SenseTime's 2023 will undoubtedly be stumbling on the thorny road.
As an important shareholder of SenseTime, SoftBank will continue to reduce its holdings in 2023, cutting meat with a blunt knife, and will not forget to continue to reduce its holdings until a few days before the release of the financial report, and Ali will reduce its holdings by 40 million shares of SenseTime in April 2023, with a cash amount of more than HK $280 million. On June 21, Alibaba sold 50 million SenseTime shares at an average price of HK$2.2670 per share due to its own business adjustments.
In November, it was shorted by Grizzly Bear, a U.S. short-selling agency, and the short-selling report said: "SenseTime is artificially (non-)intelligently inflating revenue through highly dubious revenue cycles." The report also depicts through two court cases in which SenseTime provided funds to customers, either directly or through intermediaries, which in turn were used to purchase goods from SenseTime that may have never been delivered.
SenseTime immediately responded: the report is unfounded and reserves the right to take appropriate action.
finally survived until the end of the year, but the founder and soul figure, Tang Xiaoou, unfortunately passed away.
In terms of stock price, it is not an exaggeration to say that the market value of SenseTime is hidden in the dust. In the eyes of shareholders, the former "soup king" is now mixed into a "soup boy".
1.AI is so popular, why is it so difficult for SenseTime?
The difficulty of SenseTime's news in 2023 is vividly reflected in the financial report.
According to the financial report, the company's revenue in 23 years was 3.406 billion yuan, a year-on-year decrease of 10.57%, and the net loss was 6.440 billion yuan, a year-on-year decrease of 6.54%, and the revenue and profit both declined. You know, this has been the sixth consecutive year of annual losses.
From 2018 to 2023, SenseTime has accumulated a net loss of more than 50 billion yuan, with a total adjusted loss of 13.63 billion yuan.
For high-growth technology stocks, even if the early loss can be accepted by the market, the loss of more than 50 billion is still not a small amount. What's more, the total market capitalization of the Hong Kong stock Shangshang SenseTime-W is only HK $23.763 billion.
Losing twice the current market value in six years seems to be an exaggeration.
Losses are not a problem, the problem is that we can't continue to lose money like this, at least we have to see the signs of a financial reversal. SenseTime also understands this, so in 2023, SenseTime has identified three new business segments: generative AI, traditional AI, and smart cars.
After the adjustment, the biggest change is that the business related to smart cities has been reduced. This part of the revenue exceeded 47% in the first half of 2021 and decreased to less than 10% in 2023.
The financial result is that although it lost 6.4 billion yuan in 23 years, compared with the loss of 17.1 billion yuan in 21 years, it has basically stopped the scale of the expansion of losses. Obviously, the business that once accounted for the main part of the core revenue is constantly losing money.
Stopping losses also comes at a price.
Although SenseTime took the initiative to adjust the proportion of business, this is actually not an optimization of the business structure, because the loss of such a large part of the business resulted in 23 years of revenue only 72% of 21 years. It may not be an exaggeration to say that it is a self-cutting income.
Why does SenseTime prefer to "cut off one arm" in revenue to stop losses?
The reason is actually very simple, no matter how high the revenue is, it is difficult to convert it into profits.
From 2018 to the present, SenseTime's revenue has exceeded 17 billion, but it is still difficult to turn revenue into profits. The answer may lie in the fact that the impairment of accounts receivable is too large.
Judging from the data released in the 23-year financial report, SenseTime has a lot of accounts receivable, with 3.734 billion, in contrast, the annual revenue is also 3.406 billion, and the revenue in 2022 is 3.809 billion.
In other words, almost all of the previous year's revenue became accounts receivable.
Of course, this may be due to the long delivery cycle of the business, so it takes so long for receivables to recognize revenue.
However, what is more noteworthy is the impairment situation.
In the balance sheet of receivables, the largest proportion is from the third party trade receivables, the total number of trade receivables in 23 years is 7.912 billion, and the proportion of impairment provisions accounted for 52.9%, to 4.19 billion.
To put it bluntly, a considerable part of the money cannot be recovered after the project is completed, and in the end, it can only be treated as an impairment of accounts receivable, which is reflected in the income statement, which is beyond the means of making ends meet and continuing to lose money.
According to the financial report, the impairment ratio of SenseTime's trade receivables feels a bit high.
The data shows that the proportion of loss provision for trade receivables from 6 months to 1 year is 27.9%, 43% from 1 year to 2 years, 58% from 2 to 3 years, 75% from 3 to 4 years, and 100% for more than 4 years.
