WEBVTT - Why African Startups Are Embracing DeepSeek

0:00:00.280 --> 0:00:07.600
<v Speaker 1>Bloomberg Audio Studios, Podcasts, radio news.

0:00:08.240 --> 0:00:10.960
<v Speaker 2>I'm Sarah Holder, host of The Big Take, and I'm

0:00:11.039 --> 0:00:15.680
<v Speaker 2>Jennifer Zabasaja, host of Bloomberg's Next Africa podcast. Today, we're

0:00:15.680 --> 0:00:18.880
<v Speaker 2>bringing you a story reported by our colleagues and newsrooms

0:00:18.880 --> 0:00:20.560
<v Speaker 2>throughout Africa and Asia.

0:00:20.960 --> 0:00:23.800
<v Speaker 3>It's about how the AI arms race is playing out

0:00:23.880 --> 0:00:27.640
<v Speaker 3>in Africa as Chinese AI developers try to break into

0:00:27.680 --> 0:00:33.000
<v Speaker 3>the burgeoning Africa tech scene. Earlier this year, a group

0:00:33.040 --> 0:00:36.760
<v Speaker 3>of tech executives from all over Africa got together at

0:00:36.760 --> 0:00:40.760
<v Speaker 3>the headquarters of a Nairobi based startup to consider a

0:00:40.800 --> 0:00:41.880
<v Speaker 3>business proposition.

0:00:42.280 --> 0:00:46.640
<v Speaker 1>So it was an early breakfast meeting and it had

0:00:47.000 --> 0:00:50.760
<v Speaker 1>a whole lot of African startup founders. And then there

0:00:50.800 --> 0:00:53.800
<v Speaker 1>was also a group from Huawei with one of the

0:00:53.880 --> 0:00:58.800
<v Speaker 1>top exigs that sort of runs sub foreign Africa's cloud services.

0:00:59.120 --> 0:01:03.080
<v Speaker 3>Loney princelu is a senior Africa reporter for Bloomberg based

0:01:03.120 --> 0:01:07.039
<v Speaker 3>in Johannesburg, and she says that Huawei Cloud Exac was

0:01:07.040 --> 0:01:11.600
<v Speaker 3>there to talk about one buzzy new company. In particular, he.

0:01:11.520 --> 0:01:14.920
<v Speaker 1>Came in with quite a strong pitch for deep Sick, saying,

0:01:15.240 --> 0:01:17.759
<v Speaker 1>you know, deeptic is so hot. That was literally his words,

0:01:17.760 --> 0:01:20.160
<v Speaker 1>deeptic is so hot. Now I want to talk about

0:01:20.240 --> 0:01:20.960
<v Speaker 1>anything else.

0:01:21.560 --> 0:01:25.480
<v Speaker 2>Deep Sik the Chinese AI platform that launched a breakthrough

0:01:25.600 --> 0:01:30.680
<v Speaker 2>reasoning model in January twenty twenty five, shocking its competition.

0:01:31.160 --> 0:01:34.000
<v Speaker 1>Deep Seek. Where did this one come from?

0:01:34.040 --> 0:01:37.160
<v Speaker 4>The cheaper ALI model from Chinese starts up deep Seek

0:01:37.240 --> 0:01:40.360
<v Speaker 4>could fraten us big tech. Deep Siki is a Chinese

0:01:40.400 --> 0:01:44.440
<v Speaker 4>startup that's producing these large language models that offer very

0:01:44.480 --> 0:01:47.119
<v Speaker 4>high performance at a relatively low cost.

0:01:47.480 --> 0:01:51.040
<v Speaker 2>That's Helen Yambura, Bloomberg's East Africa bureau chief.

0:01:51.360 --> 0:01:55.240
<v Speaker 4>Deep Siki is using Wawei hardware in the ecosystem, both

0:01:55.320 --> 0:01:59.880
<v Speaker 4>cheaps and cloud, so from a political perspective, Huawei is

0:02:00.040 --> 0:02:04.840
<v Speaker 4>promoting domestic AI tools that are running on domestic hardware.

