Following the "All in" large model, iFLYTEK (002230.SZ) is currently in a period of simultaneous harvest and investment.
iFLYTEK released its semi-annual report for 2024, with financial statements showing that the company achieved a revenue of 9.325 billion yuan, a year-on-year increase of 18.91%; while the net profit attributable to shareholders of the listed company was a loss of 401 million yuan, compared to a net profit of 73.572 million yuan in the same period last year, turning from profit to loss year-on-year.
Behind the loss, reporters learned from iFLYTEK's 2024 semi-annual performance briefing on August 22 that iFLYTEK's total investment related to large models in the first half of 2024 reached 1.3 billion yuan, with an increase of 400 million yuan in investment in general large model research and development.
Despite the loss in this year's semi-annual report, iFLYTEK Chairman Liu Qingfeng stated that in the first half of 2024, the company's gross profit was 3.7 billion yuan, an increase of 19.08%. At the same time, the company's per capita revenue and per capita gross profit increased by 120,000 yuan and 80,000 yuan respectively on the basis of 2023. The company achieved good results in improving quality and efficiency in the first half of 2024.
Advertisement
Large models empower revenue growth year-on-year, and actively adjust revenue structure
Financial report data shows that under the accelerated promotion of large model landing, iFLYTEK achieved a revenue of 9.325 billion yuan in the first half of 2024, a year-on-year increase of 18.91%.
iFLYTEK's main business includes the education field (educational products and services, teaching business), smart city (information engineering, smart governance, digital government), open platform and consumer business (open platform, intelligent hardware, mobile Internet products and services), operators, smart cars, smart medical care, smart finance, and other businesses.
From the perspective of business classification, iFLYTEK disclosed that in the first half of 2024, the company's education products and services achieved a revenue of 2.86 billion yuan, a year-on-year increase of 30.66%; medical business achieved a revenue of 228 million yuan, a year-on-year increase of 18.80%; open platform achieved a revenue of 2.345 billion yuan, a year-on-year increase of 47.92%; intelligent hardware achieved a revenue of 900 million yuan, a year-on-year increase of 56.61%; automotive business achieved a revenue of 350 million yuan, a year-on-year increase of 65.49%.
However, in terms of G-end business, the revenue of digital government industry application business fell by 36.33% year-on-year, and information engineering business fell by 21.76%.
iFLYTEK Director, Vice President, and Secretary of the Board Jiang Tao said at the semi-annual performance briefing that in the first half of the year, iFLYTEK actively adjusted the revenue structure, implemented a more cautious business strategy for some G-end digital government projects, and the business structure of iFLYTEK also changed significantly. Although the G-end business revenue decreased by 9%, and the business revenue ratio also decreased, the B and C-end business revenue increased by 34% in the first half of the year, especially under the empowerment of large models, the sales volume of C-end hardware increased significantly, and the landing of B-end further accelerated.In addition to maintaining a good increase in revenue, iFLYTEK achieved a gross profit of 3.748 billion yuan in the first half of this year, a year-on-year increase of 19.08%. Among them, the gross profit in the second quarter increased by 21% year-on-year, and the increase in gross profit for the entire second quarter was higher than the 14% increase in revenue for the second quarter.
Jiang Tao said that iFLYTEK's position as the national team in large models has become more solid. On the one hand, iFLYTEK's technology has maintained a leading position in the industry under the condition of national computing power, and its capabilities are fully comparable to GPT-4Turbo. On the other hand, in terms of market share, it is also in a leading position. The intelligent hyperparameter release in July 2024 shows that iFLYTEK's number of large model bids in the first half of the year and July are in a leading position in the industry. iFLYTEK also took the lead in participating in the formulation of various industry standards, such as jointly proposing the international standard ISO/IEC 5259-4:2024 with the China Electronics Technology Standardization Research Institute, which is also the first international standard in the field of artificial intelligence data quality led by China.
Investment affects short-term performance, and it is expected to maintain stable investment in the second half of the year.
Under the condition that iFLYTEK's gross profit increased by 601 million yuan in the first half of the year, the net profit attributable to the parent company decreased by 474 million yuan year-on-year, and the net profit excluding non-recurring gains and losses decreased by 179 million yuan compared to the same period last year.
