In recent years, if you were to ask which technology track is the hottest, AI would undoubtedly take the lead, having set off wave after wave of global trends.
NVIDIA has played the role of the AI king in a "solo performance," with its market value once soaring to the top in the world. The world's top technology investor, James Anderson, even predicted that NVIDIA's market value may approach 50 trillion US dollars in the next ten years. Chip giants such as AMD and Intel have also taken advantage of the AI trend, with a strong momentum to share the global computing power market.
However, in contrast, a number of AI startups have fallen into a predicament of "burning money" without making a profit. These small and medium-sized enterprises have not felt the gifts from the era, but are facing multiple crises under the Matthew effect of the AI industry.
Especially when the enthusiasm of investors fades, people are shocked to find that star AI companies are also on the verge of bankruptcy, such as Perplexity starting to sell advertisements, MistralAI quietly deleting the open-source mission, Anthropic urgently solving a financial gap of up to 1.8 billion US dollars a year, and star companies like Inflection AI and Stability AI also facing commercialization problems, with core members frequently jumping ship or resigning.
Advertisement
At the same time, including well-known AI chip manufacturers like Graphcore and Gro Intelligence, as well as many domestic and foreign AI chip companies, are facing financial difficulties and bankruptcy predicaments.
Report data shows that in the first quarter of this year, the seed round (and earlier) financing of American AI startups plummeted to 123 million US dollars, a year-on-year decline of 70%, almost returning to the level before the emergence of ChatGPT in 2022. In addition, the number of financing transactions of these American AI startups in the first quarter also decreased by more than half compared to the previous quarter, from 74 to 34.
The "cooling down" of China's AI industry financing heat is also very obvious. According to the "2023 Artificial Intelligence (AI) Industry Status Report" released by the research institution CB Insights, in 2023, the number of financing and investment in China's AI field is about 232, a year-on-year decrease of 38%; the total financing amount is about 2 billion US dollars, a year-on-year decrease of 70%.
It is not difficult to see that under the dazzling halo of AI giants, financing obstacles, mergers and acquisitions, and bankruptcies have become dark clouds hanging over AI startups. Recently, there has been continuous news of mergers and acquisitions in the AI industry, and many companies that were able to raise funds a year ago are now facing a shortage of funds, so they have turned to merger and acquisition transactions as an exit route. At the same time, well-funded technology companies are also using this market environment to strategically acquire high-quality assets.
Behind these actions, on the one hand, there is the helpless fate of small enterprises being sold, and on the other hand, it seems to announce the new attempts and ambitions of the large enterprises that initiate the acquisition.Behind the AI M&A Wave, a New Round of "Offensive and Defensive Battles" Unfolds
AMD Acquires Silo AI to Take on Nvidia Again
Recently, AMD acquired the European artificial intelligence startup Silo AI for $665 million, sparking a new wave of challenges against Nvidia.
As we all know, AMD has long aimed to compete with Nvidia in the field of AI chip technology. On the day the acquisition news was released, AMD's stock price rose by 4.7%.
It is worth mentioning that Silo AI's operating income in 2023 was 15 million euros, while the transaction price for this acquisition reached as high as $665 million.
From the generosity of the bid, it is not difficult to see AMD's AI ambitions.
Silo AI is known as the largest privately-owned AI laboratory in Europe, with 300 AI scientists and engineers. The leadership team is quite luxurious, co-founded in 2017 by Peter Sarlin, a practicing professor at Aalto University in Finland; Tero Ojanperä, former CTO of Nokia; Johan Kronberg, former CEO and Chairman of PwC; Ville Hulkko, co-founder of the Valossa Labs video AI platform; machine learning expert Kaj-Mikael Björk; and Juha Hulkko, founder of Elektrobit.
However, unlike most competitors, Silo AI does not develop language models using Nvidia's chips but uses AMD's chips, which also implies that AMD chips have strong competitiveness.
It is understood that Silo AI has business in both Europe and North America, with its global headquarters located in Helsinki, Finland, and its North American headquarters in Canada, focusing on providing customized AI models and end-to-end AI-driven solutions to help customers quickly integrate AI capabilities into their products, services, and operations.
