38,640 --> 00:05:42,080 Speaker 2: These kind of models, there's no explanation in the official 86 00:05:42,200 --> 00:05:45,400 Speaker 2: documents of how companies plan to acquire the band in 87 00:05:45,480 --> 00:05:49,680 Speaker 2: Nvidia chips, and it could all be aspirational. In a 88 00:05:49,720 --> 00:05:52,880 Speaker 2: response from Nvidia, the company said that posting a web 89 00:05:52,920 --> 00:05:57,560 Speaker 2: page about restricted products is not the same as successfully licensing, 90 00:05:57,760 --> 00:06:01,160 Speaker 2: building and operating a data center, and that Nvidia does 91 00:06:01,200 --> 00:06:05,440 Speaker 2: not provide any support or repairs for restricted products. In 92 00:06:05,480 --> 00:06:08,880 Speaker 2: the meantime, the companies listed in the filings, state officials 93 00:06:08,920 --> 00:06:13,040 Speaker 2: and central government representatives in Beijing all declined to comment 94 00:06:13,080 --> 00:06:18,080 Speaker 2: when asked to explain. Still the building continues in the desert. 95 00:06:19,480 --> 00:06:22,120 Speaker 2: Most of these data centers are located in a single 96 00:06:22,200 --> 00:06:25,560 Speaker 2: compound in Ewu County, set up by the local government 97 00:06:25,680 --> 00:06:30,200 Speaker 2: in Shinjug. These data centers house big computers, servers, and 98 00:06:30,279 --> 00:06:34,440 Speaker 2: other components and are used to process, store, and distribute data. 99 00:06:35,000 --> 00:06:37,640 Speaker 2: And while they might be located far from big cities 100 00:06:37,680 --> 00:06:41,360 Speaker 2: like Shanghai or Shenzen, startups in these cities can send 101 00:06:41,400 --> 00:06:44,560 Speaker 2: their models to the data centers for training and optimization. 102 00:06:45,440 --> 00:06:50,400 Speaker 1: Theoretically, any company in China can access that computing power. Essentially, 103 00:06:50,400 --> 00:06:53,039 Speaker 1: you can send a request to the data centers in 104 00:06:53,080 --> 00:06:56,360 Speaker 1: Shinjioan saying that, hey, we want to use how many 105 00:06:56,520 --> 00:07:00,839 Speaker 1: chips to train our models in the eastern regions, and 106 00:07:00,880 --> 00:07:03,640 Speaker 1: then the data centers and Shinjong will produce a result 107 00:07:03,680 --> 00:07:07,359 Speaker 1: to you, maybe in a few minutes or even a 108 00:07:07,400 --> 00:07:11,560 Speaker 1: few seconds. Yeah, so these are all connected because you 109 00:07:11,680 --> 00:07:14,520 Speaker 1: don't need to be present at the data center use 110 00:07:14,720 --> 00:07:15,640 Speaker 1: the computing power. 111 00:07:16,280 --> 00:07:20,040 Speaker 2: Shinjong is an ideal location for large scale data centers. 112 00:07:20,600 --> 00:07:24,000 Speaker 2: Land is cheap and rental costs are low. The region 113 00:07:24,160 --> 00:07:27,680 Speaker 2: also has cool weather, which helps offset the heat generated 114 00:07:27,720 --> 00:07:31,760 Speaker 2: by the servers. Most importantly, it's a hub for green energy. 115 00:07:32,600 --> 00:07:34,720 Speaker 2: On the way to look for the data centers, James 116 00:07:34,720 --> 00:07:40,480 Speaker 2: Mager stumbled across miles and miles of renewable energy installations. 117 00:07:40,320 --> 00:07:45,040 Speaker 3: The desert was just a dusty, rocky plane, and scattered 118 00:07:45,040 --> 00:07:48,679 Speaker 3: across the desert was just windmill after windmill, after solar 119 00:07:48,720 --> 00:07:51,960 Speaker 3: panel installation after solar panel installation, and then there was 120 00:07:52,000 --> 00:07:56,800 Speaker 3: also this power generation, which is where the company uses 121 00:07:56,920 --> 00:08:00,360 Speaker 3: mirrors to concentrate sunlight on tooward tower, which is filled 122 00:08:00,400 --> 00:08:03,880 Speaker 3: with molten sodium, and then that heat is generated by 123 00:08:04,240 --> 00:08:07,760 Speaker 3: concentrating all that sunlight is then used to generate electricity. 