0:02:05.120 --> 0:02:08.480
<v Speaker 2>Huawei and deep Seek's parent company, high Flyer have been

0:02:08.520 --> 0:02:12.440
<v Speaker 2>collaborating since mid twenty twenty three. In addition to hosting

0:02:12.480 --> 0:02:15.640
<v Speaker 2>deep Seek on its cloud, Huawei has also started selling

0:02:15.680 --> 0:02:19.280
<v Speaker 2>its storage and computing services in a bundle with access

0:02:19.320 --> 0:02:23.200
<v Speaker 2>to deep Seek's large language model. So Huawei's exec wasn't

0:02:23.280 --> 0:02:26.240
<v Speaker 2>just hyping up how hot deep seek is. He was

0:02:26.280 --> 0:02:30.040
<v Speaker 2>trying to pitch these African startups on using deep Seek's

0:02:30.040 --> 0:02:34.840
<v Speaker 2>AI platform and accessing it through Huawei's cloud servers, and

0:02:34.880 --> 0:02:37.040
<v Speaker 2>he presented a pretty compelling offer.

0:02:38.520 --> 0:02:42.480
<v Speaker 1>Basically, if you are going onto Huawei Cloud instead of

0:02:42.520 --> 0:02:45.639
<v Speaker 1>going directly onto deep Seek every day, you get two

0:02:45.760 --> 0:02:49.600
<v Speaker 1>million tokens free. So that's sort of unheard up and

0:02:50.120 --> 0:02:51.480
<v Speaker 1>then a crack pleaser.

0:02:52.000 --> 0:02:54.919
<v Speaker 2>Tokens are the unit that deep Seek uses to bill

0:02:55.040 --> 0:02:58.760
<v Speaker 2>for its services. Two million free tokens can get you

0:02:58.840 --> 0:03:01.840
<v Speaker 2>a lot of deep Seak computing power. According to the

0:03:01.919 --> 0:03:05.680
<v Speaker 2>Huawei presentation, they would cover about three thousand pages of

0:03:05.760 --> 0:03:08.240
<v Speaker 2>new content generated by deep Seek's model.

0:03:08.600 --> 0:03:12.560
<v Speaker 3>Huawei is offering incentives like this to African tech companies

0:03:12.639 --> 0:03:15.840
<v Speaker 3>because they stand to gain a lot too. They're trying

0:03:15.880 --> 0:03:19.520
<v Speaker 3>to establish a crucial foothold on the continent early in

0:03:19.560 --> 0:03:21.160
<v Speaker 3>the AI race, and.

0:03:21.160 --> 0:03:23.960
<v Speaker 2>Some critics worry that deep Seek and Huawei's bet on

0:03:24.080 --> 0:03:29.920
<v Speaker 2>African AI infrastructure could have complex long term consequences, including

0:03:30.160 --> 0:03:38.160
<v Speaker 2>for data privacy. This is the Big Take from Bloomberg News.

0:03:38.360 --> 0:03:39.760
<v Speaker 2>I'm Sarah Holder.

0:03:39.880 --> 0:03:42.600
<v Speaker 3>And this is a special co production of The Big

0:03:42.640 --> 0:03:46.640
<v Speaker 3>Take and Bloomberg's Next Africa Podcast. I'm Jennifer's Abasaga.

0:03:46.920 --> 0:03:50.000
<v Speaker 2>Today on the show, Deep Seek and Huawei's push to

0:03:50.120 --> 0:03:53.560
<v Speaker 2>dominate the artificial intelligence market in Africa.

0:03:53.640 --> 0:03:56.360
<v Speaker 3>What it could mean for the global AI race, the

0:03:56.360 --> 0:04:00.320
<v Speaker 3>soft power balanced between the US and China, and as

0:04:00.360 --> 0:04:13.200
<v Speaker 3>efforts to influence Africa's development, industry and innovation. Africa's digital

0:04:13.240 --> 0:04:16.560
<v Speaker 3>economy is valued at about one hundred and eighty billion

0:04:16.600 --> 0:04:20.600
<v Speaker 3>dollars today, but by twenty fifty it's expected to swell

0:04:20.640 --> 0:04:23.400
<v Speaker 3>to seven hundred and twelve billion dollars.

