European startups have only raised $1 billion of the €22 billion in venture capital invested in generative AI since 2019, according to data from Dealroom.
Not surprisingly, American companies attracted most of the money. A whopping $20 billion (89% of the global total) went to American startups. Their Asian counterparts raised just $790 million, while the rest of the world combined raised just $454 million. Dealroom data collection was completed on July 10, 2023.
AI is not new. It has been used in data-driven and analytical workflows with increasing sophistication for decades. But creativity and ideation were considered human skills, far removed from the capabilities of artificial intelligence.
The emergence of generative AI (GenAI) and programs like StableDiffusion and ChatGPT have turned this assumption on its head. Generative AI is a new frontier in AI, using large language models (LLMs) trained on large data sets of multimedia content (text, images, audio, video) to create new text, audio, images and more.
During their rapid emergence, generative AI startups have attracted considerable funding from investors, with more than $22 billion in funding over the past five years.
The United States is the undisputed leader in venture capital funding for generative AI, although we don’t limit ourselves to OpenAI.
GenAI immediately caused a ripple in adjacent markets. NVIDIA shares rose more than 100% in the first half of 2023 (NVIDIA is a leader in AI chips), while companies like Chegg (educational tutoring) lost more than 50% due to the disruption of their business model by from GenAI.
Medium round size
Generative AI startups show significant premium in their Seed and Series A rounds.
Notably, the average size of Series A rounds is double that of other startups. This maintenance also dates back to 2018.
Venture capital financing step by step
Early-stage investments in generative AI startups have shown a steady increase since 2016, with an acceleration starting in 2020. However, the numbers are likely to increase slightly due to some delay in reporting on smaller rounds.
Late-stage funding for GenAI increased five-fold between 2020 and 2022, but has slowed slightly in recent quarters.
Late-stage funding peaked in 2023 with over $12 billion in funding, led by OpenAI’s $10 billion funding, but also several other initiatives.
As a relatively nascent industry, most venture capital funding so far has been raised by those closest to LLM. Model developers received more than 60% of GenAI funding, followed by applications and infrastructure.
Model maker OpenAI leads the funding raised by GenAI companies, but Anthropic, Adept AI, Inflection AI, Aleph Alpha and a handful of other players have also raised significant sums. In general, considerable funding is required to support the high training and implementation costs of general LLM models.
Verticalized model makers are starting to emerge, such as Hippocratic.ai, which came out of hiding with a $50 million seed round for its health-focused LLM. Specifically developed LLM-ready sectors include healthcare, fintech, and legaltech.
Applications are the second most funded segment of generative AI after modelers. The use cases cover all media types (text, image, video, voice/audio/music, code and 3D resources). Most applications have been built around text, such as writing, customer service assistants/chatbots, and knowledge and search. Other notable segments are code generation, image generation, speech generation, and game design.
Applications are divided into those built on proprietary models and those built on third-party models.
Most apps are based on third-party templates such as Jasper and Typeface. But several startups are creating applications based on their own GenAI models. Character.ai, Runaway and Descript are examples.
Creating proprietary GenAI models can provide a hedge against competition, as applications will likely leverage collected data and user interaction to refine proprietary models. Others can create model refinement layers on top of third-party models.
The dramatic increase in the use of GenAI across multiple use cases, both consumer and enterprise, has created the need for a dedicated infrastructure that spans from rapid engineering to MLOps (training, deployment, optimization and monitoring), to data and integration .
Some of these solutions are added as add-ons to previous MLops offerings, as in the case of Scale AI.
Others meet specific GenAI needs, such as the Vector databases, which raised a record $177 million in 2023, led by Pinecone and Weaviate.
Many prominent investors have built their generative AI portfolios. Andreessen Horowitz and Sequoia have invested nearly 50% more in generative AI than anyone else to date.
Ycombinator is by far the most active accelerator for GenAI startups, with over 100 supported startups including OpenAI, Jasper, and Replit.
Main countries and cities.
The world’s leading country in generative AI funding is the United States, with a large lead, followed by Israel and Canada. They are followed by the United Kingdom, Germany, the Netherlands and Sweden.
The Bay Area has been the leading hub for generative AI, attracting more than $18 billion in less than four years. Even without OpenAI’s $12.3 billion in funding, the Bay Area attracted 8 times more funding than the next hub: New York. Tel Aviv (AI21) and London (Stability.ai) follow as the two main global centers outside the United States.
AI chips: the pillars of GenAI
The GenAI wave is increasing demand for AI chips and processors for large-scale LLM training and deployment. This sent NVIDIA stock up more than 100% in the first half of 2023 (NVIDIA is a leader in AI chips). However, even NVIDIA is two to three months behind in fulfilling new orders for cloud server chips. Training costs and availability of computing power are becoming a constraint for startups and companies wanting to train and implement LLM.
Globally, funding for AI chips began to increase in 2017-2018 and peaked in 2021-2022.
However, when filtering the rounds, 2022 was the most active year on record in terms of number of rounds.
China has been the leading geographic region for investment in AI chips.
However, despite the growing and clearly unmet demand for AI chips and the limits of semiconductor computing power, AI chip startups have, in some cases, still failed to deliver on their promises.
After being announced with much enthusiasm, Sequoia Capital reduced its valuation of British startup Graphcore to zero in April 2023, after losing a major contract with Microsoft in late 2022. Graphcore raised $682 million in venture capital financing and was valued at $2.8 billion in December. 2020.
Source: Agreement Room
And you ?
What is your opinion on the subject?
What do you think of the results of this Dealroom study? Do you consider this data to be relevant and reliable?
How can we explain the delay accumulated by Europe in terms of funding for generative AI?
McKinsey predicts that AI will add up to $4.4 trillion of value to the economy annually and $13.4 trillion of annual value added to the global economy by 2030.
Generative AI job openings in the US increased 20% in May 2023, while technology jobs decreased 43.6%, according to data from Indeed.
Generative AI is developing faster than smartphones, according to a report from InsiderIntelligence that compares their respective developments since their launch
Gartner places generative AI at the top of hype in the 2023 emerging technologies hype cycle. Emerging AI will have a profound impact on business and society.