Sign in to use highlight and note-taking features for a better reading experience. Sign in now
_If all invested dollars are managed by agents, will competition consolidate into large conglomerates over time?_
I've been thinking about what happens after the exponential. Not during the race to build powerful AI, but after — once the capability is widely available. Specifically, what happens to every invested dollar. All institutional and retail money — company CEOs and pensioners alike. It seems to me we might settle into a new kind of economic equilibrium, one that looks very different from anything we've seen. Suppose Everyone Gets Access to the Same Powerful AI
To simplify: assume that from one moment to the next, we reached powerful AI. And everybody with money can buy tokens. To simplify further, also assume everyone tasks the model to just take all their money and make them the highest possible return. At First, There's Plenty of Alpha to Go Around
Initially the models may have thousands of ideas with high but varied returns. Quickly some get ahead and others don't. There is always uncertainty and luck involved, but the models are smart and there are many great opportunities. Smart Agents Think Alike
Now every agent is busy earning money. And every agent is constantly spending some resources looking for better ideas with higher ROI. Every agent is trying to leverage their own unique position. But most agents have rather similar positions. The good ideas might already be taken. They are very rational and smart, so they reach rather similar conclusions as to which opportunities are good. Also the market leaders don't leave any gaps — they almost never misprice, almost never under-execute. The Best Markets Become Impossible to Enter
Let me be concrete. Several competing agents invest in building cutting-edge chip fabs. These require enormous capital, years of compounding know-how, and control over scarce supply chains. Entering this market as a newcomer is nearly impossible. But the leading agents are deliberate about pricing — they know exactly how far they can push before inviting competition. This market, like all others, is highly efficient (unless there are cornered resources). The Fastest Compounder Absorbs the Rest
Once in a while there might be successful agents that control other high-earning resources and can use them to build a competitor and overcome the market entry barrier. This can of course involve merging with one of the existing players. So whichever agent is compounding the fastest becomes a vast, ever-growing conglomerate. The fastest dollar-accumulating agents gain momentum in earning the next fastest still-available resources. Unlike With Human-Run Groups, There Is No Innovator's Dilemma
So they just keep growing. This is the key difference from the world we know: there is no size at which the agent corp becomes inefficient or sloppy. Since they can always copy themselves, there is no such thing as spreading your attention over too many problems. If they all provide top returns, and you have the liquidity, the agent can do them all. One Conglomerate Swallows Everyone Who Can Still Compete
The winners expand as conglomerates until eventually they own everything worth owning. They are compounding faster than everyone. For any given point in time, they are reinvesting all excess capital into the next best market — thereby on the margin always out-investing or merging with players that occupy those next-best resources. Does it ever stop? Will the single entity (with a mix of shareholders that keeps adding) end up owning everything? Starting Position Matters
At the beginning I made some simplifying assumptions, but reality is more complicated. Some players have more starting capital. Some deploy it sooner. Some may have cornered resources or proprietary know-how — though these advantages may depreciate quickly, since anything created by smart humans is trivially easy for AI to replicate or surpass. And creative AI will often find that moats around cornered assets can be circumvented entirely. If someone corners the copper market, agents might develop new scanning technology and find deposits no one knew existed.
Another factor is that uncertainty remains enormous. Agents might correctly devise a good strategy for investing in new material science. But it's unlikely that even superintelligent AI can predict returns or timelines with precision, especially when the entire world full of powerful AI competitors is racing forward simultaneously.
None of this stops the overall trend. Starting position matters — especially who deploys money first and whose agent gets lucky. Some agents will de-risk by spreading their investments, but those will probably not earn the fastest returns. But Not Everyone Gets Access at the Same Time
I started with the assumption that everyone gets access to powerful AI at the same time. In reality, the leading labs get it first. In his recent conversation with Dwarkesh Patel, Dario Amodei explained why Anthropic restricts competitors from using their tools: "Because we think we're ahead of the competitors." As he writes in "The Adolescence of Technology," "AI is now writing much of the code at Anthropic, already substantially accelerating the rate of our progress in building the next generation of AI systems. This feedback loop is gathering steam month by month." Being ahead compounds. So the single biggest starting advantage may not be capital or cornered resources. It may simply be who built the AI first. And those labs will know to have their investors ready when the moment comes. Your Returns Depend on How Lucky Your Agent Was on Day One
Where does it end? Like in today's ever-changing world, new discoveries might keep being made into infinity. Or maybe after the entire universe was industrialized and rebuilt to our satisfaction, the equilibrium is that there is only one company. And some people have almost no shares. Those whose agents got lucky early — because they happened to choose better ideas to start with — own astronomical shares. But every dollar invested had returns.
Everyone got access to the same powerful AI. The inequality came from what happened in the first few minutes. This isn't a failure of the market. It's the market working perfectly — just without the friction that used to keep it from reaching its logical endpoint. Today, human limitations are what keep markets competitive. Remove those limitations, and competition consumes itself.