Investment in generative AI startups, which develop AI-powered products to generate text, audio, video, and more, shows no signs of slowing down. However, these investments are becoming more concentrated in a smaller number of early-stage ventures. From January to July 16, 2023, 225 generative AI startups raised a staggering $12.3 billion from venture capitalists, according to data from Crunchbase. If this trend continues, generative AI companies are on track to match or surpass the $21.8 billion raised in 2023.
The breakdown of the first half of 2023 by investment stage shows 198 angel/seed deals totaling $500 million, 39 early-stage deals amounting to $8.7 billion, and 18 late-stage deals reaching $3.1 billion. Early-stage startups were the clear winners, with notable investments including Elon Musk’s xAI, which raised $6 billion in May, China’s Moonshot AI with $1 billion in February, Mistral AI with $502.6 million in June, Glean with $203.2 million in February, and Cognition with $175 million in April. According to Chris Metinko, an analyst and senior reporter at Crunchbase, investors are focusing on larger startups they believe have a higher chance of success while allowing less promising ventures to “wither away” at earlier stages.
Metinko noted that some VCs anticipate legal and regulatory challenges in the U.S. and overseas could slow the flood of AI funding. Others recall that during the mobile revolution over a decade ago, the biggest winners in the foundational infrastructure layer were well-established tech companies. The future of many generative AI businesses remains uncertain, even for the most well-funded ones. Generative AI models are typically trained on publicly available data such as images and text from the web. Companies argue that fair use protects them from legal challenges over copyrighted material. However, it’s not yet clear how courts will rule, prompting some AI companies to secure licensing deals with copyright holders.
Regardless of legal outcomes, high-quality training data is becoming harder and more expensive to obtain. As startups exhaust available data on the web and more content creators block data scraping, the cost of acquiring training data is expected to soar. One analysis estimates the AI training data market will grow from $2.5 billion to $30 billion within a decade. Training models is also becoming more costly and complex. For instance, OpenAI’s GPT-4 cost $78 million to train, while Google’s Gemini had a price tag of $191 million, according to a recent Stanford report.
Given the substantial upfront investment needed to develop flagship models, few generative AI startups are profitable, including industry leaders like OpenAI and Anthropic. OpenAI, despite reportedly generating around $3.4 billion in revenue, could face losses of $5 billion this year, according to The Information. Investors in generative AI appear to be playing the long game, especially major tech investors like Google, Amazon, and Nvidia, which view these investments as strategic bets. However, there is a looming question of whether the generative AI bubble could burst soon. If generative AI startups cannot overcome the significant challenges they face, this possibility seems increasingly likely.