124 00:08:07,960 --> 00:08:10,200 Speaker 3: And you can see this like twenty thirty kilometers away 125 00:08:10,240 --> 00:08:13,960 Speaker 3: because the sunlight is so blinding as it's concentrated onto 126 00:08:13,960 --> 00:08:14,520 Speaker 3: this tower. 127 00:08:15,760 --> 00:08:19,680 Speaker 2: This combination of green energy and advanced computing fits in 128 00:08:19,720 --> 00:08:24,840 Speaker 2: with Beijing's economic push for sustainable development. After the break, 129 00:08:25,000 --> 00:08:28,560 Speaker 2: what is China's master plan to build these data centers? 130 00:08:29,040 --> 00:08:44,080 Speaker 2: We zoom in on one company to find out. Bloomberg's 131 00:08:44,120 --> 00:08:47,560 Speaker 2: James Mager spent three days in Shinjung tracking down the 132 00:08:47,640 --> 00:08:52,800 Speaker 2: data centers. Andy Linn and his team had found on paper, Iwu. 133 00:08:52,720 --> 00:08:56,080 Speaker 3: Is pretty remote, even in Shinjong, which is pretty remote 134 00:08:56,080 --> 00:08:58,959 Speaker 3: in China. To get there, it was a four hour 135 00:08:59,320 --> 00:09:03,040 Speaker 3: plane ride from Beijing to the city of Hami, and 136 00:09:03,080 --> 00:09:05,280 Speaker 3: then from Hami to Iwu is another three and a 137 00:09:05,360 --> 00:09:08,600 Speaker 3: half for our car ride up into the mountains. 138 00:09:09,320 --> 00:09:12,480 Speaker 2: Using the data team's coordinates, James was able to locate 139 00:09:12,640 --> 00:09:16,360 Speaker 2: massive building complexes in the desert right where Andy said 140 00:09:16,400 --> 00:09:18,400 Speaker 2: the data centers were being built. 141 00:09:19,920 --> 00:09:22,319 Speaker 3: There was a lot of construction in Eyu, and there 142 00:09:22,360 --> 00:09:24,760 Speaker 3: was some construction going on up in the area with 143 00:09:24,800 --> 00:09:28,440 Speaker 3: all the solar panels. I've never seen a data center before. 144 00:09:28,920 --> 00:09:30,200 Speaker 3: I mean, this is the first time I've seen a 145 00:09:30,240 --> 00:09:33,280 Speaker 3: data center. But they look like what I imagine a data center. 146 00:09:33,040 --> 00:09:33,600 Speaker 1: Would look like. 147 00:09:34,160 --> 00:09:36,640 Speaker 3: You have a large building which is three, four or 148 00:09:36,640 --> 00:09:40,199 Speaker 3: five stories tall, has almost no windows. They obviously don't 149 00:09:40,240 --> 00:09:42,120 Speaker 3: need a lot of light. Having a lot of windows 150 00:09:42,160 --> 00:09:44,400 Speaker 3: is going to make it harder to keep the temperature 151 00:09:44,400 --> 00:09:47,880 Speaker 3: inside the data center controlled at the proper level. 152 00:09:48,120 --> 00:09:50,640 Speaker 2: It sounds like they had certainly all the hallmarks of 153 00:09:50,720 --> 00:09:51,600 Speaker 2: being a data center. 154 00:09:51,920 --> 00:09:52,199 Speaker 1: Yes. 155 00:09:53,320 --> 00:09:57,520 Speaker 2: Among all the projects examined in the investigation, one caught 156 00:09:57,520 --> 00:10:01,359 Speaker 2: the team's attention. It involves a company controlled by Nyocore, 157 00:10:01,679 --> 00:10:04,720 Speaker 2: an energy firm that's partially owned by the state government 158 00:10:05,040 --> 00:10:08,960 Speaker 2: based in Tianjin in northern China. Niacore's main business is 159 00:10:08,960 --> 00:10:12,640 Speaker 2: supplying solar and wind power. And I asked Andy, why 160 00:10:12,679 --> 00:10:16,000 Speaker 2: a green energy firm, of all companies, would be building 161 00:10:16,120 --> 00:10:16,840 Speaker 2: data centers. 162 00:10:17,600 --> 00:10:20,920 Speaker 1: Yeah. So you don't usually think that green energy companies 163 00:10:21,040 --> 00:10:25,360 Speaker 1: delve into the data standard business, but in China, Beijing 164 00:10:25,400 --> 00:10:29,480 Speaker 1: has been encouraging green energy companies to invest into AI 165 00:10:29,640 --> 00:10:34,040 Speaker 1: data centers because they want to lower the carbon foot prints. 166 00:10:34,120 --> 00:10:36,400 Speaker 1: That will create a win win. 167 00:10:37,160 --> 00:10:41,200 Speaker 2: According to the documents, Bloomberg found Nyacore plans to build 168 00:10:41,240 --> 00:10:43,960 Speaker 2: a data center with six hundred and twenty five servers. 