Published on the 31/05/2024 | Written by Heather Wright
The emperor’s new clothes…
AI companies are raking in the billions in investments – despite many having yet to produce viable products – with more established ‘old school’ tech companies also pouring the dollars into AI efforts. But for all that global investment, cracking the commercial model which can unlock revenue returns seems to be eluding many and wariness about an ‘AI bubble’ are, just like a bubble, rising.
US venture capital firm Sequoia Capital, whose investments have included YouTube, Apple, Reddit and PayPal, recently estimated that it has invested US$50 billion into the chips needed to train large language models.
CB Insights says US$21.8 billion was poured into generative AI deals last year, a fivefold increase on 2022 figures. And Pitchbook figures show valuations for early and late-stage AI companies are outpacing those in other verticals with a median early stage valuation of above US$70 million for early stage and around US$100 million for late stage companies.
“I am short chat interface and long on SaaS and enterprise implementations that use LLMs.”
Big name tech companies are also splashing the cash. Meta announced recently that it’s planning on spending $35 billion on AI this year. The company’s stocks plummeted 15 percent on the news.
OpenAI raised funds in February at an US$80 billion valuation – nearly three times its valuation just a year ago. Rival Anthropic has seen similar gains. And the market collective market capitalisation of Microsoft, Amazon, Alpahbet and Meta? According to Reuters, that’s up $5 trillion – that’s not a typo – since ChatGPT’s November 2022 release.
If you need even more eye watering figures, this week, Elon Musk’s xAI nabbed US$6 billion from investors including Sequoia, Andreessen Horowitz. And that pales in comparison with OpenAI boss Sam Altman’s reported attempts to raise up to US$7 trillion (yes, trillion) to increase semi-conductor chip building capacity globally to stop AI chips limitations continuing to slow OpenAi’s growth.
But while the money might be flowing into the AI companies via investors, the returns certainly aren’t. Revenues from genAI startups came to a paltry (in comparison) US$3 billion, Sequoia estimates.
Earlier this year, tech investment analyst Richard Windsor noted: “Capital continues to pour into the AI sector with very little attention being paid to company fundamentals in a sure sign that when the music stops there will not be many chairs available.”
In a March post, Windsor highlighted Cohere, an OpenAI competitor, and its ability to raise money at a valuation 2.4x the valuation of just nine months earlier, despite disappointing business development, as ‘the latest sign of this reckless abandonment’.
Cohere, which has raised nearly $450 million from investors, has a valuation of $5 billion, but it generated just $13 million in annualised revenue in 2023.
“Cohere’s valuation equates to a historic price/sales ratio of 384x which indicates that investors have another bad case of Fomo and are rushing into anything that can be remotely associated with AI,” Windsor said.
Windsor says he believes providers of genAI services who are raising money on the promise of selling their services for $20/month per user will be the ones to bear the brunt of the correction when it comes.
“The problem is that there are many of these, all of whom are demonstrably similar,” he says. That, combined with the multitude of free open source offerings, will put the squeeze on pricing.
“It is at this point that the flow of money is likely to slow down and then stop as falling prices will mean that targets are missed and startups go back to VCs cap in hand,” he says.
Windsor certainly isn’t alone in questioning the big investment pouring into AI. Last year, Garry Tan – dubbed one of the most successful tech investors in Silicon Valley and the CEO of accelerator Y Combinator – noted his concerns about the business case of consumer-facing AI technology.
“Personally I am short chat interface and long on SaaS and enterprise implementations that use LLMs,” he tweeted in July after reports ChatGPT’s traffic was declining.
The inability to find a commercial model was blamed for the semi-implosion of Inflection AI in March. The company had raised more than US$1 billion in 2023 at a $4 billion valuation. Then more its co-founders and most of its 70 person team jumped ship to Microsoft (a company which has already invested $13 billion in OpenAI) in a $650 million deal.
Inflection’s key offering was the Pi AI emotional support chatbot for consumers. It’s now reportedly refocusing its efforts to target business customers with APIs on Azure and other services.
AI is inherently pricey to build out, with high requirements for GPUs, data centres, energy and skills and finding a clear ROI can be a longer term proposition for all concerned.
A recent Gartner survey found nearly half of survey participants cited difficulties in estimating and demonstrating the value of AI projects as the number one obstacle to AI adoption.
Leinar Ramos, Gartner senior director analyst, says business value continues to be a challenge for organisations when it comes to AI.
On the vendor front, Meta has highlighted potential strategies ranging from subscriptions to advertising.
For now though, the money continues to flow, at least from investors.
