Travis Kalanick is again. The Uber founder, ousted as CEO, is now betting on the bodily world. His startup has advanced into Atoms, spanning manufacturing, logistics, mining, and robotics. Jeff Bezos is doing one thing related.
Why select “atoms” over extra worthwhile “bits”? I requested Joe Fath at Eclipse, a VC agency that invests in equally gnarly bodily belongings.
Fath beforehand labored alongside Henry Ellenbogen at T. Rowe Worth, the place Ellenbogen pioneered tech startup investing on the mutual fund large. Fath usually joined these offers at later phases by means of large-cap development funds.
Here is my dialog with Fath, edited for size and readability:
Traditionally, “bits”-based companies — software and web — wanted far much less capital to achieve escape velocity, the place money circulate can maintain the enterprise with out extra funding. Buyers tended to draw back from extra capital-intensive fashions as a result of it has been exhausting to gauge when (or if) escape velocity reveals up. This has historically been a harder path than the “straightforward button” of software program and client web — however that is altering shortly.
Why is that this altering?
AI is driving a few structural shifts. First, it is accelerating and materially bettering digital duties — particularly throughout white-collar service work — as LLMs proceed to advance. Second, that functionality is beginning to transfer into the bodily world, the place traditionally analog environments can now be digitized by means of progress in areas like VLA (video, language, motion) fashions, world fashions (which simulate and assist programs perceive how the bodily world works), and embodied AI. The punchline: For the primary time, AI is making bodily industries meaningfully programmable — opening the door for quicker scaling with much less capital and labor, and beginning to flip the dynamic in query #1 on its head.
AI is making it simpler to provide much more digital content material, doubtlessly denting the worth. In the meantime, bodily issues are nonetheless exhausting, so these may retain their worth. Do you subscribe to this view?
I agree together with your premise. Marc Andreessen famously stated, “Software program is consuming the world.” Now AI is consuming software program — which is a giant a part of why you are seeing stress on SaaS within the public markets. However in bodily industries, you continue to should construct and function actual issues. AI cannot totally substitute the best way it may well take over a name middle, act as a coding agent, or flip a novice right into a succesful designer by abstracting away ability. And that dynamic is now extending into authorized, accounting, finance — there’s nowhere to cover.
The punchline: in bodily industries — manufacturing, logistics, power, protection — AI is a robust complement. It makes issues quicker, cheaper, much less capital-intensive, and finally drives higher returns. Nevertheless it’s not a full substitute in the best way it is going to be throughout giant elements of the service financial system. I am a robust believer on this thesis primarily based on what I am seeing — and it is a huge purpose I left T. Rowe to return to Eclipse. You are already seeing it play out in actual outcomes with firms like Amazon, Tesla, SpaceX, OpenAI, Anthropic — and I might add Anduril to that listing. My wager is there are numerous extra coming, and that capital that after flowed primarily into software program and client web will more and more shift on this path over the approaching years.
What examples are you seeing available in the market, or inside Eclipse’s portfolio?
The pattern dimension is rising shortly. I discussed a few of the family names in #3, and we’re seeing lots of the similar dynamics throughout our personal portfolio. Just a few examples: Wayve, Anduril, True Anomaly, Redwood Supplies, Mytra — plus earlier-stage names like Thoughts, Matter, and BotCo. After which you’ve gotten the “picks and shovels” layer powering the buildout of extra superior AGI — and finally ASI — capabilities, with firms like Cerebras, Oxide, and Tenstorrent.
Any new investments you are making that match into this broad thesis?
Talked about a number of in #4, however at a high-level, Eclipse’s focus areas stay general-purpose robotics, protection tech, next-gen manufacturing, semiconductors and next-gen compute, power and batteries, and autonomy. On the similar time, we’re continuously attempting to “go searching corners” — pressure-testing new theses and deciding the place it is smart to speculate and even construct from scratch. That features areas like fusion, water, area, world mannequin builders, workforce growth for the expert trades, and neuromorphic computing, to call just a few.
The place would possibly this thesis go unsuitable?
Bodily industries aren’t straightforward — you are not simply fixing a technical or engineering downside, you then should scale it. And if you’re bodily constructing issues, scaling effectively is extremely exhausting. Elon Musk has talked about this as “manufacturing hell,” and it’s totally actual. Even with AI serving to speed up and enhance execution, discovering sufficient sensible, succesful individuals to do that at a excessive stage continues to be a problem.
On prime of that, these companies require a whole lot of capital, and there is a rising wave of entrepreneurs getting into the area — so capital allocation actually issues. You want to again the appropriate outcomes: Defensible, high-impact concepts, distinctive founders who can entice prime expertise, and groups that may scale. Lastly, whereas the AI and geopolitical shifts really feel sturdy, it is price acknowledging the counterfactual. If world relations — significantly between the US and China — have been to materially enhance, you would see some slowing in re-shoring and the push to digitize bodily industries, with a partial reversion to lower-cost, analog manufacturing abroad.
Join BI’s Tech Memo publication here. Attain out to me by way of e mail at abarr@businessinsider.com.