169 00:10:44,360 --> 00:10:46,840 Speaker 1: It is one of the largest projects we found in 170 00:10:46,880 --> 00:10:51,040 Speaker 1: the batch of the documents. Based on their investment approvals, 171 00:10:51,400 --> 00:10:56,360 Speaker 1: they need two thousand, one hundred GPU chips. This a 172 00:10:56,440 --> 00:11:00,960 Speaker 1: lot because if you imagine one at one hundred chips 173 00:11:01,000 --> 00:11:05,840 Speaker 1: cost around twenty thousand dollars, then well is a very 174 00:11:05,880 --> 00:11:11,440 Speaker 1: big amount of money just for mid sized green energy 175 00:11:11,440 --> 00:11:12,240 Speaker 1: company in China. 176 00:11:13,320 --> 00:11:17,319 Speaker 2: Niacord is selling its computer power to infinitgens Ai, one 177 00:11:17,360 --> 00:11:21,520 Speaker 2: of the largest AI infrastructure companies in China. The company 178 00:11:21,559 --> 00:11:24,280 Speaker 2: has raised one billion yun or almost one hundred and 179 00:11:24,360 --> 00:11:27,280 Speaker 2: forty million dollars in funding since it was founded in 180 00:11:27,280 --> 00:11:31,839 Speaker 2: twenty twenty three. Niacord declined to comment and infinogens Ai 181 00:11:32,080 --> 00:11:35,360 Speaker 2: couldn't be reached for a response, but last year, the 182 00:11:35,360 --> 00:11:38,960 Speaker 2: CEO of infinogens Ai said in an interview with local 183 00:11:39,000 --> 00:11:42,200 Speaker 2: media that the goal is to make computing power more 184 00:11:42,280 --> 00:11:45,400 Speaker 2: accessible to AI companies all across China. 185 00:11:46,080 --> 00:11:50,960 Speaker 1: Their CEO said they aim to create a computing power 186 00:11:51,080 --> 00:11:54,960 Speaker 1: system that allows AI developers across the main line to 187 00:11:55,400 --> 00:11:59,080 Speaker 1: just log into their system and get all the computing 188 00:11:59,160 --> 00:12:02,520 Speaker 1: power they want, just like opening a like water tap 189 00:12:02,600 --> 00:12:05,959 Speaker 1: and then you can get the water you want. 190 00:12:06,120 --> 00:12:09,960 Speaker 2: Computer power on tap. But in order to make computer 191 00:12:10,080 --> 00:12:14,400 Speaker 2: power accessible anywhere, anytime, there's an issue that needs to 192 00:12:14,440 --> 00:12:19,440 Speaker 2: be addressed first. Shinjong is China's most western region. Most 193 00:12:19,480 --> 00:12:22,280 Speaker 2: of the AI companies that need the computing power are 194 00:12:22,280 --> 00:12:25,640 Speaker 2: based in big cities like Shanghai, far in the east. 195 00:12:26,280 --> 00:12:29,240 Speaker 2: That's more than two thousand miles apart, about the distance 196 00:12:29,240 --> 00:12:33,040 Speaker 2: from Chicago to Los Angeles. Andy says. The Chinese government 197 00:12:33,160 --> 00:12:36,360 Speaker 2: plans to bridge this supplied demand gap by building what 198 00:12:36,400 --> 00:12:38,880 Speaker 2: they call computing power corridors. 199 00:12:39,440 --> 00:12:42,559 Speaker 1: You can think about this as the water resource management 200 00:12:42,720 --> 00:12:46,400 Speaker 1: system in China. They have been moving the water from 201 00:12:46,440 --> 00:12:49,240 Speaker 1: the southern region to the north region, which is suffering 202 00:12:49,280 --> 00:12:52,800 Speaker 1: from draft. So they have been moving crucial resources around 203 00:12:52,840 --> 00:12:56,000 Speaker 1: in the mainland, and now they want to move the 204 00:12:56,120 --> 00:12:56,960 Speaker 1: computing power. 205 00:12:57,720 --> 00:13:01,120 Speaker 2: So in some ways they're thinking of a computing power 206 00:13:01,280 --> 00:13:04,320 Speaker 2: like a utility, right, and the computing power corridor would 207 00:13:04,360 --> 00:13:05,560 Speaker 2: be like the power grid. 208 00:13:06,080 --> 00:13:09,040 Speaker 1: Yeah, yeah, that's what they are aiming for. They're building 209 00:13:09,040 --> 00:13:13,520 Speaker 1: infrastructure like more cables along the codors, so that they 210 00:13:13,600 --> 00:13:18,160 Speaker 1: can encourage more AI developers to build infrastructure in the 211 00:13:18,200 --> 00:13:19,080 Speaker 1: western regions. 