The Lex Newsletter: every start-up is now an AI start-up

Receive free Artificial intelligence updates

We’ll send you a myFT Daily Digest email rounding up the latest Artificial intelligence news every morning.

Dear reader, 

The speed with which generative artificial intelligence has steamrollered the US tech sector is remarkable. At a recent event, a partner at venture capital firm Index Ventures declared that he could not think of a start-up pitch he had received this year that did not mention generative AI. 

Wind back to last year and the buzz was centred on cryptocurrency and blockchain. That has now all but evaporated, along with crypto prices. VC funding is lacklustre everywhere but AI. It would take a brave company not to crowbar AI somewhere into the pitch deck.

The excitement OpenAI created when it launched its ChatGPT chatbot last year means that every meeting I go to is now filled with talk about large language models, compute power and hallucinations. The tech sector hopes AI is incubating the next cohort of trillion-dollar companies.

Look at PitchBook data on deal values for start-ups raising money, however, and it may not look like the last two quarters were anything to write home about. That is because deal counts do not include the volume of deals announced but not yet completed. Add those transactions, and there is a sudden jump from less than $1bn in the first quarter of 2022 to more than $10bn in 2023. The total so far this year is close to $13bn. By 2030, Grand View Research estimates that generative AI products could reach $110bn.

Hopes for the future size of the market and the scale of Microsoft’s $10bn investment in OpenAI have encouraged rivals to write huge cheques. Take French start-up Mistral AI, which wants to create a large language model (LLM) that can compete with OpenAI’s ChatGPT. It raised $115mn in June despite being just a few weeks old and having no product to show. 

Funding is still climbing. Recently, Amazon announced a $100mn programme that will connect customers and engineers focused on AI. It hopes that via its cloud computing business AWS it can carve out a role as the go-to platform and cloud provider for generative AI.

Which are the start-ups to watch? There is OpenAI of course. The central San Francisco start-up has raised more than $11bn so far. Cohere in Toronto is also developing LLMs for businesses to use. Others include SF start-up Anthropic, whose founders came from OpenAI; Inflection AI, which is trying to create an empathetic, AI chatbot; and Hugging Face in New York, which advocates for open source and acts as a hub for companies to find AI models.

All this excitement has lifted valuations across the entire tech sector. Most share prices have not quite reached the heights of late 2021, but they are fast erasing last year’s slump. The Nasdaq Composite index is up almost 30 per cent this year. 

But consider the challenges. Hallucinations come about because LLMs cannot tell truth from fiction. That means certain sectors — financial, legal and medical for example — are right to be cautious about implementing AI tools. Concerns about the use of deepfakes to impersonate people could create political and financial problems too. 

There is also the coming EU AI Act — a regulatory framework that could restrict generative AI software. That may hand an advantage to incumbents and make life harder for start-ups in the sector. Larger companies have more problems with compute power, however, which is proving a bottleneck.

And then there’s data. The days of AI start-ups scraping any online data they like in order to train their LLMs are coming to an end. The real owners of that data are more cognisant of its value. That is illustrated by Reddit changing access to its application programming interface. If AI companies are going to use Reddit to train their LLMs, Reddit wants to be paid. That is a fair point of view. 

The Center for AI Safety might say that mitigating the risk of human extinction from AI should be considered the final and most important challenge.

But we are at such an early stage that it remains hard to untangle hype from reality. What is less contentious is that the many AI start-ups do not yet have a clear idea about the sort of money their services will make. This is an expensive sector in which to operate. How much will VCs pour in before they start asking where the returns are?

Elsewhere in tech news 

Surely this daft cage match between Elon Musk and Mark Zuckerberg is not going to happen?

A fascinating story in The Information on investors who are offering AI start-ups access to prized Nvidia chips. 

This Slate article recalls what it was like to work as a 20-something before smartphones.

Enjoy the rest of your week,

Elaine Moore
Deputy head of Lex