Wang Xing's move not only helped Wang Huiwen unwind, allowing the latter to recuperate without further worries, but also gave those investment institutions that converged on Lightyear Beyond because of Wang Huiwen a chance to leave the company smoothly. According to Meituan's announcement, after completing the wholly-owned acquisition of Lightyear Beyond, all of the original Series A shareholders will exit at par.

In addition, the takeover by Meituan and Wang Xing will also help stabilize the existing team at Lightyear Beyond. The large model startup team, which took four months of hard work to assemble, was largely pulled by Wang Huiwen's personal charisma. With the downfall of Wang Xing, the shortcomings of a strong founder due to Wang Huiwen were once again made up.

The acquisition of Lightyear has been equally beneficial for Meituan. In the race of big tech companies competing for big models, Meituan was able to harvest an established big model team and is expected to use it to accelerate its R&D progress in the field of big models. As of June, the team size beyond light years reached about 70 people.

According to Leopard Change, at the same time Wang Xing invested in Wang Huiwen, Meituan also started to develop big model products internally. The S-team, the highest decision-making body headed by Wang Xing, is extremely concerned about the big model within Meituan. About every week or two, Wang Xing would ask the head of the algorithm team about the progress of the big model. Now, with the addition of the Light Years Away team, the pressure on the internal algorithm team of Meituan will be reduced by a lot.

What's more, Meituan, which is ramping up its content to counter rising local life invaders like Jitterbug, is making video and live streaming one of the key directions for its business right now, and the big model products represented by ChatGPT will help Meituan make up for its shortcomings in the content ecology field.

Before Meituan, Baidu founder Robin Li, who released the big model product one step ahead of time, mentioned that using artificial intelligence to generate short video content is a typical application of AIGC, which can "create opportunities for the platform to attract new users, increase user time spent and user stickiness".

However, in the path of realizing the above vision, Meituan still faces multiple challenges, including but not limited to the ongoing cost of training and running large models, as well as the reality of how to truly commercialize large model products.

In contrast to these long-term concerns for Meituan, investment institutions and some big-model entrepreneurs have already started to usher in challenges in a tangible way.

Investment institutions hope to rely on betting light years away to harvest the next BAT in the AIGC era dream, with the acquisition of Meituan abruptly ended.

Two days before this record-breaking domestic AIGC acquisition took place, on June 27, big data giant Databricks set a global record for the largest AIGC acquisition by acquiring artificial intelligence startup MosaicML for about $9.3 billion.

Compared with MosaicML's direct six-fold acquisition price, the scenario behind which Lightyear Beyond could only exit at a flat price may have long been filled with the despondency of investment institutions.

More importantly, the acquisition of Light Years Away by MosaicML objectively gives a new footnote to the recent big model debate between Zhu Xiaohu and Fu Sheng, namely that the generic big model is not a game for entrepreneurs.

Li Ge, founder of aiXcoder, an AIGC startup that focuses on automatic code writing, describes his feelings this way, "Entrepreneurs must choose their field well, and if the field you choose is just a shallow application of the big model, your moat is too low and can easily be covered off."

01

Before choosing to sell to Meituan, Wang Huiwen, who realized that something was wrong with his body, has started to take the initiative to contact with a number of major technology companies. According to Zhang Peng, founder of Geek Park, Wang Huiwen contacted Tencent, Byte, Meituan and Racer to discuss various potential solutions.

In addition to the fact that Wang Xing and Wang Huiwen used to be roommates in Tsinghua's bunk beds and co-founded Meituan, they may also be related to the Meituan system of Lightyear.

In February, after announcing that he was going off to found Lightyear and join AGI, Wang Huifen recruited the No. 2 employee of Lightyear - Liu Jiang, the vice president of Beijing Zhiyuan Artificial Intelligence Research Institute, who had been the director of Meituan's technology academy; in March, the "father of Sogou Input Method In March, the "father of Sogou Input Method", Zhankai Ma, joined Beyond Lightyear, who had been a product consultant for Meituan since 2010.

