TINO FERNÁNDEZ
At the end of 2024, artificial intelligence was decisively driving venture capital activity. EY concluded in a report that AI-related companies had raised 37% of total venture capital activity in the last quarter of 2024.
In addition, AI unicorn births doubled, and generative AI continued to be key for new unicorns. More than half of the artificial intelligence unicorns born in the third quarter of 2024 were generative AI start-ups.
EY predicted in The Top 10 Opportunities for Technology Companies in 2025 that "after the initial hype, artificial intelligence is now facing a period of consolidation, in which it will be a key element in the transformation and evolution of practically all industries.
CB Insights is now launching its annual edition of the AI 100, a ranking of the world's leading AI start-ups that only features one Spanish start-up: Multiverse Computing, which focuses on developing software to take advantage of the most advanced devices available and deliver short-term value to businesses.
The CB Insights ranking is a prediction that highlights the next wave of winners in the AI business and identifies players in this early stage of AI that show strength in terms of market traction, investor quality, and talent.
Global 'Top 10'
If the data from the AI 100 classification is cross-referenced with 20 other international sources including, in addition to CB Insights, Crunchbase, Pitchbook Data, Bloomberg Technology, MIT Technology Review, GitHub Repository Analytics or Quantum Computing Report, it is possible to extract a top 10 of the most promising AI start-ups in the world, taking into account the financing and valuation of the companies. the rate of growth and traction in the market, the demonstrable technological innovation, the founding team, the viable business applications, the market potential, and the scalability.
FIGURE. The first start-up to be included in this selection is the American Figure, founded by Brett Adcock, who also founded Archer Aviation and has recruited senior engineers from Boston Dynamics, Tesla and Apple. Adcock is recognized for its entrepreneurial track record in technology start-ups and for its vision to integrate artificial intelligence with robotics in a practical and scalable way. His leadership and experience in raising capital from major investors have been decisive for the rapid growth and international projection of Figure, which stands out for having recently raised €595 million in a Series B round led by Microsoft, Nvidia and Jeff Bezos, and which has reached a valuation of €2,290 million in less than three years.
Bezos is interested in Figure, as the company develops technologies that complement strategic areas for him, such as advanced automation, logistics, and human-machine interaction. Figure's expertise in humanoid robots can bring key capabilities to improve operational efficiency in warehouses, distribution centers, or even consumer services, areas where Bezos has historically invested and seeks to continue innovating. In addition, supporting disruptive technologies in robotics fits with their long-term vision to lead the next technological era.
HELSING. The German company has positioned itself as a leader in AI applied to defense in Europe, securing €209 million in funding, including €103 million from Spotify creator Daniel Ek.
The start up was founded by Torsten Reil and Gundbert Scherf, who have extensive experience in the defence sector and advanced technology. Torsten Reil has an outstanding track record in the management and development of technologies for national security, having worked on companies and projects related to defence and autonomous systems. Gundbert Scherf is an expert in applied artificial intelligence, with experience in research and development of AI solutions for complex and high-security environments.
Helsing's AI and sensor fusion platform enables data from multiple sources to be interpreted in real-time, creating smarter, more adaptive defense systems. It has active contracts with the defence ministries of Germany, France and the United Kingdom to modernise defence systems, something that in the current geopolitical context in Europe has accelerated its growth, with governments seeking technological autonomy in defence. Its team includes former high-ranking military officers and AI experts from DeepMind. And it is also one of the few European start-ups with authorisation to work on classified defence projects.
Unlike some start-ups that develop isolated solutions, Helsing focuses on integrating its AI technologies with existing defense infrastructures and systems, facilitating adoption and modernization without the need for costly replacements or lengthy implementation processes.
LANGCHAIN. This American start-up is used by companies such as JPMorgan, Klarna or Replit to implement AI applications on an enterprise scale. Its API allows you to build custom natural language processing chains, integrating different models, databases, and knowledge sources; and its AI agent architecture allows for the creation of AI systems that can reason step-by-step and use external tools. The founder of this American company is Harrison Chase, with a great technical experience to democratize the access and use of large language models. Chase has driven the creation of an active community of developers and positioned LangChain as a standard for building and scaling modern natural language processing solutions, making the technology accessible to enterprises, researchers, and developers alike.