0:04:25.920 --> 0:04:29.200
<v Speaker 2>For the past few years, while American AI companies have

0:04:29.320 --> 0:04:34.520
<v Speaker 2>experienced record levels of growth, African companies and African users

0:04:34.920 --> 0:04:38.680
<v Speaker 2>have been largely left behind. One reason for this is

0:04:38.720 --> 0:04:42.000
<v Speaker 2>that access to computing resources plays a big role in

0:04:42.080 --> 0:04:45.640
<v Speaker 2>African startups being able to develop their own AI models

0:04:45.760 --> 0:04:49.520
<v Speaker 2>and attract funding. The bulk of the eight hundred million

0:04:49.600 --> 0:04:53.599
<v Speaker 2>dollars African AI startups attracted last year went to places

0:04:53.640 --> 0:04:58.240
<v Speaker 2>on the continent that have more AI capabilities, like South Africa, Kenya,

0:04:58.360 --> 0:04:59.640
<v Speaker 2>Tunisia and Egypt.

0:05:00.200 --> 0:05:03.560
<v Speaker 3>And Huawei's pitch for deep Seek in that Nairobi conference

0:05:03.640 --> 0:05:07.120
<v Speaker 3>room catered to the broad spectrum of AI use cases

0:05:07.279 --> 0:05:11.799
<v Speaker 3>that African entities could need, from free options for smaller

0:05:11.839 --> 0:05:16.440
<v Speaker 3>startups to private cloud systems for governments to more compute

0:05:16.480 --> 0:05:18.839
<v Speaker 3>intensive options for app developers.

0:05:19.279 --> 0:05:22.640
<v Speaker 2>Bloomberg's Lonely Prince Lou says that in Africa, deepseek is

0:05:22.680 --> 0:05:25.800
<v Speaker 2>trying to position itself as a more flexible option than

0:05:25.800 --> 0:05:27.120
<v Speaker 2>its Western competitors.

0:05:27.920 --> 0:05:31.400
<v Speaker 1>When it comes to Google, Gemini and Rita Lama, those

0:05:31.440 --> 0:05:34.920
<v Speaker 1>waste and models, you have quite a lot of restrictions,

0:05:35.360 --> 0:05:37.680
<v Speaker 1>licensing restrictions and take restrictions.

0:05:38.000 --> 0:05:42.120
<v Speaker 3>AI offerings from US companies tend to be harder to customize.

0:05:42.240 --> 0:05:47.000
<v Speaker 3>They're largely not open source, but deep seek is that

0:05:47.200 --> 0:05:50.880
<v Speaker 3>means users can access the underlying code and modify the

0:05:50.920 --> 0:05:54.480
<v Speaker 3>software to suit their needs for no additional charge.

0:05:54.760 --> 0:05:58.919
<v Speaker 1>Something like that just suits the African environment much better.

0:06:00.000 --> 0:06:03.400
<v Speaker 1>Always wants the large language models, sometimes you want a

0:06:03.400 --> 0:06:06.640
<v Speaker 1>small language model for the types of adaptations that you

0:06:06.680 --> 0:06:07.039
<v Speaker 1>want to do.

0:06:07.120 --> 0:06:12.159
<v Speaker 2>In terms of AI, one reason African startups might want

0:06:12.160 --> 0:06:17.120
<v Speaker 2>a more customizable model is because many USAI models weren't

0:06:17.160 --> 0:06:20.599
<v Speaker 2>trained on African dialects and they often fail to pick

0:06:20.680 --> 0:06:25.080
<v Speaker 2>up on cultural nuances. Using them on unfamiliar words and

0:06:25.120 --> 0:06:29.320
<v Speaker 2>situations can be expensive. They require more tokens to perform

0:06:29.400 --> 0:06:33.560
<v Speaker 2>those tasks. This all makes deep Seek's open source model

0:06:33.680 --> 0:06:37.640
<v Speaker 2>more appealing, and using the model is also much cheaper.

0:06:37.920 --> 0:06:42.160
<v Speaker 1>We spoke to one sort of in Nigeria called Equalize AI,

0:06:42.279 --> 0:06:44.760
<v Speaker 1>and so basically what deep seek is charging is twenty

0:06:45.160 --> 0:06:48.880
<v Speaker 1>seven usints to press a million words for a query,

0:06:49.279 --> 0:06:52.039
<v Speaker 1>and then a dollar a team for a million words.

