The Big Model To B Siege

You have a total of 89 million people with a total of one hundred people with technology, plus hundreds of millions of water and electricity benefits, but you only earn me 4 to 50 million a year. Do you say I still need to raise you?

After hearing the big boss's half -joking fun, as NetEaseFor a member of TO B, Zhang Zhi (pseudonym) had to re -examine his situation and seek retreat as soon as possible.He aimed at a rare track that needed to enter the game now -a big model.

The big model sales circle in the TO B direction was very small. Zhang Zhi quickly pushed the friends in the circle to get several OFFER and HC interview opportunities.

However, the result after he runs around is not satisfactory, either he hopes that there is still a certain technical background on the basis of traditional sales skills and relationships, and he can output a complete pre -sales technical solution by himself; either a fraud scheme, the so -called belt, the so -called belt, the so -called belt, the so -called belt, the so -called belt, the so -called belt, the so -called belt.HC's interview is just a routine to obtain commercial thinking, and the interview has also become two people performing each other.

The former is intended to get the sale of the sale, and it is not necessary to say more; although the latter is the meaning of the drunkard, it still reveals that the main task of the model service provider is at this stage.B -side seeks the reality of AIGC's realization.

The benchmark effect of the large factory competes with the big customers, and the entrepreneurial unicorns have also begun to focus onTO B direction transfer: Even the 900,000 government -enterprise direction project system is not allowed even if it is a 900,000 government -enterprise direction project.

Dangdang modelThe C -side story moves towards the B side, which seems to be another reincarnation of the last round of AI boom.The point is that, with the corporate API falling into the vortex of price war, only the remaining project orders that can bring about cash flow.The structural contradictions of the project system have not changed due to the addition of large models, but instead have become more and more trend.

Unfortunately, the model service provider at this stage really rely onFutures -like project ordering orders are difficult to survive.

Central factoryGhost story

Even if the market value has entered the top five of China stocks many times, the annual revenue exceeds 100 billion yuan, and NetEase is still used to claim to be a Chinese factory.

This is not exactly self -modest, but itThe business characteristics of strong branches and weakness have caused NetEase to be limited to organizational ability and habits when exploring new businesses, and it is difficult to mobilize sufficient resources at one time to complete the work.Taking the commercialization department of large models as an example, 60 % of the production and research needs to cooperate with the game business, and the remaining production and research resources that cooperate with commercialization can be imagined.

Today, the department seems to have left only two or three leaders who are responsible for shooting, and have been completely marginalized in the organization. This process is only over a year.time.We try to restore this from different anglesXiao Fu's middle plant, chasing the story of the big model.

In early 2023, ChatGPT's red border was north and south, opening a new round of scientific and technological growth narrative.At that time, almost all of the sounds from the major Internet factories were to seize this round of opportunities.Regardless of whether it can open up a new growth curve, it is enough to reuse the results of the commercialization of science and technology in the past.

Always carrying the game business, very fewThe same is true of NetEase that does not do business.However, NetEase AI needs to be reused not to technology and research, but the ability of commercialization.

A person familiar with the matter told the photon planet that last year, NetEase, NeteaseThe to B team began to explore the commercialization of large models.Most people inside Netease are born in production and research and can make C -end products, but the overall overall is not mature in the commercialization of To B, he said.

In the development path, NetEase is completely opposite to other model service providers.Friends are the first of the model base, and then expand the scene of the scene after having a good general ability.On the other hand, NetEase is simply encapsulated to the expert model used internally. Although the internal function reuse is made by a large factoryTo B's underlying logic, but the game tracks in NetEase are separated from other industries, and the scene leap is a great challenge.

It is worth mentioning that NetEase'sAI Studio Fuxi belongs to NetEase Thunder.As early as the organizational structure, it showed that NetEase's positioning of the game as the first landing test field for AI.

Taking Dan Qing, a model of Wenshengtu released by NetEase in the first half of last year, as an example, as a product that is quickly packaged, it is a product that directly goes to Malaysia to commercialize.

Specifically, NetEase has once had a Wensheng map model with a fast -moving consumer brandThe cooperation of API calls, it is reported that it should beThe amount of calls used during the launch can not be over 10,000, fromThe caliber of UV and PV is not very optimistic.It ’s not to mention that since mid -year, the API calls of the basic model have already been made into cabbage prices, and the revenue of TO B commercialization can eventually only fall into customized projects.

The roots behind this lies in the splitting of the scene. The result of the large models of itself is uncontrollable, and the scene problem further exacerbates this.

Now in the game field, we can see the intelligence such as "Inverse Water Cold"NPC, face pinch, and the achievements of AI teammates such as "Everlasting and Forever" have launched C -terminal AIGC products based on business attributes, but these capabilities cannot be quickly moved to other companies.

Born in business but trapped in business, the real NetEaseAIGC seems to have been living in the shadow of other star business. It is difficult for the outside world to have a sense of the story, and it is difficult for the outside world to perceive it.

The gap between products and solutions

For the commercialization of an innovative business, NetEase's zero -zero time to the department for a year and a half, it is already more patient and determined in the industry.If it is not for the main business to face some revenue pressures and the needs of the marketing and resource replacement of the game business, the operating space for the team may be greater.