In addition, Silo AI has also created open-source multilingual large models on the AMD platform, such as the Poro, Viking, and SiloGen model platforms. Previously, Silo AI also officially announced a partnership with the large model developer Mistral AI, combining Silo AI's end-to-end AI capabilities with Mistral AI's large models.Silo AI's clientele includes major companies such as Allianz, Philips, Rolls-Royce, and Unilever, and it is reported that they have delivered over 200 production-level AI projects to date.
After being acquired by AMD, Silo AI CEO Peter Sarlin will continue to lead the Silo AI team as part of AMD's AI division, reporting to AMD Senior Vice President Vamsi Boppana.
Many in the industry have expressed that this acquisition is intended to continue to compete with NVIDIA.
Looking at the recent dynamics, competitors like AMD and Intel have been investing heavily in hardware and in securing more market share in AI solution offerings.
For instance, in terms of hardware, AMD's MI300X is considered to be capable of rivaling NVIDIA's H100 in certain tasks. In addition, AMD has promised to launch the MI325 with 288GB HBM3E memory this year, and the MI350 next year, which will compete with NVIDIA's Blackwell B200.
However, in terms of software, despite AMD and Intel's attempts to compete with NVIDIA in software, they are still far behind. Industry experts have indicated that even if AMD chips can match NVIDIA in key technical data, NVIDIA's software optimization will give them a significant efficiency advantage (software can provide a 30-fold speed advantage).
This acquisition by AMD is aimed at filling the gap in software capabilities. It is known that Silo AI trains its language models on the Finnish supercomputer Lumi and AMD hardware, which is achieved through a specially developed software layer.
AMD CEO Dr. Lisa Su also emphasized: "The acquisition of Silo AI significantly strengthens our artificial intelligence software capabilities."
Experts from Silo AI will further develop the software in the future, which customers can use to program complex AI models and train them on computers equipped with AMD chips. This specialized technology in the field of AI software is very valuable to AMD and its customers.
In fact, to close the gap, AMD has been investing and acquiring heavily in the past period. In the past year, AMD has invested over $125 million in more than a dozen AI companies to enhance the leadership of AMD's computing platform. Among them, in August 2023, AMD announced the acquisition of the French AI software company Mipsology, and in October, it acquired the American AI startup Nod.ai.This time, it has made a one-time extravagant investment of 665 million US dollars to acquire Silo AI. AMD Senior Vice President Vamsi Boppana stated, "This transaction will not have a practical impact on AMD's financial performance, but it has the potential to unleash tremendous commercial potential in the future."
It is evident that AMD is continuously expanding its AI ecosystem and advancing its leadership position in computing platforms.
SoftBank Takes Over the "British Version of Nvidia"
On the other side, the news of SoftBank's acquisition of the AI star company Graphcore has dropped a bombshell in the industry.
After widespread speculation and anticipation in the market, the Japanese SoftBank Group finally announced the acquisition of the British artificial intelligence chip company Graphcore, adding a giant to the AI chip field and accelerating the prosperity of the global AI ecosystem.
The development history of Graphcore is full of challenges and opportunities.
As a startup company focusing on the development of AI chips, Graphcore has stood out in the fiercely competitive AI chip market with its unique IPU technology. Compared to traditional GPUs, IPUs are more focused on AI computing tasks in design, providing more efficient and flexible solutions to meet the growing demand for AI applications.
In the past few years, Graphcore has won widespread recognition in the industry with its technological innovation and excellent product performance, attracting the favor of many well-known companies and investment institutions. Its valuation has quickly risen to 2.8 billion US dollars, becoming one of the most promising startups in the UK. It was once regarded as a treasure for competing with Nvidia and was even called the "British version of Nvidia."
In the face of various market speculations and rumors, Graphcore has always maintained a low-key and cautious attitude, focusing on technology research and development and product iteration.