With the disclosure of Meituan's acquisition announcement, more details about Lightyear Beyond were revealed. In terms of equity, AI Age, which is wholly owned by Wang Huiwen, accounts for 76.72%, Qimai, which is wholly owned by Wang Xing, accounts for 0.43%, Sequoia China accounts for 2.44% and other investors account for 20.41%.

It is worth noting that since Wang Xing and Sequoia China founding partner Shen Nanpeng are both shareholders of Lightyear Beyond and directors of Meituan, both of them abstained from voting on Meituan's board resolution to acquire Lightyear Beyond.

The approximately $2.065 billion acquisition payment made by Meituan includes approximately $234 million from the offshore shareholders, approximately $367 million in domestic convertible bond debt, and the $1 required to be paid to Wang Huiwen.

Specifically, the approximately $234 million acquisition payment outside of China includes $5 million from Wang Xing, $28 million from Sequoia China, and approximately $200 million from other investors.

About 367 million yuan of debt paid within the territory of Meituan, in the opinion of industrial investor Jiang Yi, is very likely to be the convertible bond debt attached to the first-class technology acquired outside of light years, and the main purpose of the first-class technology to issue convertible bonds to the outside world, perhaps for financing.

According to the announcement of the acquisition of the American group, the main asset of the territory beyond the light years is the equity of first-class technology. after the completion of the acquisition of first-class technology in April, 2022, first-class technology audited book value and net asset value of about 15.59 million yuan and 570,000 yuan, respectively, the same period first-class technology loss of about 19.18 million yuan, in a state of excess income over expenditure.

As of June 29, Lightyear Beyond still had about $285 million in cash on hand, equivalent to about RMB 2.1 billion, which also means that Meituan acquired Lightyear Beyond at almost zero cost.

02

Despite picking up the slack of a good brother, Wang Xing and Meituan still have a lot of challenges to face.

After the acquisition is completed, the financial results of Light Years Beyond will be consolidated into Meituan's accounts. The training costs and operating costs incurred in the development of large models may be directly reflected in Meituan's financial results for the third quarter of this year, and more importantly, the blood transfusion to Light Years Beyond will affect Meituan's overall profitability for a long time.

Behind ChatGPT's ability to be so smart and get its products off the ground quickly is an expensive supercomputer tailored by Microsoft. in 2019, when Microsoft invested $1 billion in ChatGPT developer OpenAI, it agreed to build a massive, cutting-edge supercomputer for the artificial intelligence research startup.

Even with Microsoft's help, Open AI, which has already burned through billions of dollars in funding, still cautions the public that it may need at least $100 billion more to finally lead to the AGI era.

For Meituan, perhaps not with such a huge scale of capital investment, but want to be among the first echelon of large models, a steady stream of financial support, will be a necessary condition.

Another challenge is that, compared to Baidu, Ali, Tencent and even bytes and other domestic technology makers, Meituan's shortcomings in cloud services are also one of the obstacles to its competitiveness in enhancing the big model.

Meituan once also wanted to do a big job in cloud services, and in 2013 launched a public cloud computing service platform, two years after the trial operation, Meituan cloud officially independent. In 2017, Meituan Cloud officially announced its platform strategy to join the public cloud competition, and launched GPU cloud hosting to enter the AI field.

But under the delayed success of market development, in March 2020, Meituan Cloud made an announcement to the public that it would stop the service and support to users and recycle resources in May of the same year, and in the future, Meituan Cloud would only serve internal own business.

After the arrival of the big model wave triggered by ChatGPT, Baidu, Ali, Tencent and bytes, on the one hand, launched their own research on big model products, on the other hand, launched big model service platforms based on cloud services one after another.

In contrast, Meituan, which accelerated in the big model track by acquiring light years away, is missing a path to promote the commercialization of big models.

03

The long-term challenge in the field of big model is not just a problem for Meituan.

In order not to miss the wave of the times that is enough to compare with the mobile Internet change, investment bigwigs and entrepreneurial bigwigs are quickly moving towards alliance.