For example, the team hasBD has proposed some sales clues of the segmented industry, and the size of the project ranges from millions to tens of millions, but most of the projects have not been able to enter the eyes of the boss, because it does not match the tone of the game industry or cannot be in marketing in marketing.Make resource replacement at the level.

However, just like the current market is full of hundreds of thousands to one million levelsLow -price orders are generally. Even if the screening of customers is released, it still cannot bring much about flowing water and profits under the underlying logic of product delivery.

Because the problem that restricts NetEase and even most model service providers has never been unable to make products, but products that are difficult to drive by technology to switch to operating drivers solutions.The shortcomings are not in the product itself, but to support human resources such as services and operations that support the entire long delivery cycle.

The so -called solution is not only a product matrix composed of a single scene and problem in the actual transaction scenario, but a different product and services assembled for customer needs.Complete delivery.In fact, it is exactly the continuous paymentpayThe service narrative allows SaaS to get out of a hammer to buy and sell the quagmire, and continues to build cash flow for To B companies.

A big model sales also revealed our own voice:Most customers whether they are my own resources or ecological resources, at least they hope to give them one to explain, so as not to let them smash their rice bowls.However, many model service providers delivered after -sales service products that will be smashed by Party A's rice bowl.

From another perspective, this is also this round this yearIn the tide of AGENT, model service providers of different scale of Alibaba, byte bean bags, and wisdom spectrum AI actively expand the endogenous reasons of ecological partners.Ecological partners are middleware between self -technology and landing. After the last mile, they also take off the hard work of subsequent services.

To put it bluntly, holding a hammer to find a nail, the big factory is unwilling to do it, and the Chinese factory can't do it. You can only throw it to a startup.

It should be pointed out that it is not a good thing to help find a nail.Along with the composition of ecological partners, the division of the composition of ecological partners is not much friendly than before, but it can be streamlined from the perspective of entropy reduction; on the other handProduct hedging that must be lagged behind in different directions such as RAG and productization cycles is constantly overdrawing the expectations of AI in various industries.

When an educational group contacts a model service provider, he proposed the groupLOGO is put into the need to generate a fixed position of the picture.The service provider team kept PROMPT, and the conference room gradually fell into embarrassment.

The level of AI can at least reach 60%of the design, but you can't even do 20%. In its expected, AI, which is full of stormsThe situation is not even reduced.

Big modelTo B, still traditional to B

GPT5 is far away, and Sora has not got out of the futures stage.The main contradictions of the model service provider to make large models commercialization are no longer technology, but from Party A.Especially, SaaS, as the shortest path of AI on the side of the enterprise, has been wrapped in recent years.

It can even be said that large models are used in essence and traditionThere is no difference between to B service: there is no additional premium, no differentiation of the track, and even under the uncertainty of the black box, the pressure on the delivery cycle and the cost side will be stronger than before.

A certain medicineAs an example, the SaaS enterprise is originally 3,000 yuan/year. After introducing the ability to call API, in order to quickly break the payment wall, the single store price is reduced to 300 yuan/year.

As for the delivery cycle, in the case of relatively large project orders, when there are more monks and less porridge, the head customers are as traditionalTO B generally performs.

CTO and CIOs first went to the reports of the solutions for the solutions of Ali, Tencent, Huawei and other large factories. It took almost two months after a round.After walking, you have to push the internal push to the boss. When two weeks will be established, there will be three or five million in the first phase of the project.

Zhang Zhi Tu Tu Tu Tu Tu Tu Tu Tao, for half a year,The cost of BD and customized production research determines that the project will inevitably lose money, but for the benchmark effect of big customers, you have to do money.Some large customers can not appear on the customer wall of the service provider for special reasons. For example, Netease once used the structure of the mall and has done the e -commerce app of Moutai. This project is rarely known so far.

IfIn the form of Agent integration, it is more friendly to the corporate side from the perspective of the simple SaaS service and try to reach the corporate side.It's just that this will also encounter the problem that traditional To B encounters when iterated production tools -internal tissue resistance.

Taking the scenes of large models on the side of the government and enterprises as an example, the scene of the book on the side of the government and enterprises is an example. The demand of this part of Wen Shengwen can easily affect the distribution of interests in the organization.

A person familiar with the matter told the photon planet that Alibaba Cloud had come into contact with a design institute customers until the display, demand polishing, and preliminary development before the internal projects were unusually smooth, but they couldn't move in the process of internal project establishment.This is related to the rice bowls of the key departments of the bidding bidding. The IT department could not push it at all, and the bidding meeting was half -dead.

In addition, if traditionalTo exam the model service of to B, it is not difficult to find that the importance of ecological partners is emphasized by different model service providers. In addition to human and organizational considerations, it may be more about losing their pricing capabilities.Generally, the price of calling with API, the low competition affects the future of the commercialization of large models, and the self -consistent logic needs more exploration.

As for the startups of the model layer, when the market continues to cool down, it will not be able to take out enough results, and maybe only the large manufacturers have been acquired.