0:06:52.120 --> 0:06:54.719
<v Speaker 1>In terms of the response that it generates.

0:06:54.720 --> 0:06:58.320
<v Speaker 2>That's a lot less than Chat GPT, which costs more

0:06:58.400 --> 0:07:01.400
<v Speaker 2>like five dollars for a million words of processing and

0:07:01.520 --> 0:07:04.080
<v Speaker 2>fifteen dollars for a million words generated.

0:07:04.600 --> 0:07:08.400
<v Speaker 1>That is two thousand and seven hundred dollars per month

0:07:08.520 --> 0:07:11.440
<v Speaker 1>if it us deep seek, vers is twelve thousand, five

0:07:11.520 --> 0:07:15.480
<v Speaker 1>hundred dollars if it uses chat GPT four for the

0:07:15.520 --> 0:07:18.160
<v Speaker 1>same toxic So it's a massive difference.

0:07:18.520 --> 0:07:21.920
<v Speaker 2>Deep Sex's platform is four and a half times cheaper

0:07:22.040 --> 0:07:26.520
<v Speaker 2>for Equalized AI to use, which raises the obvious question.

0:07:27.480 --> 0:07:31.480
<v Speaker 3>What's the financial appeal for deep sek. What are they

0:07:31.520 --> 0:07:35.360
<v Speaker 3>seeing financially when it comes to some of these rois.

0:07:35.680 --> 0:07:38.760
<v Speaker 1>Look, they're still charging, so they must be playing more

0:07:38.800 --> 0:07:42.640
<v Speaker 1>of a volumes game. And in terms of volumes, that

0:07:42.760 --> 0:07:46.000
<v Speaker 1>is what Africa can offer. We have a massive amount

0:07:46.000 --> 0:07:49.040
<v Speaker 1>of young people on the continent. They're all interested in

0:07:49.080 --> 0:07:52.320
<v Speaker 1>what technology can do for them. There's a lot of

0:07:52.360 --> 0:07:57.680
<v Speaker 1>restrictions in terms of actual traditional infrastructure. So technology is

0:07:57.680 --> 0:08:01.240
<v Speaker 1>where everyone's sort of looking towards. So it's a longer game,

0:08:02.000 --> 0:08:04.840
<v Speaker 1>I think for deep Sick more than it is for

0:08:04.960 --> 0:08:08.200
<v Speaker 1>open Ai, who is running for profits.

0:08:08.320 --> 0:08:12.600
<v Speaker 2>Basically, open Ai didn't respond to a request for comment

0:08:12.680 --> 0:08:17.200
<v Speaker 2>on the story. Meanwhile, Deep Sik's numbers game seems to

0:08:17.240 --> 0:08:18.080
<v Speaker 2>be working.

0:08:18.280 --> 0:08:23.520
<v Speaker 1>When it comes to startups. Specifically, every African stortup we

0:08:23.560 --> 0:08:27.160
<v Speaker 1>spoke to prefer Deep Sick at this point when it

0:08:27.240 --> 0:08:32.040
<v Speaker 1>comes to training their models, adapting their models. So what

0:08:32.040 --> 0:08:34.800
<v Speaker 1>we're seeing on the ground is that they are definitely

0:08:34.960 --> 0:08:37.680
<v Speaker 1>numbers in favor for deep sic currently.

0:08:39.320 --> 0:08:42.760
<v Speaker 3>And Huawei isn't just pitching startups on deep Sek, they're

0:08:42.760 --> 0:08:46.120
<v Speaker 3>also trying to get African governments to adopt their platform.

0:08:46.520 --> 0:08:48.079
<v Speaker 3>Here's Bloomberg Talent Yamburra.

0:08:48.640 --> 0:08:52.480
<v Speaker 4>We don't have credible evidence that African governments are formally

0:08:52.600 --> 0:08:57.840
<v Speaker 4>adopting deep sick tools for their public service and official news,

0:08:58.280 --> 0:09:02.240
<v Speaker 4>but Huawei is pushing what it calls its National Government

0:09:02.360 --> 0:09:06.400
<v Speaker 4>Cloud Solution in Southern Africa. So this is platforms to

0:09:06.440 --> 0:09:13.080
<v Speaker 4>support public services and cloud infrastructure and digitization of government services.

