00:00We're going to be profiling Minimax, which had that big IPO here in Hong Kong in January,
00:04and we are lucky to have the co-founder Ye Ye Yun. Thank you so much. Also the president of
00:10the
00:10company. As I said, you listed in January. Tell me what your priority has been since that in the
00:17last five months. Actually, we are just focusing on achieving our mission, which is intelligence
00:23with everyone, making the frontier model accessible to everyone. On the one hand,
00:28we are building the four stack foundation models. We are probably one of the very few labs can have
00:34that kind of capability. And on the other hand, we are building user friendly products. So currently
00:40more than 300 million users are using our product and more than 1 million enterprise customers and
00:47professional developers are using our models right now. So which is five times versus six months ago
00:54before our IPO. And also our revenue growth year over year is much faster than the spending. So year
01:02over year, last year, the revenue growth is 160%, much higher than the 33%, which is the R&D year
01:09over
01:09year spending. So you will see the sustainable of the business and the group. So how do you improve
01:15your prospects for profitability when you have rising inference costs versus that lower pricing pressure
01:23on API, right? And then again, that's all despite revenue nearly tripling in 2025?
01:32Actually, so our ARR more than doubled in past two months, even probably out of our expectation.
01:41And our gross profit margin doubled to 25% in the past year. The last model we released as a
01:50language
01:50model and 2.7, the gross profit margin of that model, even higher than that specific number. So you
01:57will see all this drives the potential probabilities in the future. And also, I think the foundation models,
02:04the model performance is all the key for this. And our innovations on producing these kinds of model is a
02:11key. For example, we are doing the end to end optimization, both to hardware and software, even
02:18for the whole group. How can you prevent, though, these commoditized AI agents from essentially getting
02:26into a price war, like what we're seeing in EVs and other areas? I'm seeing Deep Seek will make
02:32permanent a steep discount on its flagship V4 Pro model. And then again, you have a difficulty when
02:41prices are coming down to get margin. But you're saying even on your API product, which is one-fifth
02:48the cost of the similar models in America, you're saying you're having higher margins now? How so?
02:55Yeah, yeah. So I think so agent and the models are really important for monetization. But definitely,
03:03the foundation model is a key. You will see a better performance models with specialization and
03:09differentiation will drive the token consumption, also drives the enterprise and the consumers retention
03:15and the consumption. The model performance product monetization will become the flyware. So you will see,
03:22we did the end-to-end optimization with the whole group. There are lots of innovation, technical
03:28innovation inside. So you will see we can provide probably similar performance models with probably
03:33lower price, even sometimes higher margin. Yeah. So how do you change your revenue mix where most of
03:40your revenue is coming from your consumer-facing products, like the chat box, Xingye, and also
03:46Hylio, which... That's last year's... That's like the text-to-video generation. But next,
03:52Shira, as you go into your Model 3, right, from 2.7 to 3, how is your revenue mix going
03:59to change?
04:00Yeah, yeah. So the number you mentioned is last year's number. But right now, it's almost like 50
04:04enterprise and 50 consumers. So the enterprise increase a lot. And also, yes, the model is a key.
04:12So we think the model is our product. No matter it's a B2B or G2C, it's all the channels for
04:18the
04:18commercialization. So we've spent most of our resources and spending on the model layer. Yes,
04:25we are going to release M3 very, very soon in a few days, which probably, I think, which is the
04:30first
04:31open source native multi-model. Yeah. Now, you also know very well, and the market has responded quite
04:37favorably, that the Hang Seng will be adding you to the Hang Seng tech index as of June 8th,
04:44I believe. And because of weighted structure stock listings, you might have to wait until August
04:49to be potentially added to the Stock Connect. What kind of southbound stock investor interest do you
04:58expect to see once you're added to that Stock Connect at a time when China's trying to clamp down
05:06on cross-border capital flows? At first, we can only control what we can control, right? So we're
05:14just focusing on providing a really great model to the market, provide the user-friendly products to
05:21the market, and also probably the best alternative solutions to some really expensive models with
05:27a better ROI to the market. All this stuff, what we can control. So others, we just left to the
05:34market. And to regulators, because there's also concern about those capital flows coming out. But
05:41again, I want to come back to the Stock Connect. How important is that to be able to have your
05:47stock
05:48potentially available to mainland southbound investors? Right now, they're not able to in a
05:55great large number because of restrictions on the availability of money from China to Hong Kong.
06:01Yeah, I think that's great to have more capital inflow from the domestic market. Yes, it is. But I think
06:09all these great investors, they focus on your company's performance to see what kinds of model with
06:16differentiation you are going to deliver and to see what kinds of products you are delivering. And also, probably in
06:22the
06:22past year, we're only spending like 1% or 2% capital spending compared to the big gens. Yeah, so
06:29you will see the
06:30capital and a group efficiency.
Comments