WAYVE. Experts believe Britain's Wayve has revolutionized the approach to autonomous driving with its learning-from-scratch technology, which contrasts with competitors that rely on detailed high-definition maps and manually programmed rules. The company has raised €230 million from Microsoft, Virgin and Baillie Gifford, and its technology applies end-to-end deep learning that allows vehicles to adapt to new environments without prior specific programming, demonstrating the ability to generalize. In addition, Microsoft has signed a $1 billion contract to integrate its technology into autonomous transportation initiatives.
What really sets Wayve apart in this highly competitive industry is its "learning first, sensors second" approach. While other companies such as Cruise or Aurora invest enormous resources in equipping their vehicles with ultra-advanced sensor technology, Wayve has opted to develop more sophisticated learning algorithms that can work with simpler hardware, which would facilitate a wider and cheaper adoption of this technology.
Wayve uses deep learning techniques to teach its vehicles to drive through direct experience in complex urban environments, rather than relying solely on predefined maps and expensive sensors. This gives them an advantage in adaptability and scalability.
TOGETHER AI. Fundada por Vipul Ved Prakash, junto con antiguos investigadores de Google, Meta y Stanford, ofrece una plataforma cloud para entrenar e implementar modelos de IA a escala que ha conseguido atraer a más de 15.000 desarrolladores y empresas, ofreciendo una alternativa viable a soluciones propietarias como OpenAI. En febrero, la compañía recaudó 305 millones de dólares en una operación liderada por General Catalyst, que elevó la valoración de la empresa a 3.300 millones de dólares. Prakash tiene una trayectoria destacada, y es cofundador de Cloudmark, especializada en seguridad digital, que fue adquirida por Proofpoint.
Together AI represents a commitment to the democratization of advanced AI technology, allowing more organizations to harness the potential of large language models and other AI technologies without facing prohibitive traditional economic and technical barriers.
HIPPOCRATIC AI. The startup was founded in 2023 by Munjal Shah, a serial entrepreneur with previous experience in AI. Shah previously co-founded and sold Like.com to Google and was CEO of Health110, acquired by UnitedHealth. From the outset, Hippocratic AI's mission has been to solve the global shortage of healthcare workforce, and it specializes in developing large language models (LLMs) specifically designed for the healthcare sector. It also develops AI-powered virtual healthcare assistants that can perform non-diagnostic communication tasks in the healthcare field. It has investors such as General Catalyst, Andreessen Horowitz (a16z), and Menlo Ventures, and has achieved unicorn status in less than two years. It is that significant investment that it has received in such a short time along with its specialization in a critical sector, which, according to experts, positions the start-up among the most promising of those dedicated to AI.
In addition, their model has outperformed GPT-4 and Claude in tests of medical information accuracy, a critical differentiator in an industry that does not admit errors. It should be noted that the company has developed a system of guardrails (protective barriers) that prevents the system from providing potentially dangerous medical advice or from going beyond providing basic health information.
MULTIVERSE COMPUTING. The only Spanish start-up in the AI 100 classification can be considered, according to experts, within the world's top 10 among the most promising AI companies. Considered the largest quantum software company in Europe by number of employees and one of the most relevant worldwide in its sector, Multiverse Computing was founded in 2017 by Enrique Lizaso-Olmos, Sam Mugel and Román Orús, and develops quantum computing software and algorithms – and quantum-inspired – that allow large corporations to solve complex problems in sectors such as finance, finance, finance, energy, automotive, industry, health or sports.
Among its innovations is the possibility of offering these algorithms in a simple way, integrated into platforms that can be used from tools as common as Excel, and that allow companies to take advantage of quantum potential without the need for advanced knowledge in this technology.