212 00:13:19,800 --> 00:13:23,320 Speaker 2: This is all part of China's plan to dominate in AI. 213 00:13:23,840 --> 00:13:26,199 Speaker 2: But in order for China to get there, it needs 214 00:13:26,280 --> 00:13:29,160 Speaker 2: all these things to come together. Its own high tech 215 00:13:29,240 --> 00:13:32,840 Speaker 2: chips that can rival in NVIDIAs, computer power on tap, 216 00:13:33,240 --> 00:13:37,120 Speaker 2: and an industry with cutting edge innovations. And Andy says 217 00:13:37,120 --> 00:13:41,280 Speaker 2: building such a national network could encourage more Chinese companies 218 00:13:41,360 --> 00:13:44,040 Speaker 2: to jump into the AI space without needing to build 219 00:13:44,040 --> 00:13:47,440 Speaker 2: their own data centers. That would help startups come up 220 00:13:47,440 --> 00:13:51,240 Speaker 2: with more innovations like Deepseek did with its chatbot, and 221 00:13:51,400 --> 00:13:56,239 Speaker 2: ultimately narrow the gap between the US and China. 222 00:13:56,280 --> 00:13:59,200 Speaker 1: I think one of the most powerful aspect of this 223 00:13:59,480 --> 00:14:03,800 Speaker 1: utility for AI computing powers that it will not be 224 00:14:03,920 --> 00:14:08,320 Speaker 1: controlled by a couple of tech giants only. So if 225 00:14:08,320 --> 00:14:11,680 Speaker 1: you make a comparison between US and China. In the US, 226 00:14:11,840 --> 00:14:15,959 Speaker 1: computing power is mostly owned by a handful of tech giants. 227 00:14:16,200 --> 00:14:20,960 Speaker 1: Open ai madea alphabet, but if you make it as 228 00:14:20,960 --> 00:14:26,200 Speaker 1: a utility, every developer across the mainline can just log 229 00:14:26,280 --> 00:14:29,920 Speaker 1: into a system and then got as much computing power 230 00:14:30,000 --> 00:14:33,520 Speaker 1: as they want. Then it will create a very distributed 231 00:14:33,600 --> 00:14:38,640 Speaker 1: system for AI innovations, so that anyone who have the 232 00:14:38,720 --> 00:14:42,960 Speaker 1: idea can just create its own model, and then we'll 233 00:14:43,000 --> 00:14:47,320 Speaker 1: have a more collaborative environment for these developments, which is 234 00:14:47,400 --> 00:14:49,600 Speaker 1: totally different from what we have in the US. 235 00:14:50,360 --> 00:14:53,640 Speaker 2: There's a lot of uncertainty about this model. Will China 236 00:14:53,720 --> 00:14:56,440 Speaker 2: be able to create a cheap and reliable power source 237 00:14:56,720 --> 00:14:59,560 Speaker 2: that everyone can tap into, will it be able to 238 00:14:59,560 --> 00:15:03,760 Speaker 2: acquire computer chips that AI development demands, and will China 239 00:15:03,840 --> 00:15:06,680 Speaker 2: be able to make those chips itself rather than relying 240 00:15:06,720 --> 00:15:08,480 Speaker 2: on a US company Like in video. 241 00:15:09,200 --> 00:15:13,360 Speaker 1: What we found in Shinjiai and Shanghai for these scale 242 00:15:13,440 --> 00:15:17,280 Speaker 1: of data CENTTA development shows that Beijing is trying very 243 00:15:17,320 --> 00:15:21,880 Speaker 1: hard to nurture is on domestic AI industry so that 244 00:15:21,920 --> 00:15:33,320 Speaker 1: they can match the tech giants from the wests. 245 00:15:39,120 --> 00:15:42,840 Speaker 2: This is The Big Take Asia from Bloomberg News. I'm wanha. 246 00:15:42,920 --> 00:15:45,680 Speaker 2: To get more from The Big Take and unlimited access 247 00:15:45,720 --> 00:15:48,680 Speaker 2: to all of Bloomberg dot Com, subscribe today at Bloomberg 248 00:15:48,720 --> 00:15:52,200 Speaker 2: dot com slash podcast offer. If you like the episode, 249 00:15:52,280 --> 00:15:54,720 Speaker 2: make sure to subscribe and review The Big Take Asia 250 00:15:54,880 --> 00:15:58,120 Speaker 2: wherever you listen to podcasts. It really helps people find 251 00:15:58,160 --> 00:16:00,480 Speaker 2: the show. Thanks for listening, See next time.