Key Takeaways:
SenseTime is progressing in its shift to generative AI, which made up 60% of revenue in the first half of this year, even as it faces challenges from price wars and GPU restrictions
Despite the company’s pivot away from more sensitive vision AI technologies, it continued to lose money in the first half, though its loss narrowed
The release of OpenAI’s ChatGPT in late 2022 sparked a global revolution in artificial intelligence (AI). Overnight, the term “AI” became synonymous with the generative form of the technology, representing a sudden shift away from the previous focus on computer vision, an area where China had been leading.
Something went wrong.
We’ll take a deeper dive into SenseTime’s generative AI push shortly, including what’s driving it and the challenges it could face in the transformation. But first we’ll review the company’s history and finances from the past few years, which have led to today. SenseTime , once hailed as one of China’s four “AI Dragons,” found itself scrambling to adapt. The company swiftly decided to go all-in on generative AI, joining the earliest group of Chinese companies investing in the technology.
SenseTime initially focused on visual recognition AI after its founding in 2014, hitching its wagon to China’s aggressive spending on surveillance systems as part of efforts to enhance public security. The company provided advanced software, including facial recognition algorithms, for integration into digital cameras and other systems. Financial results, announced last week, show signs of progress in the transformation. For the first half of this year, the company’s revenue from generative AI grew sharply to account for 60.4% of its total, as the segment’s business surged.
This trend was particularly pronounced between 2019 and 2021, when the pandemic drove increased demand for such systems as China used the technology in its efforts to limit the spread of the virus. As it rode that government-led spending wave, SenseTime’s annual revenue reached a peak of 4.7 billion yuan ($663 million) in 2021, more than double the 1.85 billion yuan in 2018.
SenseTime was never profitable to begin with, and its losses peaked in 2021 at a staggering 17 billion yuan. But as it faced the dual challenges of slowing demand and U.S. sanctions, it began focusing on controlling costs and narrowing its losses to ensure its longer-term viability.
In its latest financial report, SenseTime put a positive face on the ongoing decline for its “traditional AI” segment, mainly computer vision products, by saying it reflected a successful transition to generative AI. The traditional segment accounted for about 30% of its revenue in the first six months of this year, less than half the 73.5% it made up a year earlier when products from that segment fell under its Smart Business, Smart City, and Smart Life categories.
The company’s generative AI business moved in the other direction, surging 255.7% to 1.05 billion yuan, accounting for 60.4% of its revenue. That helped to lift the company’s total revenue for the six-month period by 21.4% to 1.7 billion yuan, putting SenseTime on track to return to annual revenue growth after two years of declines.
SenseTime was never profitable to begin with, and its losses peaked in 2021 at a staggering 17 billion yuan. But as it faced the dual challenges of slowing demand and U.S. sanctions, it began focusing on controlling costs and narrowing its losses to ensure its longer-term viability.
The second area involves development of its own large language model (LLM), SenseNova. First unveiled in April 2023, it rolled out its fifth version of the LLM in April.
Building on SenseNova, SenseTime has created a ChatGPT-like chatbot service called SenseChat. While SenseChat is currently free for public use, SenseTime monetizes the technology by selling application programming interface (API) services via its cloud platform. Such services allow enterprises and developers to leverage SenseNova to create AI applications tailored to their specific needs.
Two Focuses
However, investors shouldn’t get too excited just yet about such numbers, since the generative AI market in China is still in its early stages and such rankings could change dramatically in the future. Nvidia, then hosting them in data centers and renting their computing power with its own software to enterprises via cloud platforms. The
business model is straightforward: acquire GPUs, set up infrastructure, and lease computing power to businesses needing AI resources.
One dynamic to consider is the fierce price war unfolding in China’s LLM space, where developers are racing to lower prices to attract customers. In such a competitive environment, startups like SenseTime may have difficulty competing with giants like Baidu and ByteDance, which have much deeper war chests.
In GPU rental services, SenseTime, like other firms in this business, is being constantly challenged by tightening restrictions from the U.S. government. Such restrictions have already led to bans on the sale of some advanced Nvidia GPUs to China.
Despite the challenges, there are also some positive signs. SenseTime’s shift and investment into generative AI doesn’t appear to be worsening its financial losses. For the first half of 2024, the company reported a net loss of 2.5 billion yuan, narrowing from 3.1 billion a year earlier. This improvement came despite a 6% increase in R&D costs and a 24% rise in the cost of sales, which includes expenses for purchasing GPUs.
SenseTime’s progress in generative AI reignited investor interest in its stock earlier this year, powering a 40% surge in the shares the day it released the latest version of SenseNova in April.
However, investor reaction to its latest report last week suggests the momentum may be waning, with the shares edging down 1.6% the day after the report’s release. Its closing price of HK$1.18 on Friday is more than double its HK$0.60 low earlier this year but still represents just half the HK$3.85 IPO price when it listed in December 2021.
For SenseTime to win back investors, it will need to prove it has a stable footing and can keep growing in the rapidly evolving generative AI market, where it will have to weather financial challenges in a race that’s almost certain to end in a painful consolidation. Simultaneously, it will need to navigate carefully, managing a controlled decline in its traditional computer vision business while shifting more resources to its generative AI segment.
Despite the challenges, there are also some positive signs. SenseTime’s shift and investment into generative AI doesn’t appear to be worsening its financial losses. For
the first half of 2024, the company reported a net loss of 2.5 billion yuan, narrowing from 3.1 billion a year earlier. This improvement came despite a 6% increase in R&D costs and a 24% rise in the cost of sales, which includes expenses for purchasing GPUs.
SenseTime’s progress in generative AI reignited investor interest in its stock earlier this year, powering a 40% surge in the shares the day it released the latest version of SenseNova in April.
However, investor reaction to its latest report last week suggests the momentum may be waning, with the shares edging down 1.6% the day after the report’s release. Its closing price of HK$1.18 on Friday is more than double its HK$0.60 low earlier this year but still represents just half the HK$3.85 IPO price when it listed in December 2021.
For SenseTime to win back investors, it will need to prove it has a stable footing and can keep growing in the rapidly evolving generative AI market, where it will have to weather financial challenges in a race that’s almost certain to end in a painful consolidation. Simultaneously, it will need to navigate carefully, managing a controlled decline in its traditional computer vision business while shifting more resources to its generative AI segment.