0:09:13.559 --> 0:09:16.280
<v Speaker 3>If deep seek were to expand into the government space.

0:09:16.559 --> 0:09:20.040
<v Speaker 3>It could be handling more sensitive data, and critics at

0:09:20.120 --> 0:09:23.719
<v Speaker 3>universities and legal firms have raised concerns about how that

0:09:23.800 --> 0:09:26.680
<v Speaker 3>data is being gathered and how it could be used.

0:09:27.200 --> 0:09:30.760
<v Speaker 3>A Wall Street Journal investigation found that Huawei technicians have

0:09:30.840 --> 0:09:35.840
<v Speaker 3>provided African governments with information on their political opponents. For example,

0:09:36.760 --> 0:09:40.400
<v Speaker 3>Huawei and the governments, though deny the claims, Deep.

0:09:40.240 --> 0:09:43.600
<v Speaker 2>Seak has already been blocked in several European countries over

0:09:43.720 --> 0:09:47.160
<v Speaker 2>data privacy concerns and what regulators say is a lack

0:09:47.160 --> 0:09:51.240
<v Speaker 2>of transparency about the relationship between the Chinese government and

0:09:51.280 --> 0:09:53.199
<v Speaker 2>companies like Huawei and deep Seek.

0:09:54.280 --> 0:09:57.360
<v Speaker 3>One of the central concerns is that the Chinese government

0:09:57.480 --> 0:10:01.360
<v Speaker 3>could gain access to deep seek data store on Chinese servers.

0:10:01.840 --> 0:10:05.920
<v Speaker 3>Italy's Data Protection Authority asked for information on deep Seek's

0:10:05.960 --> 0:10:10.720
<v Speaker 3>data collection, use and storage policies, which deep Seek did

0:10:10.760 --> 0:10:15.080
<v Speaker 3>not produce. German regulators asked for evidence that German user

0:10:15.200 --> 0:10:19.040
<v Speaker 3>data sent to China was protected in line with European standards,

0:10:19.440 --> 0:10:21.880
<v Speaker 3>which Deep Seek was unable to provide.

0:10:22.200 --> 0:10:25.840
<v Speaker 2>When asked by Bloomberg about data privacy concerns, Deep Seek

0:10:25.880 --> 0:10:31.800
<v Speaker 2>and Huawei both declined to comment. Huawei's efforts in Africa

0:10:31.880 --> 0:10:34.720
<v Speaker 2>are aligned with a similar playbook that the Chinese government

0:10:34.760 --> 0:10:37.280
<v Speaker 2>has used for over a decade as a means of

0:10:37.320 --> 0:10:41.440
<v Speaker 2>gaining soft power after the break what that approach could

0:10:41.440 --> 0:10:45.360
<v Speaker 2>mean for China's influence and its position in the global

0:10:45.400 --> 0:10:46.120
<v Speaker 2>AI race.

0:10:56.440 --> 0:10:59.360
<v Speaker 3>For over a decade, the Chinese government has made a

0:10:59.440 --> 0:11:04.760
<v Speaker 3>concerted effort to invest in foreign infrastructure, increasing its influence

0:11:04.840 --> 0:11:09.280
<v Speaker 3>in developing countries. It's a strategy known as the Belt

0:11:09.320 --> 0:11:10.440
<v Speaker 3>and Road initiative.

0:11:11.320 --> 0:11:15.280
<v Speaker 4>China is going to the developing world, especially in Africa,

0:11:15.760 --> 0:11:20.800
<v Speaker 4>and you know, extending soft power through building infrastructure, but

0:11:21.040 --> 0:11:25.000
<v Speaker 4>using technology rather than you know, the hard airports and

0:11:25.080 --> 0:11:28.400
<v Speaker 4>ports and roads, Andre Luise, this now is you know,

0:11:28.480 --> 0:11:32.080
<v Speaker 4>the software to be ahead of the yules.

