iFlytek’s real dilemma: customers become rivals, technological advantages gradually disappear, and profitability track is uncertain

As former customers such as BAT have become rivals, iFLYTEK has had to adjust its business model from 2B to 2C, while its leading edge in voice technology is rapidly diminishing.

iFlytek (002230.SZ) is one of the companies that has been at the center of attention in the past month.

On August 28, iFLYTEK's stock price plunged 6%. Its market value evaporated by nearly 8 billion yuan in two days. A few days later, the capital market turned around and chased the rise. Starting from September 1, it surged for three consecutive days. As of the close of September 15, the stock price was 53.00 yuan, and the market value was about 73.6 billion yuan.

In the past two months, iFlytek's stock price has risen by 116.5%, and its market value has once approached 100 billion yuan. Since its listing in 2008, the company's market value has increased by more than 23 times.

Surrounding the rise and fall of the stock price are the market's very different views on the company's prospects. Bears believe that iFlytek's main business is weak and its profits are unsustainable, and that it is a "big company in the stock market, but a small company in the financial statements" with huge risks; bulls believe that iFlytek is at the forefront of the artificial intelligence trend, with an early start and deep accumulation, and its voice artificial intelligence technology is unmatched, and its future is limitless.

As the business demand for mobile voice processing increases, large companies that have demand for voice technology are trying to use their own voice technology to support their own products. This puts third-party technology output platforms such as iFlytek in an awkward position, and the business model that has been running for many years has been severely challenged.

As a countermeasure, iFlytek began to develop direct-to-consumer C-end business two years ago, which made it a rival of BAT. In the C-end market, iFlytek is far inferior to them in terms of resources and experience. What is more worrying is that with the continuous change of new technologies, thanks to deep neural networks and cloud computing, these rivals are likely to outflank them. Technology is no longer iFlytek's advantage in winning the competition.

Under pressure, iFLYTEK has stepped up its pace of external expansion and has continued to make acquisitions in recent years, but problems such as inefficiency and mixed business have also followed.

These are the underlying reasons behind the stock price fluctuations, and also reflect the market situation of this pioneer in voice intelligence, which is either a big advance or a big retreat.

But iFlytek said it is confident in meeting these challenges. On September 10, in the spacious and bright new office building, Jiang Tao, senior vice president and secretary of the board of directors of iFlytek, told Caixin reporters that no matter what the outside world says, iFlytek will not stop and will continue to invest.

In the past month, the team of iFLYTEK's Beijing branch has grown from 300 to 800 people, and the company had to move from Wudaokou to an office building covering 10,000 square meters in Zhongguancun Software Park.

Tough battle for position

Last October, at Luo Yonghao's Smartisan smartphone M1 launch event, an intelligent voice input method from iFlytek became popular. At the time, iFlytek's market value had exceeded 40 billion yuan, but compared with Smartisan, it was not a star company. The reason why it attracted attention was partly for the opposite reason: consumers could not remember the last time this company appeared.

With the voice technology it has accumulated since its establishment in 1999, iFLYTEK has been accustomed to making money in silence. Before 2014, it did not even have its own marketing department, did not engage in marketing, and did not have an aggressive market strategy.

For a long time, the company's main business has been to provide Chinese voice technology solutions to industry customers including universities, governments, operators and banks.

In order to expand its B-side business, the company launched the iFlytek Voice Cloud Platform in 2010, opening up the underlying technology of voice recognition to developers for them to develop applications. This is also the first intelligent voice platform in China. Tencent QQ, AutoNavi, Ctrip, etc. have all been iFlytek's customers.

Internet companies' reliance on its technology not only helped iFlytek to earn revenue, but also enabled it to dominate the mainstream market for a long time. According to its financial report, from 2010 to 2014, iFlytek's net profit grew at a compound rate of more than 30%. iFlytek claims that it currently has a 70% share of the Chinese voice market.

But the good times seem to be over. Since 2015, given the importance of voice intelligence technology, BAT has entered the market, which has brought a double dilemma to iFlytek: it has lost users and gained competitors. Since 2015, iFlytek's net profit has been declining year by year, averaging only 10%, which is lower than the growth rate of revenue.

In the upcoming artificial intelligence market, the primary beneficiary is not this old company that has been deeply engaged in technology for many years, but BAT, which owns big data and is rich in resources.

In theory, voice recognition technology follows the Matthew effect: the more sound data there is, the stronger the recognition capability. For BAT, a large number of stable C-end users will be a natural treasure trove of voice recognition data.

This technology is also regarded by giants as an entry-level and most easily implemented strategy. Baidu was the first to see the opportunity. In 2012, Baidu established the Deepin Research Institute and quickly put the relevant technology into its search business.

This technology company, which had previously been hiding behind Internet companies, had to step into the spotlight.

On the one hand, it began to vigorously attract small and medium-sized enterprises to find application scenarios for artificial intelligence, and on the other hand, it began to turn to the consumer market and create its own entry-level products. In 2010, it launched the iFlytek input method, and in the following years, it released a number of smart hardware products.

