Travis joins James Li’s Breaking Points to discuss the issue.
Travis Oliphant is the creator of the foundational packages that make modern AI and machine learning possible: Numpy and SciPy. He is also the creator of OpenTeams. If you ask him if we’re in an AI bubble, the answer is simple: yes.
“We are witnessing a tremendous amount of money being poured into a system that is not clearly sustainable.”
Travis joined James Li’s Breaking Points to discuss the issue.
According to Travis, to panic and dismiss AI as hype is to fundamentally misunderstand what is happening. Technology is not the bubble, the business model is.
For Travis, the bubble is a misallocation of resources on a historic scale, fueled by a level of capital expenditure that is not justified by revenue. The prevalence of “circular deals” or “round-tripping” within the AI ecosystem is the largest indicator.
The mechanism operates as follows:
This creates a closed-loop financial system where capital flows from a giant corporation, through a startup, and back to the giant corporation, artificially inflating revenue figures. This creates an illusion of revenue growth and demand that disguises the lack of organic market uptake.
This model is failing.
The current valuations assume AI will be “job-replacing”. However, Travis Oliphant argues, this vision fundamentally misunderstands the technology’s current capability and the nature of human labor.
The transition from “I can respond intelligently to a conversation” to “I can actually do a job for you” is much harder than expected.
According to Travis, AI today is a “wonderful tool to enhance humans”, not automated, job-replacing robots that justify $½ trillion valuations. The burn rates required to train these models are unsustainable.
Forgetting this is capital’s great mistake. Every breakthrough, from the underlying frameworks to the foundational concepts, grew from cutting-edge communities of open source contributors. The sustainable path for AI is through them.
If we want to achieve a utopian vision of AI-enabled prosperity, we must build on the principles that make powerful technology like AI possible: distributed ownership and distributed decision-making. This is the exact opposite of concentrated business models.
The panic over monetization is a failure of imagination. After 20-plus years building cutting-edge tools, Travis is adamant that you don’t have to trade independence for convenience. Saas model AI is not the only option. You do not have to be trapped in endless subscriptions, with no control over your data, your models, or your future.
The alternative is Sovereign AI. It’s a model where organizations and governments own their intelligence. This fosters a horizontally integrated ecosystem of partners, instead of a vertically integrated monopoly.
Consider the difference. Under the current model, a hospital pays an endless subscription to a “black box” service, sending its sensitive patient data to a third party. Under a Sovereign AI model, that same hospital uses open-source tools to train its own model on its own data, inside its own secure systems. The intelligence, and the data, never leaves. This fosters a horizontally integrated ecosystem of partners, instead of a vertically integrated monopoly.
The bubble bursting is not a crisis for AI. It is a necessary correction. Capital is waking up to the fact that it has displaced sustainable approaches. The cutting edge communities that make modern AI possible are big enough and strong enough to carry this technology forward. The only question is how much economic disruption this speculative detour will cause before we get back to the real work of building a distributed, transparent, and prosperous future for everyone.
Every January, the global economic conversation moves to a small alpine town in Switzerland. Heads of state, founders, and technologists gather in Davos to debate the future. This year, beneath the familiar headlines about geopolitical tension and economic uncertainty, another theme dominated nearly every private conversation: artificial intelligence.