0:11:33.200 --> 0:11:36.800
<v Speaker 2>This new AI software initiative isn't technically a Belt and

0:11:36.920 --> 0:11:40.679
<v Speaker 2>Road project, but it takes a similar approach, and it's

0:11:40.720 --> 0:11:43.320
<v Speaker 2>one the Chinese government has also used in the past

0:11:43.360 --> 0:11:46.840
<v Speaker 2>to expand the reach of Chinese phone companies like Huawei

0:11:46.880 --> 0:11:48.880
<v Speaker 2>and Zte into Africa.

0:11:49.440 --> 0:11:53.080
<v Speaker 4>The Chinese companies Juawi and Zte were not the first

0:11:53.160 --> 0:11:57.400
<v Speaker 4>on the continent, but these companies came with a cheaper,

0:11:57.480 --> 0:12:01.080
<v Speaker 4>more cost effective solution, so it's the same thing that

0:12:01.120 --> 0:12:02.640
<v Speaker 4>we're seeing with Deep Seek.

0:12:02.800 --> 0:12:06.960
<v Speaker 3>Belt and road projects in Kenya, Ethiopia, and Djibouti gave

0:12:07.080 --> 0:12:12.240
<v Speaker 3>China economic leverage over these developing countries. Chinese companies and

0:12:12.320 --> 0:12:16.000
<v Speaker 3>the Chinese government offered cheap loans in order to roll

0:12:16.000 --> 0:12:22.040
<v Speaker 3>out infrastructure and equipment across the continent to cement their position. Ultimately,

0:12:22.520 --> 0:12:26.040
<v Speaker 3>those nations had to majorly restructure their debt in order

0:12:26.120 --> 0:12:27.439
<v Speaker 3>to repay their loans.

0:12:28.080 --> 0:12:30.800
<v Speaker 2>Other nations who weren't able to pay back their loans,

0:12:30.840 --> 0:12:33.800
<v Speaker 2>like Sri Lanka, have had to turn over control of

0:12:33.840 --> 0:12:37.480
<v Speaker 2>their infrastructure to China. When it comes to Huawei and

0:12:37.520 --> 0:12:41.360
<v Speaker 2>Deep Seek, the leverage these companies could have is access

0:12:41.440 --> 0:12:42.199
<v Speaker 2>to information.

0:12:42.960 --> 0:12:46.640
<v Speaker 4>It's all about capturing the vast amount of data that

0:12:46.760 --> 0:12:50.560
<v Speaker 4>will shape AI in the future. Some critics are worried

0:12:50.880 --> 0:12:54.360
<v Speaker 4>that this will elcho that built and road where we

0:12:54.520 --> 0:12:58.160
<v Speaker 4>saw China forcing some poor countries to handle over critical

0:12:58.160 --> 0:13:03.320
<v Speaker 4>infrastructure to pay for all those projects. So we don't

0:13:03.360 --> 0:13:06.400
<v Speaker 4>know how these might pan out for African companies and

0:13:06.480 --> 0:13:09.160
<v Speaker 4>governments in future in terms of cost.

0:13:10.520 --> 0:13:13.080
<v Speaker 2>And there are other side effects that come from relying

0:13:13.160 --> 0:13:17.960
<v Speaker 2>too heavily on one country's digital infrastructure. For example, for

0:13:18.040 --> 0:13:21.400
<v Speaker 2>the past two years to prevent cheating during China's national

0:13:21.520 --> 0:13:26.680
<v Speaker 2>university entrance exams. China has restricted certain technical capabilities of

0:13:26.720 --> 0:13:28.600
<v Speaker 2>domestic tech companies.

0:13:28.240 --> 0:13:32.600
<v Speaker 1>Ali Baba, Black Dance, Taint, Saint and Deepseek, whose service

0:13:32.600 --> 0:13:35.640
<v Speaker 1>are all located in China. They disabled things like the

0:13:35.679 --> 0:13:39.680
<v Speaker 1>image recognition and the instant Q and A features on

0:13:40.040 --> 0:13:43.400
<v Speaker 1>sort of consumer facing products to prevent cheating.

0:13:43.720 --> 0:13:46.719
<v Speaker 2>But the move didn't just impact Chinese students.

