Look, just when everyone decided AI was a product category, a revenue line, and a boardroom buzzword, the smartest people in the room are quietly grabbing their coats and heading back to the lab. No demos. No roadmaps. No promises. Just research, and billions of dollars chasing it.
For the last few years, AI has been defined by shipping. Bigger models. Faster responses. Slicker demos. Quarterly updates framed as breakthroughs. But here is the tell that something more fundamental is happening. The people who actually know how this stuff works are walking away from the most powerful platforms on the planet to start labs that do not even pretend to be businesses, at least not yet.
They are not building features. They are chasing intelligence itself.
Over the past several months, a steady stream of respected AI researchers and leaders have left marquee roles to found what some investors and analysts are calling “NeoLabs.” These are not startups in the traditional sense. Many have no products. Many have no revenue. Some have no clear plans to generate either anytime soon. And yet they are raising enormous sums of money at multi-billion-dollar valuations.
One clear example is David Silver, a senior leader at Google DeepMind, who recently departed to found a new company called Ineffable Intelligence. As reported by Fortune, Silver is not chasing a near-term commercial win. His focus is on deeper questions around intelligence and learning, work that does not map neatly to a product roadmap or a sales funnel.
That move alone raised eyebrows. What makes it more interesting is how common it has become.
According to reporting by Kate Clark in the Wall Street Journal (behind a paywall), a growing number of venture-backed AI labs are being funded with the explicit understanding that they may not produce products or revenue for years, if ever. Investors are backing research itself, not applications of existing research. That is a meaningful shift.
One of the examples highlighted in that reporting is Flapping Airplanes, a research lab founded by Ben Spector, a PhD student at Stanford University. Spector did not walk into investor meetings with a traditional pitch deck or a plan to monetize quickly. He had a lab, an unconventional approach to training AI models inspired by biology, and a desire to assemble a small team to tackle foundational problems. Venture capital firms rushed to fund him anyway.
And this is not limited to established researchers with long resumes. Students finishing, or not finishing, their PhDs at places like Stanford and MIT are increasingly leaving academia early to found labs of their own. The opportunity to raise large sums of capital quickly has pulled talent out of universities and into venture-backed research organizations at a pace many academics say they have never seen before.
Then there is Safe Superintelligence, the lab founded by Ilya Sutskever, a co-founder and former chief scientist at OpenAI and one of the key figures behind the modern large language model era. Sutskever has been unusually direct with investors. He is investigating ideas he believes are promising. He is not guaranteeing a product, a timeline, or a revenue stream. He has said openly that AI is returning to an age of research after years dominated by scaling and commercialization.
Investors, at least for now, seem comfortable with that answer.
This represents a reversal of how AI innovation has traditionally been funded. In the past, the most ambitious AI research lived inside academia or corporate research arms like DeepMind. Startups focused on packaging that research into applications that could make money. The AI boom has flipped that model. Capital is now flowing directly to research labs, with the expectation that if a real breakthrough happens, everything else will follow.
Not everyone believes this ends well. The technical challenges these NeoLabs are attempting to overcome are enormous. Several investors quoted in the Journal expressed skepticism that most of these efforts will produce results that matter. Being incrementally better is not enough. Talent retention is also a serious concern. When the largest tech companies are offering compensation packages north of nine figures to recruit AI experts, keeping researchers committed to a long, uncertain research journey is not easy.
We are already seeing stress fractures. High-profile researchers have left new labs to return to Big Tech, rattling investors and prompting tougher questions about incentives and long-term commitment. Are founders and teams in this for a decade of research, or until the first life-changing offer comes along?
Still, stepping back, it is hard to ignore what this moment signals.
AI spent decades in the lab before neural networks and large language models exploded into commercial reality. Long winters of slow progress preceded the sudden breakthroughs that reshaped our industry and, frankly, our world. The last few years have been about harvesting that work at scale.
Now, the smartest people and the smartest money appear to agree that to get the next wave, we have to go back to fundamentals. Back to research. Back to asking questions that do not have immediate answers or obvious business models attached.
Most of these labs are chasing some version of superintelligence. Whether that goal is achievable, or even desirable, is a debate for another column. What matters is that the critical mass is there. Capital, talent, and attention are converging around foundational AI research at a scale we have not seen before. And this time, it is happening alongside rapid advances in other frontier areas like quantum computing, robotics, and scientific automation.
Shimmy’s take is simple. We waited decades for the last great AI leap because the work was underfunded, underappreciated, and largely confined to academic corners. This time, the lab lights are on, the money is flowing, and the competition for breakthroughs is fierce.
If you think AI has been interesting lately, just wait. The next chapter is being written back in the lab.