What is the loss provision for trade receivables in 2023? That's 103% of revenue in '23.
After a whole year of hard work, the money earned becomes bad debts, can the company not lose money?
Of course, this can't be blamed on SenseTime, after all, Party A's father doesn't give money, and no matter how hard Party B works, Party A has no right to speak in front of him and still can't get money in the end, so it can only be said that this track is not an industry that is easy to commercialize.
In addition to the impairment of accounts receivable, another core factor affecting profits is cost. According to the financial report, the company's cost of sales will increase by 50% to 1.905 billion in 2023. In addition, other administrative expenses and R&D expenses have decreased.
The rise in the cost of sales is actually understandable, after all, the company is in the transition period of the old and new power, narrowing the business related to smart cities, and then relying on generative large models and automobile-related businesses, the early sales investment is very necessary.
In addition to the cost of sales, another cost that is difficult to avoid is depreciation and amortization.
Companies that do AI business need to purchase a large number of hardware devices, such as AI computing cards, which are essentially consumables and have high depreciation. In other words, with the expansion of the AI business, this part of the depreciation expense must rise.
What is the depreciation and amortization of SenseTime in 23 years? 1.250 billion.
From 2020 to 2022, they will be 570 million, 658 million, and 968 million respectively. In the future, the generative AI business will grow again, and the cost of this part will also be higher, which is actually a rigid expense for companies in the AI industry.
Rigid cost expenditure, coupled with the problem of bad debts, is one of the important reasons why it is difficult for SenseTime to make positive profits. Therefore, if the company wants to continue to operate, it still has to rely on the financing given by investors.
From the perspective of cash flow, from 2021 to 2023, the net cash flow generated by SenseTime's operating activities will be -2.485 billion, -3.084 billion, and -3.234 billion respectively, and the net cash flow from financing activities will be 9.378 billion, 3.329 billion, and 1.084 billion.
With such a large cash outflow, coupled with the fact that the company is not really making money, financing seems to have become the "lifeline" to support the company's operation.
According to the financing information of Tianyancha APP, before going public in Hong Kong in 21, SenseTime also received a cornerstone round investment of 500 million US dollars, and from 2021 to 2023, the net cash flow generated by financing activities has also been decreasing.
Therefore, whether it can continue to get financing is very important for SenseTime.
In terms of business, in fact, SenseTime has also been exploring new possibilities and imagination, at that time, in the pre-IPO prospectus, it had mentioned the "metaverse" 47 times, in March last year, it cooperated with Thailand's T&B Global Media Group to build a metaverse platform, and this year it has made efforts to generate large models.
It can be seen that SenseTime may be eager to make a shift in focus on its business to cater to market hotspots and outlets, while also exploring different directions for the commercialization of its business.
While SenseTime has to do this from a financial standpoint, it could also lose some continuity and stability from a strategic point of view.
At present, AI is still a long-cycle, high-cost investment, and low-commercialization industry, but it needs more strategic determination. For companies that lack the support of other cash cow businesses, this path is obviously more dangerous and the challenges they face will be more demanding.
2. Generative model: SenseTime completes the key blow from "soup boy" to "soup king".
Although SenseTime's overall financial performance is not good, there are also bright spots, such as the generative AI business revenue of 1.2 billion yuan, which is the fastest new business of SenseTime to exceed 1 billion revenue volume since its establishment ten years ago.
In the popular track of generative large models, SenseTime not only has fast technology, but also has very good commercialization capabilities, which is not only the honor of SenseTime, but also retains the face of the "AI Four Little Dragons".
As for running so fast, it is inseparable from the early accumulation of SenseTime:
On the one hand: as early as 2021, SenseTime launched SenseCore, a large device of SenseTime, which is composed of three parts: the computing power layer, the platform layer, and the algorithm layer, and the computing power is even fuller. At that time, the huge amount of capital investment was once criticized, but in the era of large models, computing power is a scarce resource, and the strategic value is highlighted.
Although SenseTime started as a visual technology, it seems that it has little relevance to large models, and it can't stand the computing power enough to play, and SenseTime's technical background is the premise for SenseTime to be able to take the lead in the large model track.
On the other hand, SenseTime has accumulated a large number of high-quality customers in the AI 1.0 era, and these customers in the traditional AI era have rapidly converted into generative AI customers in the past one or two years. Previously, the cooperation between the two became an "endorsement", bringing a siphon effect, and more and more brand customers began to choose SenseTime's large-scale model services.
Technical background + potential customer group is the key to the commercialization of SenseTime's generative large model.