0:13:47.000 --> 0:13:50.680
<v Speaker 1>We also saw then in Lagos in Nairobi who all

0:13:50.720 --> 0:13:54.920
<v Speaker 1>experienced slowdowns in terms of service interruptions and you can't

0:13:54.960 --> 0:13:56.120
<v Speaker 1>do anything about that.

0:13:56.880 --> 0:14:01.200
<v Speaker 2>And then there are the privacy concerns. Most big Western

0:14:01.240 --> 0:14:04.840
<v Speaker 2>AI models are supposed to comply with European laws governing

0:14:04.880 --> 0:14:09.959
<v Speaker 2>transparency around data collection and processing, but deep Seek's Chapot

0:14:10.080 --> 0:14:14.480
<v Speaker 2>does not. Instead, it typically stores user data on servers

0:14:14.520 --> 0:14:18.480
<v Speaker 2>in China, which under Chinese law, could be accessed by

0:14:18.480 --> 0:14:22.440
<v Speaker 2>the Chinese government. And while several European countries have taken

0:14:22.480 --> 0:14:26.640
<v Speaker 2>steps to curb deep Seek adoption, African countries have not.

0:14:27.320 --> 0:14:30.320
<v Speaker 4>I think most countries are really you know, at the

0:14:30.360 --> 0:14:35.440
<v Speaker 4>starting stage of developing the regulation and so on, so

0:14:35.560 --> 0:14:39.600
<v Speaker 4>there's a lot of adoption without the legal background, without

0:14:39.640 --> 0:14:46.280
<v Speaker 4>the regulatory backbone to cover that, so the technology has

0:14:46.320 --> 0:14:53.360
<v Speaker 4>sort of front run the regulation, which might be a problem.

0:14:53.640 --> 0:14:57.960
<v Speaker 3>Then there's the looming geopolitical backdrop to this. Bloomberg's Helen

0:14:58.040 --> 0:15:01.680
<v Speaker 3>Yamburra told me that if Africa nations and tech companies

0:15:01.800 --> 0:15:05.480
<v Speaker 3>start relying too heavily on Chinese tech, they could get

0:15:05.520 --> 0:15:08.560
<v Speaker 3>caught in the crossfires of the US China trade war.

0:15:08.880 --> 0:15:12.360
<v Speaker 2>For instance, if the US government tries to disrupt China's

0:15:12.440 --> 0:15:17.440
<v Speaker 2>AI capabilities by limiting access to chips or ratcheting up terrace,

0:15:18.080 --> 0:15:21.440
<v Speaker 2>that could also impact Africa's AI capabilities.

0:15:22.000 --> 0:15:24.320
<v Speaker 4>The sticks are just really too high for a single

0:15:24.360 --> 0:15:27.560
<v Speaker 4>country to rely on just one country, whether it's the

0:15:27.720 --> 0:15:28.720
<v Speaker 4>US or China.

0:15:29.920 --> 0:15:33.400
<v Speaker 2>The US has tried to pursue its own tech infrastructure

0:15:33.440 --> 0:15:35.560
<v Speaker 2>projects in Africa in the past.

0:15:35.760 --> 0:15:39.400
<v Speaker 1>And so big deals being announced in thin specifically, there

0:15:39.480 --> 0:15:42.920
<v Speaker 1>was a big AI data center being announced of more

0:15:42.960 --> 0:15:44.320
<v Speaker 1>than a billion dollars.

0:15:44.480 --> 0:15:48.160
<v Speaker 2>But under the Trump administration, those efforts have stalled.

0:15:48.400 --> 0:15:51.560
<v Speaker 1>And that's opened the way again for China. China, as

0:15:51.560 --> 0:15:54.120
<v Speaker 1>we said, has been here for a long time. That've

0:15:54.160 --> 0:15:57.440
<v Speaker 1>been in wasting for twenty years, so that they do

0:15:57.600 --> 0:15:59.800
<v Speaker 1>own a lot of the equipment in our five G

0:16:00.720 --> 0:16:05.080
<v Speaker 1>in our fiber, in our data centers in the US

0:16:05.160 --> 0:16:08.240
<v Speaker 1>companies that are here. A lot of them sort of

0:16:08.280 --> 0:16:12.360
<v Speaker 1>have set up shop in South Africa and some in

0:16:12.520 --> 0:16:15.160
<v Speaker 1>Kenya and some in Nigeria, but they don't really advance

0:16:15.520 --> 0:16:17.960
<v Speaker 1>to other African countries. And as we know, there's fifty

0:16:17.960 --> 0:16:21.440
<v Speaker 1>four of them. So China has really played the game

0:16:21.480 --> 0:16:22.640
<v Speaker 1>also in smaller countries.