Industry insiders believe that this is closely related to iFLYTEK's increased investment in artificial intelligence. It can be seen that in the first half of 2024, the company's R&D investment reached 2.19 billion yuan, a year-on-year increase of 32.32%, accounting for 23.5% of revenue.
"The most important factor for high investment is that the management believes that the current period is still a key historical opportunity for the layout of artificial intelligence, and it must be firm in the source of the model and the landing of applications." Jiang Tao introduced that in the first half of the year, iFLYTEK's total investment related to large models exceeded 1.3 billion yuan, of which the investment increased by 400 million yuan around the N part of the general large model 1+N. The R&D investment around the large model training and reasoning platform engineering increased by 160 million yuan, the investment around the promotion and application of large models increased by 120 million yuan, and the investment around the productization of large models by each BG and BU increased by more than 600 million yuan. These investments will affect the company's operating performance in the short term, but the company's path and rhythm in the commercialization of large models and self-blooding are becoming clearer and clearer.
Liu Qingfeng emphasized that increasing R&D investment is due to confidence in the future. "If we choose to be conservative and save all these investments, it seems that profits will come out immediately. We are not unaware, we are all good at mathematics, we have been entrepreneurs for so many years, and we are all shareholders of the company, why don't we do this? Why do we explore in many places? It is because we do not want to miss this major revolutionary opportunity and stand on the same platform with all the top players."
Regarding the expected scale and scale of investment in the second half of the year, Liu Qingfeng said that the investment will basically remain stable. Looking at the investment of each BG and BU around the productization of large models, this link, as the first curve of large model monetization, will completely follow the market business strategy and will not suddenly invest a large amount of funds. In terms of investment in the underlying capabilities of large models, it will always be compared with GPT-4 and GPT-5, and the investment in R&D personnel will be enough to maintain the current team.
Liu Qingfeng said that in the next few years, the company will not have a large increase in its own funds for the development of large models, and it can maintain a leading position in the industry in terms of large model performance with the support of relevant policies on computing power. According to the Scaling law research, the larger the training parameters of the large model, the stronger the underlying capabilities of the model may be. As the national team of artificial intelligence undertaking national strategic tasks, iFLYTEK has taken the lead in building a Wan card cluster Feixing No.1 through independently controllable artificial intelligence chips, and for the first time, it has trained a large model with hundreds of billions of parameters through independently controllable computing power, clearly proposing "to give full play to the advantages of the new national system and support有实力的 enterprises to take the lead in major research and development tasks". It is expected that iFLYTEK will receive policy support for computing power from the state and governments at all levels in the future. In the next few years, relying on the major layout of computing power supported by the state and the government, the company will not have a large increase in its own funds for computing power, and it can also maintain a leading position in the comprehensive performance of the Spark large model in the industry.
"iFLYTEK still has a gap in computing resources compared to Internet giants, but it will still maintain a leading position in the comprehensive performance of large models, and will train stronger models with less computing power. The improvement of computing power efficiency in the process of training large models is one of iFLYTEK's core competencies in the field of artificial intelligence." Jiang Tao said that iFLYTEK's overall investment in computing power in the future is controllable. The company will be confident in maintaining the industry-leading level of basic general large models and winning the commercialization battle of large models without a large increase in its own computing power asset investment in the future.It should be noted that large models have boosted company revenue growth, but not only iFLYTEK is in a situation of incurring losses due to high investment.
The reporter noticed that 4Paradigm (06682.HK) recently disclosed its financial report for the first half of 2024, showing that the company achieved a revenue of 1.867 billion yuan in the first half of the year, a year-on-year increase of 27.1%. It has made progress in the commercialization of large model technology, launching more than 30 artificial intelligence products and achieving large-scale application in 14 industries. However, the net loss was 169 million yuan. At the same time, the company's R&D investment in the first half of the year was 850 million yuan, a year-on-year increase of 28.4%.
Internet industry analyst Ding Dao Shi analyzed for the reporter that for current large model companies, they are basically in the strategic investment stage. Strategic investment is for future strategic profits. Current profits are not so important for large model companies. All large model companies are racing to lay a solid foundation instead.
Relevant people judge that in the next one to three years, large models that achieve large-scale commercialization and landing have parameters ranging from tens of millions to hundreds of billions. In the future, the combination of general models and special models, by understanding semantics and calling models with hundreds of millions to tens of billions of parameters to complete tasks, will have better results than models with trillions of parameters.
Leave A Reply