Wen Jirong, executive director of the High Tide Institute of Artificial Intelligence at Renmin University of China, even optimistically predicts that in the next 5-10 years, there will definitely be a super intelligent brain, "Now that the principle is through, all that remains is the engineering progress problem. (After the arrival of AGI) everything now, will be rewritten."

To avoid falling behind in the flood of rewriting destiny, investment institutions use a huge check to bet on those entrepreneurs they trust.

Wang Huiwen is one of them. With the exit of Lightyear Beyond, in Jiang Yi's view, it might be a good thing for other big model startups in the same latitude, as they will hopefully share the resources that originally belonged to Lightyear Beyond.

A founding partner of the fund mentioned in an interview with Tencent Technology that on the ten or so big model startups where Wang Huiwen is located, which are receiving attention and injection of capital from investment institutions across the industry, "everyone is focused on investing in these ten or so projects, and the ones that can't finance the money can't finance the money."

The reason why investment institutions are afraid to cast a wide net is that the big model boom has been prevalent for half a year, VCs have not seen the real bright application scenarios and enough closed-loop business models, as Zhu Xiaohu recently argued with Fu Sheng "how big is the value of big model startups" said, entrepreneurs should not be superstitious generic big model. "Scene first, data is king".

But the reality is that there are very few startups that meet the above eight characteristics. aiXcoder founder Li Ge told Box Lunch Finance that the big model wave will spread for a long time to come, but the key is that "entrepreneurs have to find their own position in this wave, so you can't say that you're particularly anxious when Meituan acquires Lightyear, or when another event happens. The thing that needs to be anxious is no head."

Li Ge also agrees with Zhu Xiaohu's view that "entrepreneurs are not suitable to do generic large models". In his opinion, entrepreneurs must choose their own field, and this field has enough depth. Code automation is the field Li Ge chose for himself, "it is not a big model everything done so a situation, it must be considered in the whole application environment more complex many problems."

Translated with www.DeepL.com/Translator (free version)

Wang Xing's move not only helped Wang Huiwen unwind, allowing the latter to recuperate without further worries, but also gave those investment institutions that converged on Lightyear Beyond because of Wang Huiwen a chance to leave the company smoothly. According to Meituan's announcement, after completing the wholly-owned acquisition of Lightyear Beyond, all of the original Series A shareholders will exit at par.

In addition, the takeover by Meituan and Wang Xing will also help stabilize the existing team at Lightyear Beyond. The large model startup team, which took four months of hard work to assemble, was largely pulled by Wang Huiwen's personal charisma. With the downfall of Wang Xing, the shortcomings of a strong founder due to Wang Huiwen were once again made up.

The acquisition of Lightyear has been equally beneficial for Meituan. In the race of big tech companies competing for big models, Meituan was able to harvest an established big model team and is expected to use it to accelerate its R&D progress in the field of big models. As of June, the team size beyond light years reached about 70 people.

According to Leopard Change, at the same time Wang Xing invested in Wang Huiwen, Meituan also started to develop big model products internally. The S-team, the highest decision-making body headed by Wang Xing, is extremely concerned about the big model within Meituan. About every week or two, Wang Xing would ask the head of the algorithm team about the progress of the big model. Now, with the addition of the Light Years Away team, the pressure on the internal algorithm team of Meituan will be reduced by a lot.

What's more, Meituan, which is ramping up its content to counter rising local life invaders like Jitterbug, is making video and live streaming one of the key directions for its business right now, and the big model products represented by ChatGPT will help Meituan make up for its shortcomings in the content ecology field.

Before Meituan, Baidu founder Robin Li, who released the big model product one step ahead of time, mentioned that using artificial intelligence to generate short video content is a typical application of AIGC, which can "create opportunities for the platform to attract new users, increase user time spent and user stickiness".

However, in the path of realizing the above vision, Meituan still faces multiple challenges, including but not limited to the ongoing cost of training and running large models, as well as the reality of how to truly commercialize large model products.

