Not everyone finds the return that generative AI can promise valuable. Based on the latest funding numbers from PitchBook, however, apparently many investors do.

In Q3 2024, VCs committed to $3.9 billion in deals across 206 generative AI startups, PitchBook found. And, of that, $2.9 billion of funding went to U.S. companies in 127 deals.

Some of the biggest winners in Q3 were coding assistant Magic (August $320 million), enterprise search provider Glean (September $260 million), and business analytics firm Hebbia (July $130 million). In August, Moonshot AI, of China, raised $300 million, while Japanese startup Sakana AI, focused on scientific discovery, closed a $214 million tranche last month.

Critics say generative AI has too broad of an umbrella with all its varied technologies -- ranging from text and image generators to coding assistants, cybersecurity automation tools, and even more beyond that. Experts question the reliability of the tech, and, in the case of generative AI models trained on copyrighted data without permission, the legality of the practice.

But VCs are effectively making a bet that generative AI will get traction in large and profitable industries, and that the long-tail growth it enjoys today will not be overwhelmed by its challenges.

Maybe they're correct. A Forrester report predicts that 60% of the skeptics toward generative AI will have the tech — knowingly or unknowingly — to work with for tasks ranging from summarization all the way to creative problem solving. That's a lot more sanguine than Gartner's prediction earlier this year that 30% of generative AI projects would be abandoned after proof-of-concept by 2026.

"Big customers are now rolling out production systems which exploit startup tooling and open source models." Brendan Burke, senior analyst for emerging tech at PitchBook said to TechCrunch in an interview. The latest wave of models shows that new generations of models are possible and may excel in scientific fields, data retrieval, and code execution.

One of the significant challenges to the large-scale adoption of this emerging generative AI technology is its huge computational requirements. According to a new study by Bain analysts, it will be challenged by this technology to build gigawatt scale data centers - data centers that consume 5-20 times what an average data center consumes today - stressing an already-strained labor and electricity supply chain.

Already, generative AI-driven demand for data center power is propping up coal-fired plants. Morgan Stanley estimates that if this trend were to continue, global greenhouse emissions between now and 2030 could be three times higher than if generative AI hadn't been invented.

But it may be decades before those investments pay off. Several of the world's biggest data center operators-Microsoft, Amazon, Google, and Oracle-have each said they'll invest in nuclear to help counterbalance their growing consumption of nonrenewable energy. Last month, Microsoft said it will draw electricity from the infamous Three Mile Island nuclear plant.

Investments in generative AI startups show no signs of slowing down — negative externalities be damned. The viral voice-cloning tool, ElevenLabs, is allegedly in funding discussions at a $3 billion valuation, while Black Forest Labs, the company behind the X notorious image generator, is said to be in talks for a $100 million funding round.