However, the AI hardware industry is fiercely competitive, and the demand for funds is huge. Graphcore has failed to grow into a European AI chip giant as expected, and its performance has been under pressure in recent years. The company's co-founder and CEO, Nigel Toon, said that the scale of investment needed in the AI field is "astonishingly huge," far beyond the scope that Graphcore can bear as an independent company.In February of this year, it was revealed that Graphcore was in trouble. Graphcore's revenue in 2023 plummeted by 46%, and its losses expanded further. In October 2022, Graphcore attempted to raise funds from investors to supplement operations and cover losses, but the negotiations ultimately failed.
Moreover, in November 2023, Graphcore announced its withdrawal from the Chinese market, significantly reducing its business in China, which undoubtedly cast a shadow over the company's prospects.
In February of last year, Graphcore actively sought investors outside the UK and prepared to sell a large number of shares in exchange for urgently needed capital. Rumored potential transaction counterparts included companies such as OpenAI, Japan's SoftBank Group, and Arm, with Graphcore planning to raise more than $500 million.
It was not until recently, with the news of the acquisition by SoftBank, that the speculation surrounding Graphcore came to an end.
For SoftBank, the acquisition of Graphcore is undoubtedly an important step in accelerating its AI ecosystem layout. Since the controversial acquisition of Arm in 2016, SoftBank has clarified its AI strategic direction. As a global leader in semiconductor IP supply, Arm's technology is widely used in consumer electronics such as smartphones and tablets, providing a solid foundation for SoftBank's AI layout and directly obtaining ownership of intellectual property rights surrounding semiconductor design.
In recent years, SoftBank has been continuously increasing its investment and layout in the AI field. From announcing the development of high-performance semiconductor infrastructure and improving computing power to establishing a special fund for AI company investment, every step taken by SoftBank reveals its firm belief and ambition for the future of AI.
Especially this year, SoftBank not only increased its investment in AI technology research and development but also promoted the establishment of a new department at Arm focused on the development and production of AI chips, further strengthening its competitiveness in the AI chip field.
In April, it was reported that SoftBank was preparing to develop computing devices required for generative AI, with an investment of 150 billion yen (about $900 million) by 2025 to develop high-performance semiconductor infrastructure, aiming to increase computing power by several tens of times the current level.
Subsequently, SoftBank announced its intention to invest $9 billion annually to increase investment in artificial intelligence companies.
In May of this year, reports revealed that Arm would establish a new department to develop AI chips, with the goal of manufacturing a prototype before the spring of 2025 and starting mass production in the fall of that year.In June, Masayoshi Son declared at the shareholders' meeting that he would spare no effort to create a "super artificial intelligence era," stating that this move might succeed or might fail miserably, but SoftBank had no choice but to try. This bold statement has also set a clear goal for SoftBank's AI grand plan.
Industry investors and analysts believe that through this acquisition, SoftBank is expected to further expand its influence in the semiconductor landscape and achieve synergistic effects with Graphcore. On one hand, this will help SoftBank achieve faster growth and broader layout in the AI field. On the other hand, with the support of SoftBank, Graphcore will have more technological innovation and market opportunities, promoting its continued leadership in the AI chip field.
The capital market has also shown affirmation to SoftBank's "AI dream." Since February of this year, SoftBank's stock price has risen by nearly 70%.
At the same time, this transaction will also provide beneficial references and insights for other enterprises and investors, promoting competition and development in the AI chip market, and driving the further prosperity of the global AI ecosystem.
As Dan Ridsdale, the head of technology at Edison Group, said: "NVIDIA has already taken a dominant position in the generative artificial intelligence field, but there are still other opportunities in the field of artificial intelligence, and the industry needs strong competitors. This acquisition is good news for AI and both companies."
Apple, the king of AI acquisitions?
Data statistics show that as of 2023, Apple has acquired as many as 32 AI startups.
After Apple was exposed to cut the automotive project team, Apple CEO Tim Cook said at the company's annual shareholders' meeting on February 28 that Apple plans to disclose more plans related to the use of generative artificial intelligence later this year.