In contrast to these long-term concerns for Meituan, investment institutions and some big-model entrepreneurs have already started to usher in challenges in a tangible way.

Investment institutions hope to rely on betting light years away to harvest the next BAT in the AIGC era dream, with the acquisition of Meituan abruptly ended.

Two days before this record-breaking domestic AIGC acquisition took place, on June 27, big data giant Databricks set a global record for the largest AIGC acquisition by acquiring artificial intelligence startup MosaicML for about $9.3 billion.

Compared with MosaicML's direct six-fold acquisition price, the scenario behind which Lightyear Beyond could only exit at a flat price may have long been filled with the despondency of investment institutions.

More importantly, the acquisition of Light Years Away by MosaicML objectively gives a new footnote to the recent big model debate between Zhu Xiaohu and Fu Sheng, namely that the generic big model is not a game for entrepreneurs.

Li Ge, founder of aiXcoder, an AIGC startup that focuses on automatic code writing, describes his feelings this way, "Entrepreneurs must choose their field well, and if the field you choose is just a shallow application of the big model, your moat is too low and can easily be covered off."

01

Before choosing to sell to Meituan, Wang Huiwen, who realized that something was wrong with his body, has started to take the initiative to contact with a number of major technology companies. According to Zhang Peng, founder of Geek Park, Wang Huiwen contacted Tencent, Byte, Meituan and Racer to discuss various potential solutions.

In addition to the fact that Wang Xing and Wang Huiwen used to be roommates in Tsinghua's bunk beds and co-founded Meituan, they may also be related to the Meituan system of Lightyear.

In February, after announcing that he was going off to found Lightyear and join AGI, Wang Huifen recruited the No. 2 employee of Lightyear - Liu Jiang, the vice president of Beijing Zhiyuan Artificial Intelligence Research Institute, who had been the director of Meituan's technology academy; in March, the "father of Sogou Input Method In March, the "father of Sogou Input Method", Zhankai Ma, joined Beyond Lightyear, who had been a product consultant for Meituan since 2010.

With the disclosure of Meituan's acquisition announcement, more details about Lightyear Beyond were revealed. In terms of equity, AI Age, which is wholly owned by Wang Huiwen, accounts for 76.72%, Qimai, which is wholly owned by Wang Xing, accounts for 0.43%, Sequoia China accounts for 2.44% and other investors account for 20.41%.

It is worth noting that since Wang Xing and Sequoia China founding partner Shen Nanpeng are both shareholders of Lightyear Beyond and directors of Meituan, both of them abstained from voting on Meituan's board resolution to acquire Lightyear Beyond.

The approximately $2.065 billion acquisition payment made by Meituan includes approximately $234 million from the offshore shareholders, approximately $367 million in domestic convertible bond debt, and the $1 required to be paid to Wang Huiwen.

Specifically, the approximately $234 million acquisition payment outside of China includes $5 million from Wang Xing, $28 million from Sequoia China, and approximately $200 million from other investors.

About 367 million yuan of debt paid within the territory of Meituan, in the opinion of industrial investor Jiang Yi, is very likely to be the convertible bond debt attached to the first-class technology acquired outside of light years, and the main purpose of the first-class technology to issue convertible bonds to the outside world, perhaps for financing.

According to the announcement of the acquisition of the American group, the main asset of the territory beyond the light years is the equity of first-class technology. after the completion of the acquisition of first-class technology in April, 2022, first-class technology audited book value and net asset value of about 15.59 million yuan and 570,000 yuan, respectively, the same period first-class technology loss of about 19.18 million yuan, in a state of excess income over expenditure.

As of June 29, Lightyear Beyond still had about $285 million in cash on hand, equivalent to about RMB 2.1 billion, which also means that Meituan acquired Lightyear Beyond at almost zero cost.

02

Despite picking up the slack of a good brother, Wang Xing and Meituan still have a lot of challenges to face.