0:16:24.960 --> 0:16:28.240
<v Speaker 3>And in part because of that dynamic, the expansion of

0:16:28.320 --> 0:16:31.480
<v Speaker 3>Chinese AI tools in Africa hasn't been met with the

0:16:31.520 --> 0:16:35.240
<v Speaker 3>same suspicion as it has in European countries or the

0:16:35.360 --> 0:16:40.080
<v Speaker 3>United States, because Loni says, many people in Africa are

0:16:40.320 --> 0:16:42.600
<v Speaker 3>just as skeptical of US tech.

0:16:43.280 --> 0:16:48.040
<v Speaker 1>In Africa, people don't always mistrust the East versus the West.

0:16:48.360 --> 0:16:53.320
<v Speaker 1>Many also distrust the West in terms of privacy and

0:16:53.840 --> 0:16:57.840
<v Speaker 1>you know, take and all of that. So African general

0:16:57.880 --> 0:17:02.160
<v Speaker 1>has a better relationship with China than a europe for instance.

0:17:02.720 --> 0:17:06.600
<v Speaker 1>But yeah, I would just I would like to just

0:17:06.680 --> 0:17:10.120
<v Speaker 1>see whether this adoption of deep seek in the sort

0:17:10.160 --> 0:17:11.720
<v Speaker 1>of space continues.

0:17:12.359 --> 0:17:15.720
<v Speaker 3>For now, Alen says she's watching to see how Africa's

0:17:15.800 --> 0:17:17.520
<v Speaker 3>AI industry develops.

0:17:17.840 --> 0:17:20.520
<v Speaker 4>What will be interesting is to see whether Africa can

0:17:20.880 --> 0:17:24.600
<v Speaker 4>come up with its own, you know, homegrown AI industry.

0:17:24.880 --> 0:17:28.000
<v Speaker 4>There's very many hurdles to this. It doesn't have the

0:17:28.119 --> 0:17:30.600
<v Speaker 4>money that you know, China or the US will throw

0:17:30.720 --> 0:17:34.840
<v Speaker 4>at AI. It also doesn't have the skilled labor to

0:17:34.960 --> 0:17:38.600
<v Speaker 4>do this or the infrastructure. But it would be good

0:17:38.640 --> 0:17:41.679
<v Speaker 4>for Africa to be at least to try for some

0:17:41.880 --> 0:17:43.720
<v Speaker 4>independence on this front.

0:17:48.680 --> 0:17:50.679
<v Speaker 2>This is the Big Take from Bloomberg News.

0:17:50.920 --> 0:17:54.800
<v Speaker 3>I'm Sarah Holder and I'm Jennifer's Abisada with Next to Africa.

0:17:54.960 --> 0:17:58.399
<v Speaker 2>Follow The Next Africa podcast now and join them every

0:17:58.400 --> 0:18:03.320
<v Speaker 2>Friday on Apple, podcast, Spotify, iHeart, or wherever you listen.

0:18:03.720 --> 0:18:06.320
<v Speaker 3>To get more from The Big Take and unlimited access

0:18:06.400 --> 0:18:09.720
<v Speaker 3>to all of Bloomberg dot Com, subscribe today at Bloomberg

0:18:09.800 --> 0:18:12.240
<v Speaker 3>dot com slash podcast Offer.

0:18:12.560 --> 0:18:15.000
<v Speaker 2>If you liked this episode, make sure to follow and

0:18:15.080 --> 0:18:18.280
<v Speaker 2>review The Big Take or Next Africa wherever you listen

0:18:18.320 --> 0:18:21.040
<v Speaker 2>to podcasts. It helps people find the shows