This is also the strongest signal to date that the iPhone maker is embracing the generative artificial intelligence boom.
Because Apple has always maintained a low-key investment and merger and acquisition strategy and usually does not disclose the details of its transactions, some clues can still be captured through the business and direction of the investment targets.According to data from IT Juzi, as early as 2014, Apple announced the acquisition of Novauris, a UK-based voice recognition technology company. The Novauris team has been working at Apple since the fall of 2013, focusing on improving the Siri virtual assistant service based on Siri speech technology.
A year later, to continue optimizing Siri services, Apple acquired VocalIQ, a UK AI voice company focused on the automotive market.
In 2017, Apple acquired RealFace, an Israeli facial recognition technology company, which mainly provides biometric login services and can also help users on different platforms select and organize the best photos in their photo albums.
This year, Apple has announced the acquisition of several artificial intelligence companies, including French AI startup Datakalab, Canadian AI company DarwinAI, and German AI startup Brighter AI.
Datakalab focuses on low-power, high-efficiency deep learning algorithms, developing extremely advanced visual recognition and information processing technologies, and claims to provide a "new method for compressing computer vision neural networks" that can run without relying on the cloud, thereby enriching its local AI capabilities. In the future, it may provide technical support for Apple's iPhone, iPad, and Vision Pro to achieve better AI functions.
Datakalab has a lot of development in the commercial field, especially in developing technology that can analyze human emotions through facial recognition and visual data analysis.
DarwinAI, founded in 2017, is a rapidly developing visual quality inspection enterprise that provides AI visual quality inspection systems for electronic manufacturers to improve product quality and production efficiency. It is reported that the technology features making AI models more lightweight and efficient, and this research result may provide strong technical support for Apple to introduce generative AI in iOS18 this year.
DarwinAI's patent for explainable artificial intelligence (Explainable AI, referred to as "XAI") platform has been adopted by many Fortune 500 companies, which can easily integrate AI that they can trust.
Apple's acquisition of DarwinAI is an important step in its active layout in the AI field. After the acquisition, the core team members of DarwinAI joined Apple, including the company's co-founder and chief operating officer, Arif Virani, and chief scientist Alexander Wong. Industry experts believe that DarwinAI's technology is expected to be applied to Apple's supply chain management, further improving the efficiency of component manufacturing, thereby maintaining a leading position in the global AI market competition.
The other German AI startup, Brighter AI, acquired by Apple, mainly engages in anonymization/privacy data processing. This technology may be helpful to Apple. With the help of AI technology, Apple can erase privacy data without blurring photos. The purpose is to use this acquisition to strengthen the privacy features in products such as Apple Vision Pro and Apple Maps.According to an analysis by IT Orange, Apple Inc.'s merger and acquisition (M&A) strategy is characterized by several key features:
1) The scale of M&A is usually not large, focusing mainly on small technology companies;
2) The primary purpose of M&A is to acquire talent, technology, and intellectual property, with products and commercialization as secondary considerations. This may also be an important factor in Apple's long-term maintenance of innovation;
3) M&A is highly related to the main business, starting from the overall business and strategy, to improve existing businesses or develop new ones.
Overall, Apple's external M&A activities are very low-key, consistently not wanting to talk about investments, and focusing on developing its own products. These characteristics are not only reflected in the field of AI. It is understood that in the past decade, Apple has shown more than 82% of its investment behavior in the form of M&A.
Although Apple is "tight-lipped" about its internal AI strategic layout, in the ongoing AI arms race, there are reports that Apple is discussing and trading with more AI startups to gain an advantageous position in future development. By acquiring promising AI startups, Apple has gained top talent and core innovative technologies, consolidating its position in key AI fields and ensuring a competitive edge in the rapidly changing technological environment.
It is reported that Apple has started testing large language models internally and integrating generative AI into features such as Siri, Messages, and Apple Music. Apple plans to add functions for automatically creating presentations and completing text blocks to its software. There are also reports that Apple is developing a new version of the Xcode programming software, which uses AI technology to help developers write code.