Looking back, let's look at SenseTime's 2023 financial report, although the fundamentals are indeed very stretched, but the long board of generative large models is long enough, and what is more valuable is that strategically SenseTime has changed so many directions over the years, and finally found a direction that it is good at and easy to commercialize.
However, when it comes to continuous commercialization, the challenges faced by SenseTime are like the fundamentals of the 23-year financial report - there are not many surprises, and there are constant shocks.
First of all: the core reason why SenseTime was able to promote commercialization so quickly is that the layout was early enough.
But the big model track is a marathon, not a 100-meter sprint. SenseTime, which was laid out in advance, ran in front, but it could only be regarded as a lead, since it was a marathon, it was also a physical race and an endurance race. It has seized the opportunity period, but it will be erased by time, after all, the giants are now collectively exerting computing power.
In terms of Baidu, it has the largest single intelligent computing center in Asia, the Yangquan Intelligent Computing Center, with a computing power scale of 4EFLOPS (40 billion floating point calculations per second), a construction area of about 120,000 square meters, and the Shenyang Intelligent Computing Center.
In addition, it is very important to note that computing power can be smashed out with money.
Secondly, and more importantly, the big model is a thief and burns money, how much money can it burn? ChatGPT has spent nearly $200 million in the nine months since its birth! Some media have even threatened that OpenAI is likely to go bankrupt by 2024. It's a joke now, but the essence behind the joke is that large models do burn money.
Behind OpenAI is Microsoft, and behind Wenxin Yiyan is Baidu. The core of Baidu's repeated failures in the field of AI is that it is not bad for money, and the search business can provide a steady stream of cash support.
Living long enough means a better chance of lasting until the day when the willows are shining, no, there are rumors that Baidu will recently provide AI functions for Apple's iPhone 16, Mac system and iOS 18 released this year.
And how long can SenseTime last?
This is a question worth discussing. From a purely financial perspective, the answers to these questions are not difficult to answer.
SenseTime's total operating expenses in 2023 will be 5.864 billion, net current liabilities will be 13.653 billion, cash and equivalents in hand will be 10.733 billion, financial assets of 500 million and receivables of 37.34 billion. In addition, the total net cash flow generated by the company's operation, investment and financing in 23 years was 1.532 billion.
From the perspective of cash covering operating expenses and current liabilities, the company currently has enough money to operate.
However, if we consider the annual loss of about 6 billion yuan in the past three years, if the scale of the loss is still like this, then the company's current capital reserves may be difficult to support the long-term operation of more than 5 years.
If we also consider that the financing prospects of the entire AI industry in the secondary market are not optimistic, I am afraid that it will still need to borrow from banks to maintain the company's operation.
At present, SenseTime has actually borrowed a lot of bank loans, and the financial report shows that SenseTime's current secured bank loans are 4.469 billion, and the borrowing interest rate is 1.95%-3.80%. The interest rate is not high. In addition, the company has not yet drawn 8.26 billion yuan of bank credit.
In other words, when it is really a last resort, there is still room for reversal in business. It's just that it's still that problem, the company's operation can't always rely on taking money from shareholders or finding bank loans, after all, it still needs to find its own hematopoietic ability.
At present, whether it can bring this hematopoietic ability may depend on whether the generative large model business can be supported.
But the question is, the big model itself burns money like this, even if it grows to a business scale of 1 billion now, how much cash flow and profit increment can it bring in the future?
I'm afraid it's hard to say.
From the perspective of commercialization, the prospect of the landing of generative large models in the automotive industry is not clear, but it is a good signal that SenseTime helps Xiaomi SU7.
In terms of external capital, Microsoft, Ali and other shareholders have successively reduced their holdings, according to the previous blunt knife to cut meat, visually, it is really difficult to say when this wave of reduction will be completely cleared.
In terms of stock price, SenseTime is currently hovering around HK$0.7, and its market value cannot go up, and its future financing ability and liquidity may be overshadowed.
This is a race against time, running through the race of opportunities, entering the physical race, before the endurance race, how SenseTime can beat the time, how to find better investors, how to speed up the commercialization process, is the key battle whether SenseTime can rely on generative large models to turn around.
Finally, I wish Shang Tang to turn the tables against the wind as soon as possible and return to the spirit of the "soup king". Here, I also pay tribute to Tang Xiaoou, and hope that SenseTime people can carry forward Tang Xiaoou's unfinished ambitions.
"When the story is staged to the end, we will also be like the end of "The Long Season", sitting on a small train in the bud rice field, looking forward, don't look back, and drive forward on the road of AI!"