𝐀𝐈 𝐅𝐨𝐫𝐞𝐯𝐞𝐫

Below is a chart from Coatue’s latest East Meets West (EMW) conference (June 2025) which makes the argument AI is a new wave of innovation. And it is.

My beef is as follows:
⏩ Using the AI mantra 𝐢𝐧𝐝𝐢𝐬𝐜𝐫𝐢𝐦𝐢𝐧𝐚𝐭𝐞𝐥𝐲, without reminding the audience of the limitations of the current technology.
⏩ 𝐌𝐢𝐬𝐬𝐞𝐥𝐥𝐢𝐧𝐠 AI technology. Take AI applied to health: Microsoft touts it can “outperform doctors”. The reality is more prosaic: AI come ups with diagnostics much more cost-efficiently. Of course, AI still misses at least 15% of cases (today). AI will be deployed first for cost reasons, let’s have an honest discussion about it.
⏩ 𝐎𝐯𝐞𝐫𝐞𝐬𝐭𝐢𝐦𝐚𝐭𝐢𝐧𝐠 the impact of “AI” short-term (and in the process burning $ billions of investors’ money). Of course, longer term, AI (and I am not even sure this will resemble “AI” as we know it today) will change the world. Just like the PC, mobile or the cloud did.

Yes, we can make the argument that this time the speed is different. In an exponential world, AI’s impact will show through more quickly.

However:
⏩ The market is 𝐜𝐨𝐧𝐟𝐮𝐬𝐢𝐧𝐠 all models – see Coatue’s presentation ending up with “Innovation is everywhere” (no kidding, duh). AI is mingled with data centres, GPU rental (e.g. CoreWeave), crypto (e.g. Circle), etc
⏩ 𝐏𝐫𝐢𝐜𝐢𝐧𝐠 is clearly out of whack – OpenAI valued at 30x revenues, Thinking Machines Las’s valuation at $10bn for a seed round and no plans, etc.
⏩ Much of the hype is driven by 𝐬𝐞𝐥𝐟-𝐬𝐞𝐫𝐯𝐢𝐧𝐠 corporates – much of the investment in “foundational models” is from the very companies which will benefit the most from deploying these models over eye-wateringly expensive (and wasteful) data centres.

What are the smart investors to do?
⏩ 𝐒𝐢𝐭 𝐨𝐮𝐭 the foundational AI craze - See Balderton and Cathay Innovation new (and large!) funds focused on AI applications. This is not dependent on AGI, but the pragmatic approach that AI will be good enough in many cases. In short, the cost argument, but you will also have the productivity or staff augmentation argument, etc.
⏩ 𝐆𝐨 𝐟𝐨𝐫 𝐭𝐡𝐞 𝐧𝐞𝐱𝐭 𝐰𝐚𝐯𝐞 of physical AI – AI has been fed all the world’s digital data for training. Next is to feed it data from the real world. There is plenty of value in financing embodied AI.
⏩ 𝐅𝐢𝐧𝐚𝐧𝐜𝐞 𝐭𝐡𝐞 𝐩𝐢𝐜𝐤𝐬 𝐚𝐧𝐝 𝐬𝐡𝐨𝐯𝐞𝐥𝐬 – Much of the reluctance of enterprises to deploy is down to their inability to tap their own data. Remember data lakes? There is still plenty to do.

📎 Sources:
https://www.coatue.com/blog/company-update/coatues-2025-emw-conference

‍