It is also worth noting that Apple, with its billions of smart terminal devices worldwide, has established a solid ecosystem of developers and users. This means that once new AI features are integrated into iOS, they will quickly cover billions of devices and affect hundreds of millions of users. This market access advantage is something other companies do not possess.
Now, as its AI layout deepens step by step, Apple, which has always been "half a beat behind" in AI competition and has always worried investors, may open up a new world in the near future.
AI M&A continues to ferment.In addition, the M&A in the AI industry continues to ferment.
Recently, OpenAI announced two acquisitions in just four days: the real-time analytics database startup Rockset and the remote collaboration company Multi;
There is the latest news that the AI unicorn Character.AI is facing financing difficulties and is considering being acquired, with plans to sell to Google and Meta, etc.
In addition:
Big data giant Databricks acquired MosaicML for $1.3 billion: MosaicML will become part of Databricks' Lakehouse platform, providing enterprises with a unified platform for managing data assets.
The world's largest professional information service provider Thomson Reuters acquired Casetext: Thomson announced that it will acquire the AI assistant startup Casetext for legal professionals for $650 million, as part of its long-term investment in generative AI.
Business intelligence software provider ThoughtSpot acquired Mode Analytics for $200 million: The main business of both companies is to provide data analysis services for enterprise users.
HubSpot acquired Clearbit for $150 million: Marketing software manufacturer and customer relationship management platform HubSpot announced the acquisition of B2B data provider Clearbit to strengthen its AI platform.
Homestay giant Airbnb acquired GamePlanner.AI: Airbnb acquired GamePlanner.AI for $200 million, which focuses on developing artificial intelligence and machine learning technologies to improve corporate operational efficiency and service quality. The technology and talent of GamePlanner.AI will mainly be used to enhance Airbnb's capabilities in the field of intelligent accommodation services.
Cloud computing giant Snowflake announced the acquisition of Neeva, a generative AI search startup founded by two former Google employees.The following article has been translated into English:
There are too many examples to list.
Looking at the domestic AI market, there have been instances such as Guangnian handing over to Meituan and Lingxin Intelligence being acquired by Zhi Pu AI.
It can be seen that the trend of reshuffling and integrating resources in the AI industry has always been present, but foreign manufacturers have recently had a concentrated outbreak, while the domestic market may still have uncertainties due to the recent API price war.
However, some voices have indicated that the head effect of the current domestic large model market has initially emerged, and the characteristics of the first echelon are very obvious: they have unicorn-level valuations, have their own large models, and application products have begun to land. For other players not in the first echelon, they are prone to face difficulties in aspects such as funding and profitability.
Mergers and acquisitions are very important for both parties. The "seller" can avoid the fate of the company going bankrupt, which may also be one of the destinations that AI startups are looking forward to; on the other hand, for the "acquirer," it is also an extremely important supplement. If they can find a highly potential startup and incorporate it into their own, they can often shine in key places.
Especially in the context of the cold financing market, the trend of mergers and acquisitions in the AI industry may accelerate. Because for any startup, they always have to face the issue of whether the operating cash flow can cover the expenses.
Regarding the account book, there may be a miracle that works for a while, but there is no magic that can be applied for a long time.
As the AI storm sweeps the world, global technology giants have fully embraced AI, setting off a race against time. A little faster, a little faster, in addition to increasing R&D, cash-rich giants often choose to directly buy a startup to quickly make up for their shortcomings.
Everyone is eyeing each other.NVIDIA, Continuously Strengthening Its Moat
Upon review, it is evident that the investment and acquisition cases in the AI sector by giants are on the rise. As a leader in the AI race, NVIDIA is also actively strategizing its position.
Data indicates that NVIDIA invested in approximately over 30 startups in 2023 alone, doubling the number from the previous year, participating in a total of 38 financing rounds with a total value exceeding $5 billion.
Looking at the major AI financing deals, NVIDIA's presence is almost always noticeable.
Jensen Huang has repeatedly emphasized, "The technology industry develops rapidly; if NVIDIA does not invest in the distant future early on, it may face catastrophic consequences." Therefore, NVIDIA's investments cover almost the entire AI industry chain.