In October 2021, the company raised €10 million in a funding round, which was joined by another in March 2024, of €25 million, led by Columbus Venture Partners. In March of this year, the Government of Spain, through the digital SEPI, announced an investment of 67 million euros in Multiverse Computing, as part of its commitment to technological transformation.
QDRANT. This Berlin-based start-up specializes in the development of a high-performance vector database, designed specifically for semantic search and artificial intelligence applications. Its founder and CEO is Andrey Vasnetsov, who is considered influential in the technology sector for his expertise and insight in the field of vector databases and similarity-based information retrieval.
Compared to competitors such as Pinecone, Weaviate, Milvus, and others in the vector database space, Qdrant distinguishes itself with its combination of high performance, ease of use, and its business model that balances open source with commercial cloud services. In reality, the start-up provides a vector database solution that enables lightning-fast similarity searches. efficient metadata filtering, scale-out to handle large datasets, and advanced capabilities for modern AI applications.
TWELVE LABS. This start-up founded by two Korean engineers has raised €86.4 million, including €44 million from Nvidia, which has helped it develop AI-based video understanding technology. Its platform allows semantic searches in audiovisual content with unprecedented precision, understanding both the visual content and the narrative context. Its Marengo API is used by streaming platforms and media companies to automatically catalog vast video files. It reduces search time in audiovisual files from hours to seconds, and its multimodal models can understand gestures, emotions, actions and context in video, outperforming solutions from tech giants in specific benchmarks. In addition, he has formed strategic alliances with some Hollywood studios to optimize the post-production process.
LAKERA. Founded in 2021 by David Haber, Mateo Rojas-Carulla and Matthias Kraft, this German company has raised €8.8 million in funding to develop security technology specifically designed to protect applications based on large language models (LLMs). Its Lakera Guard product protects against multiple attack vectors, including malicious prompt injection, sensitive data extraction, and content filter evasion. Its free tool Gandalf has become a standard for AI security education. It went viral by challenging users to try to extract a password from a protected LLM model, demonstrating the security risks and generating more than 1.5 million attempts in a few weeks. It is used by more than 500,000 people to understand vulnerabilities.
Basically, Lakera specializes in security for artificial intelligence, developing advanced solutions to protect AI models against attacks, manipulations, and vulnerabilities specific to this technology. Its objective is to guarantee the integrity, privacy and reliability of AI-based systems.
Key trends
The ranking of the most promising AI start-ups of the moment prepared by CB Insights offers some fundamental keys to this sector. One concerns how AI agents dominate the conversation: these applications, which automate tasks and processes for human users, constitute the next generation of AI. Having been incorporated into virtually every horizontal and enterprise function, AI agents are also incorporated into vertical infrastructure and applications.
AI agents and supporting infrastructure account for 21% of companies this year, with investors considering this space a priority.
- Another conclusion is that the future is physical: "Artificial intelligence will evolve beyond software AI agents to a physical state. Advances in various areas of AI development, such as robotics, multimodal image and voice models, edge computing, synthetic data, and spatial intelligence, provide the foundation for physical AI, which combines artificial intelligence software with hardware to act in physical environments." The CB Insights study concludes that industrial humanoids represent an early manifestation of this, while future permutations could include fully autonomous defense drones, domestic companion robots, and more.
- In addition, machine learning security has become a safe bet. The need to secure AI applications has grown with the proliferation of generative AI and AI agents. Security is seen as the main obstacle to adopting generative AI, and machine learning security firms are hardening AI algorithms and foundational models, such as LLMs, while defending against increasingly sophisticated AI-powered attacks.
- Added to this is the fact that vertical applications are booming. In 2024, horizontal companies in this AI 100 cohort received more funding than their vertical and infrastructure counterparts: $1.6 billion, compared to $1.2 billion for both infrastructure and verticals. So far in 2025, however, the funding landscape is very different, with major companies leading the way with $1.1 billion in funding raised.
Check it out here: https://www.expansion.com/expansion-empleo/emprendedores/2025/05/04/68150cb2e5fdeaa6518b457a.html#