In order to support this strategy, the cost of marketing and other aspects is naturally not small, and the company has increased its capital operations to support this rapid expansion. Since its listing, iFLYTEK has raised a total of 5.207 billion yuan.

The decline in net profit is partly due to the aggressive marketing strategy in recent years. The financial report shows that iFLYTEK's marketing investment has been rising year by year since 2014. In 2016, the overall marketing investment increased by 60%, exceeding 600 million yuan, almost equal to the company's R&D investment.

Jiang Tao said that it was this kind of investment that led to the decline in net profit. Take this year as an example, a large part of the 300 million yuan gross profit in the first half of the year was spent on national channels and marketing networks.

In 2016, Baidu turned to artificial intelligence and invested 20 billion yuan, including a series of investments and talent introduction plans. Baidu offers a considerable annual salary to professionals, averaging 220,000 US dollars, which is the average level of similar companies in the United States.

Tencent entered the market the latest among BAT, but it has also developed rapidly. Ma Huateng once publicly stated that artificial intelligence has become popular, and there are more than four teams in the company conducting AI research and development. In the past two years alone, Tencent has invested in seven companies in related fields.

Industry insiders said that once BAT's voice technology matures, they will use their own technology. In addition to being reluctant to hand over the entrance to others, independent research and development can also reduce costs in the face of huge future demand.

This means that iFLYTEK will lose a major customer and its endogenous growth may be insufficient.

In 2014, iFLYTEK's voice platform provided underlying voice technology support for Sogou Voice Assistant. In less than two years, Sogou has established its own voice team and launched real-time voice translation technology.

Caijing reporters learned that Tencent QQ, another partner of iFlytek, has been a client of iFlytek since 2006, but now all of Tencent's voice terminals use its own technology.

On the other hand, BAT is open-sourcing their own voice development platforms, which has created pressure, at least in the field of voice, which is iFlytek's main business. Observers told Caijing that although the company has advantages in corpus database accumulation and debugging, it will not be able to withstand open source for long.

Disappointing performance followed one after another. The financial report shows that in 2016, the contribution of artificial intelligence-related main business was only about 20%. The bulk of the revenue still came from non-artificial intelligence businesses such as information operation and maintenance, and the "gold content" of the company's revenue was getting lower and lower.

Technological advantages are no longer obvious

iFlytek's intelligent voice technology, on which it is based, is also surrounded by powerful competitors.

In China, iFLYTEK was the first company to study deep neural networks and the first company in the world to apply deep learning to Chinese speech recognition.

Deep neural networks are the biggest breakthrough in artificial intelligence in recent years. This technology has greatly improved the efficiency of previous model training. On the other hand, the emergence of cloud computing has made large-scale data computing possible.

In the past, iFlytek's accuracy could only improve by two or three points a year, but now, thanks to the new technology, it can improve the performance of the past five years in the same period of time. This means that the technical barriers are disappearing.

Many industry insiders said that when it comes to artificial intelligence technology alone, iFLYTEK is not much different from other companies.

Li Zhifei, founder and CEO of Mobvoi, told Caixin: "More than 90% of technologies do not have unique weapons, they are more of a ticket." This speech recognition company has tasted the sweetness of open source technology and has its own independent research and development technology only five years after its establishment.

Despite years of industry experience, iFLYTEK still has advantages in scene data accumulation and model parameter debugging, and is currently the only speech recognition company that can recognize more than 20 dialects. But the key is how long this advantage can last?

Jiang Tao also had to admit this reality. He told Caijing reporters, "Technology is no longer the competitive advantage in the future. The key is to see who runs faster."

However, he still believes in the technological potential of iFlytek, saying that iFlytek's industry-university-research mechanism, which has been accumulated for more than ten years, will provide it with continuous technological innovation capabilities. He listed three joint laboratories at the University of Toronto, Roberts University, and UC Berkeley.

"We keep in touch with the latest theoretical breakthroughs in core technologies in the industry, and these are fed back to the research team in a timely manner," he said.

In fact, iFlytek's R&D revenue ratio far exceeds that of BAT, reaching 20% ​​almost every year. It also spares no effort in acquiring technology companies, and has nearly 1,000 employees focusing on basic research, which also exceeds the latter.

However, after investing heavily, iFLYTEK has not shown any special technological breakthroughs in key issues such as natural semantic understanding and acoustic processing.

Acoustic processing is directly related to the current difficulty of speech recognition: far-field speech, that is, how to break through the noise and recognize the sound in an environment of 5 to 8 meters.

Founded in 2016, SoundAI Technology is a front-end company focusing on acoustic technology. CEO Chen Xiaoliang told Caijing that among his partners are not only iFlytek but also its competitor AISpeech. He said that iFlytek is better at the back-end, that is, the artificial intelligence part.

Li Zhifei believes that the time it takes for giants to deploy voice intelligence technology is getting shorter and shorter. The first to achieve large-scale semantic recognition of voice recognition was Google, but Baidu had its own a few months later. He believes that the advantage of converting technology into products and business models is the main barrier in the future.

Unisound is a startup company that uses Internet thinking to do voice recognition. Just two years after its establishment, it claims to have attracted 20,000 developer users. In the platform and industry fields, Unisound has become a direct competitor with iFlytek.