Following the investment in over 30 startups last year, NVIDIA recently invested in/acquired two AI startups, Run:ai and Deci, indicating an intensifying trend in AI mergers and acquisitions.
It is reported that Run:ai, founded in 2018, is a provider of workload management and orchestration software based on the open-source container orchestration platform Kubernetes. By acquiring Run:ai, NVIDIA can further simplify the deployment process of AI solutions, improve GPU utilization, and enhance GPU infrastructure management, significantly reducing costs.
Currently, Run:ai's solutions have been integrated with NVIDIA's DGX, DGX SuperPOD, Base Command, NGC containers, and AI Enterprise software, allowing customers of NVIDIA DGX and DGX Cloud to use Run:ai's AI workload capabilities.
Another AI startup acquired by NVIDIA, Deci, is also committed to achieving "cost reduction and efficiency improvement" for AI chips.
It is understood that Deci, founded in 2019, is an Israeli deep learning service provider. Its core technology lies in accelerating the inference process of AI models through means such as data preprocessing, model architecture, and hyperparameter optimization. In addition, Deci is also responsible for model deployment, services, monitoring, and interpretability, providing customers with comprehensive AI solutions. Its accelerator can redesign models, creating new models with multiple computational paths, all of which are optimized for specific inference devices, thereby greatly improving the model's operational efficiency.At present, Deci has reached cooperation with multiple tech giants such as Microsoft, Intel, AMD, and Amazon.
It is not difficult to see that the purpose of these two transactions by Nvidia is to help its customers make more effective use of their own AI chip products.
Recently, Nvidia made another move, acquiring the American software startup Shoreline for about 100 million US dollars. This is another major initiative after Nvidia's market value surpassed Microsoft and Apple to become the most valuable listed company in the world. Lingxi Investment Research believes that this acquisition will further enhance Nvidia's strength in the AI field, aiming to reduce dependence on cloud computing giants like Amazon and accelerate the pace of building its AI empire. At the same time, Shoreline.io will also leverage Nvidia's strong R&D capabilities and global resources to accelerate the iteration and upgrade of its products, providing users with more efficient and stable services.
In addition, Nvidia has also made a series of other investments in fields such as hardware, software, data center management platforms, robotics technology, security analysis, and mobile capabilities.
Overall, through continuous product iteration, technological innovation, and acquisition strategies, Nvidia has successfully built an ecosystem centered on AI. Its stock price has soared by more than 200% in the past year, with a market value once exceeding 3.3 trillion US dollars, making it one of the most valuable listed companies in the world. Market analysts are optimistic about Nvidia's future, believing that its market value is expected to continue to rise.
In conclusion, the emergence of this round of "acquisition wave" reflects the multiple challenges faced by the AI industry.
High R&D costs, uncertainty in business models, and fierce competition with tech giants have put even AI companies with valuations over 1 billion US dollars under tremendous pressure. Shortage of funds, poor market response to products, and pressure from investors, various factors are intertwined, forcing these companies to re-examine their development strategies.
Many founders and investors have realized that some early companies are only interesting experiments, but most startups have low retention rates and profitability levels, and are still unable to escape the fate of bloodsucking.
In this situation, with a large number of entrepreneurial projects pouring in, the elimination competition has officially begun.In addition to the aforementioned M&A cases, many AI star companies that were once in the limelight have begun to fall into difficulties. On the subpage AI Graveyard of the AI tool aggregation website "DANG!", as of June this year, it has listed as many as 738 failed AI projects.
The time left for small and medium-sized AI startups may not be much.
For AI startups, pursuing technological innovation is a means of breaking through, and high-quality products and sustainable profit models are the core of winning.
From the perspective of investors and large technology companies, acquiring AI startups to obtain talent, technology, and resources has become an important feature of this round of reshuffling. However, how to obtain truly scarce AI capabilities in an environment of intense competition and increasing regulation is still an important issue that needs to be considered in the next stage.
Comments