“That does seem frothy,” Thomas Rice, the portfolio manager for Perpetual’s Global Innovation Share Fund, told The Australian Financial Review.
“Extreme valuations of companies that haven’t actually done anything yet are signs of a potential bubble in the start-up space.”
Even Sam Altman, the co-founder and CEO of OpenAI, now uses “AI” and “bubble” in the same sentence.
“It is definitely like the new bubble – all the people that were working on crypto a few years ago are now working on AI,” Mr Altman told The Financial Review last week.
‘It’s completely nuts’
Daniel Petre, a former Microsoft vice president who was heavily involved in the dotcom boom of the 1990s and went on to co-found AirTree Ventures, one of Australia’s largest venture capital firms, said start-ups scrambling to cash in on AI reminded him of the “big data” trend of last decade, and the blockchain trend more recently.
“At one point you just had to say ‘big data’ in a pitch deck and people were swarming all over you. ‘Blockchain’ was one of those terms, too. And now it’s ‘machine learning’. It’s completely nuts,” he said.
Ramin Marzbani, who was one of Australia’s leading analysts in the dotcom era and now runs a small venture capital firm, said the investment pitches he was seeing from software companies were “deja vu all over again”.
This tech bubble is different, says Daniel Petre, co-founder of AirTree Ventures. Louie Douvis
“We’ve had a couple of pitches from businesses that have been writing software for around 10 years, and now they’re rebranding the exact same product as AI just to get a higher valuation, when nothing has fundamentally changed about the product they were making,” he said.
Despite the bubble forming around AI start-ups, the technology itself is profound and very real.
Mr Petre said: “This is actually a genuine thing, unlike web 3.0, which kind of disappeared without a trace.
“Machine learning will have far more impact at the technology platform level than blockchain ever did. Or in fact, probably more than any other underlying technology that we’ve ever seen since humanity worked out how to make fire.”
Ironically, though, many of the start-ups cashing in on AI to inflate their value might actually see their long-term value undermined by AI, he warned.
Machine learning was making it much easier for start-ups to write new applications, but that also meant it was making it much easier for competitors to emerge, he said.
And unless start-ups had their own differentiated data sets to train their AI models on, a lot of the value would be captured by the companies making the AI platforms, such as Google and OpenAI, he said.
“You might get a million users, but someone could start tomorrow and get a million users as well. There’s no differentiation so is differentiation going to be a pricing race to the bottom?” Mr Petre asked.
“If, however, you have a scaled and differentiated data set that you’re training your algorithm on, then that’s where the Holy Grail is.”
Perpetual’s Mr Rice agreed that the AI boom was different from the past few tech bubbles, and that the key to investing in it would be to look at the data sets being used.
“A lot of people will look at this AI boom and say, ‘Oh, I’ve seen this story before. I saw it with the metaverse, I saw it with blockchain’,” he said.
“But you need to ask, can we see real-world use cases of this technology in practice? And I think the big difference here is, I could never really see that with blockchain, but I can see plenty of examples of real-world usage with AI and large language models.
“If I was looking at the start-up space, I would be trying to focus on what is the unique thing that this company is doing, what’s the unique data set they’re providing.
“I think there’s an opportunity to add value there,” he said.