Jiang Tao said frankly: "The advantages of these companies are that they have ready-made product capabilities, resource integration and marketing promotion capabilities."

Running on the right track is the key

If iFlytek wants to maintain higher value in the future, it needs continuous innovation and accurate profit path. However, at present, the commercialization is single and not optimistic.

"There will be bubbles in any industry. After the concept of artificial intelligence became popular, a large amount of capital poured in, which will increase the valuations of some companies. But the future still depends on how these companies expand and change," Internet expert Liu Feng told Caixin reporters.

According to the classic technology company value analysis, due to technological innovation or new business models, the market has a higher tolerance for profitability in the early stages of establishment, but ultimately it still depends on the changes and growth of the company's profits.

Take Facebook and Twitter as examples. Their poor initial performance did not match their high market value. However, we still need to look at their performance after several years. Facebook will eventually find a profitable path and take off. Otherwise, like Twitter, it will fade out of the market.

In fact, companies in the industry are all looking for a track that can generate stable profits. AISpeech began to provide customized platform services for enterprises, while Unisound focused on the future home IoT and medical fields.

Analysts believe that if iFlytek sticks to the voice market, survival is not a problem, but it will be very difficult to continue to grow at a high speed. This means that iFlytek must quickly find ways to expand the application scenarios of its voice technology or move to other more profitable fields.

Jiang Tao introduced that the company's current largest investment is in education, medical care and consumer businesses. However, only education has seen signs of profitability. This year's interim report shows that education services and products contributed more than 30% of revenue, a year-on-year increase of 57.7%, and gross profit doubled.

iFLYTEK has built a knowledge graph cognitive platform that integrates classroom, tutoring, examinations, etc. The system covers more than 15,000 schools across the country, with more than 15 million active teachers and students.

Take tutoring as an example. The machine can compile the homework completed by students into data that can be processed and analyzed by computers, and provide personalized tutoring solutions for them. iFlytek said that it is digging deep into the commercial operation of this area.

At present, the education team has 2,000 people, and it will expand in the future. Jiang Tao told Caijing that in the future, all the top 100 schools in the country will use an automatic marking system of iFlytek. Currently, the annual report discloses that 68 schools have such system, and this number is still rising.

iFlytek has squeezed out some competitors in the education market. AISpeech announced that it has completely divested its education business. Its former co-founder Yu Kai told Caijing that the education market is huge, but the deeper you go, the more you will be tested on products and data.

This is also iFlytek’s earliest artificial intelligence testing ground. After seven years of hard work, it was able to finally take off thanks to the application data and resources accumulated over the years.

However, in the fiercely competitive field of smart healthcare where iFLYTEK is a latecomer, it is still hard to say whether iFLYTEK can succeed.

In 2016, iFLYTEK entered the medical field with a high profile, bringing in industry leader Tao Xiaodong, a doctor from Johns Hopkins University in the United States. In addition to voice medical record entry, image recognition and medical assistance are also in Tao Xiaodong's product planning.

Competitors are doing the same thing. Take Unisound as an example. Since the company was founded in 2012, medical care has always been a strategic market. Whether in terms of technology, market or business capabilities, it is no less than iFlytek.

Internet companies are also actively making plans. In the past few months, Baidu has released its medical brain, and Ali Health has launched products such as the clinical medical research and diagnosis platform. Tencent is also making efforts. Its first esophageal cancer early screening project using AI medical imaging has entered the clinical pre-trial stage.

These markets have huge potential in the future, and the key to securing a position is who can first obtain more process data, which also tests the ability of resource integration and product operation. Therefore, expansion into new areas is not easy, which means that iFlytek's subsequent growth remains to be seen.

As for the consumer goods sector, which has long lacked a profit model, there is no sign of iFlytek's breakthrough. This business, which required huge initial investment, only contributed 26.9721 million yuan in revenue in the first half of this year, accounting for less than 1%. This is far from the revenue of several hundred million yuan of the leading companies in this field. Since it began to enter the consumer market in 2015, iFlytek has made few breakthroughs. Its genes and organizational structure cannot support its competition with BAT.

China's artificial intelligence market is still growing rapidly, and there are huge variables in the market ranking. Given iFlytek's technology and market accumulation, continued growth is not a problem. For this company, whether it can endure loneliness, isolate the noise and temptation of the capital market, and focus on the company's technology accumulation, industry layout and exploration of business models that suit its genes will determine its position in the future intelligent industry chain.

As a winner of Toutiao's Qingyun Plan and Baijiahao's Bai+ Plan, the 2019 Baidu Digital Author of the Year, the Baijiahao's Most Popular Author in the Technology Field, the 2019 Sogou Technology and Culture Author, and the 2021 Baijiahao Quarterly Influential Creator, he has won many awards, including the 2013 Sohu Best Industry Media Person, the 2015 China New Media Entrepreneurship Competition Beijing Third Place, the 2015 Guangmang Experience Award, the 2015 China New Media Entrepreneurship Competition Finals Third Place, and the 2018 Baidu Dynamic Annual Powerful Celebrity.