Product
More
Legal
© 2026Daily24.All rights reserved.
AI is becoming critical infrastructure for companies and governments. But when the data, models and computing power sit in someone else’s hands, who is really in control?
Cohere co-founder and CEO Aidan Gomez joins CNBC’s Arjun Kharpal to explain why the next phase of the AI race will be shaped by sovereignty: where sensitive data lives, who controls the infrastructure and whether access to vital systems could be restricted or switched off. Gomez, one of the authors of the landmark 2017 “Attention Is All You Need” paper, argues that cost, governance and trust will matter as much as raw model performance as businesses move toward more autonomous AI.
He also assesses China’s rapid progress and explains AI’s double-edged role in cybersecurity—finding and exploiting vulnerabilities at scale, while also helping organizations detect and patch them.
Subscribe to “The Tech Download” wherever you get your podcasts.
#CNBC #CNBCPodcast #TheTechDownload
-----
Subscribe: @cnbci
CNBC International Live: @cnbcinternationalive
LinkedIn: https://www.linkedin.com/showcase/cnbc/
TikTok: https://www.tiktok.com/@cnbc
Facebook: https://www.facebook.com/cnbc
Instagram: https://www.instagram.com/cnbc
Threads: https://www.threads.net/@cnbc
X: https://twitter.com/CNBC
Telegram: https://t.me/cnbc
One silly thing about solving problems: there's two categories. Usually programmers talk about "Latency vs Bandwidth" but I'll change the names to "Inherent vs Capacity" for a more general audience. The main reason for the distinction is that capacity problems can be solved with money, but inherent problems really cannot. Usually these problems require time, or some clever solution to work around the part of physics that's in your way. You can sometimes convert inherent problems to capacity problems (ex: hiring a bunch of engineers to have the capacity to solve the root problem) but that starts to get into the weeds very quick. In terms of AI research, the main point is that you cannot assume that the dominant problems are capacity problems. If you throw money at the problem, but it's not a capacity problem, then you're not going to get a solution. Worse, if an engineer needs to make a trade-off, and they have a surplus of cash, then they will make the trade to be extremely cash-inefficient (because you always trade away whatever you have a surplus of). So... yeah. Money can only buy certain things. If only 5% of your problem is capacity, then throwing money at it will only buy you a really inefficient solution that doesn't address 95% of your actual problem.
Each time on any topic i hear more than three times ‘China is closing the gap” it turned out China is far far ahead and the West in denial. Then we start ‘blaming’ the Chinese for all sorts of ridiculous things, then impose constraints.
This problem has actually been solved. Cloud Engines. Fully on a decentralized blockchain, data storage and compute, including software applications, end to end, full stack. Nodes can run within one or multiple countries. AI agents and inference running on chain. The User (company, institution, government) can own all of the infrastructure and control their data and AI. Fully customizable to comply with governing agencies and/or institutional policies. No big tech (whether foreign or domestic), no 3rd party dependencies. Full stack sovereignty. The Internet Computer.
The one thing he keeps repeating is that we need more AI companies from countries other than the US and China for national security. That is the only way his company can survive and that is why his company is focusing on selling to governments. But it is going to be very difficult to compete in the consumer market that way
The lead is not the problem, we are leading, and been leading on AI since day 1, the problem is the ridiculous self defeating AI doomerism that this cntry is in the grip of.
I would let China lead on AI so that I might have a chance to see anybody copy it or not?😊
Grifters, everywhere!!!!