At A.Team's first Generative AI Summit a panel of four top VCs explored the future of AI businesses, focusing on the potential for profit-making.
Nearly every player in the AI ecosystem is burning cash in the name of growth. Their computational demands are costly, with even industry giants like Google struggling to meet the needs of every application.
Successful AI startups must differentiate through execution, build communities, and leverage unique data capture to gain a competitive edge in industries where AI is already being adopted.
Join us on June 15th for our Generative AI Salon — a monthly gathering of top AI builders.
Here’s a question that might stump ChatGPT: “How do you turn a profit with a new AI product?”
At A.Team's first Generative AI Summit in New York City four venture capitalists pondered the future of the artificial intelligence business. When moderator Morgan Blumberg of M13 posed the profitability question, “Where are the profitable AI companies today?” nervous laughter rippled across the room.
“Sorry, what is, uh, ‘profit’?” grinned Matt Turck, Managing Director at FirstMark, and the organizer of New York City’s Data Driven meetup.
“They’re in Taiwan,” quipped Madison Hawkinson, an investor at Palo Alto-based Costanoa Ventures, referencing the chip manufacturers whose processors must bear AI’s exponentially increasing computational demands.
In 2023, generative AI tools have become the tech world’s collective obsession. “Pretty much every engineer I know is hacking on some AI side project,” said Grace Isford, partner at Lux Capital in New York. “It’s changing the game.”
But is AI a smart bet for investors? How do you profit from something at once free to use, yet obscenely expensive to operate? And how can companies build with AI and win in this chaotic arms race?
It’s inscrutable, paradoxical—the b-school case study equivalent of a zen koan.
Yet all four panelists—Blumberg, Turck, Hawkinson, Isford—offered clever answers. Here’s how they are making sense of the startup world swirling around generative AI.
A Very Costly Free-for-all
Our AI appetites are bigger than our wallets.
As we sit at home and mess around with chatbots, it’s easy to forget that each silly question typed into ChatGPT and every whimsical sketch generated by Midjourney requires leviathan computational efforts that run up electricity bills and strain CPUs in data centers far away—costing someone real money.
“I was at Google last week, and even they don’t have enough computing power to use AI for every single aspect of every single application,” said Isford.
So far, practically every player in the AI ecosystem is burning cash in the name of growth. But the party may not last forever. Isford says that before she invests in a company, it must have a rock-solid “compute contract” with one of the major cloud providers that do the actual data-crunching. In other words, the company has a bead on the ultimate cost to them as they grow. “It’s about where the buck stops, right?”
Want to Get Ahead? Build a Good Business
When Turck started at FirstMark Capital a decade ago, neural networks set techie hearts pounding. A few years later, artificial intelligence buffs were consumed by crypto, then chatbots, then crypto again. Now we’re back to chatbots, observed Turck. “This is my third hype cycle for AI,” he chuckled.
Successful AI startups will beat competitors the same ways as their predecessors did, Truck argued, “by differentiating through execution, building community, iterating, going fast."
“Will Jasper be defensible or not defensible?” he asked rhetorically, referencing the fast-growing copywriting software startup that helps marketers harness large language models for their work. Who knows? “They just need more time on earth—to build a community, a reputation, and a ‘logo wall,’” the same things all good software companies build.
“Like all new things, [AI models] are amazing, but ultimately, it’s software,” Turck said. “It’s more or less the same thing at the end of the day.”
As in previous data technology booms, visionary startups will use the powerful capabilities of the new technology—in this case, open-source AI models—to their niche’s needs and tastes in fresh ways. One example: Runway—a portfolio company of Isford’s Lux Capital—provides graphic designers and other visually-driven creatives with over 30 AI “Magic Tools” they can use to generate and edit images and video.
Like all new things, AI models are amazing, but ultimately, it’s software.
The most vulnerable industries, predicted Hawkinson, will be those where today’s market leaders compete on brand and price—while all are already leveraging the same technology. She added she’s got her eye on travel and e-commerce.
Another key differentiator may be the data itself. AI models are what they eat, essentially—solid and well-defined training data makes superior tools—and a defensible edge. “I’ve been obsessed with who has unique data capture,” said Isford. “Do you have unique subterranean sea data? Or biodata? Or data that no one else has access to?”
In a world where no one has the corner on artificial intelligence, real smarts still provide the only natural edge.