After the acquisition is completed, the financial results of Light Years Beyond will be consolidated into Meituan's accounts. The training costs and operating costs incurred in the development of large models may be directly reflected in Meituan's financial results for the third quarter of this year, and more importantly, the blood transfusion to Light Years Beyond will affect Meituan's overall profitability for a long time.

Behind ChatGPT's ability to be so smart and get its products off the ground quickly is an expensive supercomputer tailored by Microsoft. in 2019, when Microsoft invested $1 billion in ChatGPT developer OpenAI, it agreed to build a massive, cutting-edge supercomputer for the artificial intelligence research startup.

Even with Microsoft's help, Open AI, which has already burned through billions of dollars in funding, still cautions the public that it may need at least $100 billion more to finally lead to the AGI era.

For Meituan, perhaps not with such a huge scale of capital investment, but want to be among the first echelon of large models, a steady stream of financial support, will be a necessary condition.

Another challenge is that, compared to Baidu, Ali, Tencent and even bytes and other domestic technology makers, Meituan's shortcomings in cloud services are also one of the obstacles to its competitiveness in enhancing the big model.

Meituan once also wanted to do a big job in cloud services, and in 2013 launched a public cloud computing service platform, two years after the trial operation, Meituan cloud officially independent. In 2017, Meituan Cloud officially announced its platform strategy to join the public cloud competition, and launched GPU cloud hosting to enter the AI field.

But under the delayed success of market development, in March 2020, Meituan Cloud made an announcement to the public that it would stop the service and support to users and recycle resources in May of the same year, and in the future, Meituan Cloud would only serve internal own business.

After the arrival of the big model wave triggered by ChatGPT, Baidu, Ali, Tencent and bytes, on the one hand, launched their own research on big model products, on the other hand, launched big model service platforms based on cloud services one after another.

In contrast, Meituan, which accelerated in the big model track by acquiring light years away, is missing a path to promote the commercialization of big models.

03

The long-term challenge in the field of big model is not just a problem for Meituan.

In order not to miss the wave of the times that is enough to compare with the mobile Internet change, investment bigwigs and entrepreneurial bigwigs are quickly moving towards alliance.

Wen Jirong, executive director of the High Tide Institute of Artificial Intelligence at Renmin University of China, even optimistically predicts that in the next 5-10 years, there will definitely be a super intelligent brain, "Now that the principle is through, all that remains is the engineering progress problem. (After the arrival of AGI) everything now, will be rewritten."

To avoid falling behind in the flood of rewriting destiny, investment institutions use a huge check to bet on those entrepreneurs they trust.

Wang Huiwen is one of them. With the exit of Lightyear Beyond, in Jiang Yi's view, it might be a good thing for other big model startups in the same latitude, as they will hopefully share the resources that originally belonged to Lightyear Beyond.

A founding partner of the fund mentioned in an interview with Tencent Technology that on the ten or so big model startups where Wang Huiwen is located, which are receiving attention and injection of capital from investment institutions across the industry, "everyone is focused on investing in these ten or so projects, and the ones that can't finance the money can't finance the money."

The reason why investment institutions are afraid to cast a wide net is that the big model boom has been prevalent for half a year, VCs have not seen the real bright application scenarios and enough closed-loop business models, as Zhu Xiaohu recently argued with Fu Sheng "how big is the value of big model startups" said, entrepreneurs should not be superstitious generic big model. "Scene first, data is king".

But the reality is that there are very few startups that meet the above eight characteristics. aiXcoder founder Li Ge told Box Lunch Finance that the big model wave will spread for a long time to come, but the key is that "entrepreneurs have to find their own position in this wave, so you can't say that you're particularly anxious when Meituan acquires Lightyear, or when another event happens. The thing that needs to be anxious is no head."

Li Ge also agrees with Zhu Xiaohu's view that "entrepreneurs are not suitable to do generic large models". In his opinion, entrepreneurs must choose their own field, and this field has enough depth. Code automation is the field Li Ge chose for himself, "it is not a big model everything done so a situation, it must be considered in the whole application environment more complex many problems."

Translated with www.DeepL.com/Translator (free version)