Systems of agents bring Service-as-Software to life READ MORE


The decade of centicorns: why scaling to $100B is the new normal

Ideas / Points of View

09.05.2025 | By: Joanne Chen, Leo Lu

The most valuable resource known to humanity just got cheap: intelligence.

We’re entering a moment where AI is no longer just a tool for organizing workflows, but is now being deployed as a worker capable of handling tasks that once demanded human brainpower—everything from conducting market research to coordinating complex projects. As costs of compute and inference continue to fall and AI native applications continue to permeate the market, access to intelligence is expanding.

Public discourse has been: how to build a $1B company with one person.

We think this is the wrong discourse to have. The next generation of startups will be built with an entirely new input: abundant intelligence. Like electricity or the internet, when intelligence becomes cheap and prolific, we don’t use less of it—we find exponentially more and new things to do with it.

In this scenario, we shouldn’t limit ourselves to 1:1 human-to-agent substitution for efficiency. We should invite more people to think beyond replacement, where they can harness and guide the agents to experiment with net-new use cases previously unimaginable. Abundant intelligence resets the ceiling on what small teams can achieve.

The right team, using AI, won’t just settle for a billion-dollar company; it will aim for $100B. The limiting factor is no longer human hours; it’s vision and execution. AI won’t constrain ambition. It will accelerate it—fueling bigger ideas, faster growth, and new market categories we couldn’t previously attempt.

Welcome to the age of $100B startups. 

Named after the 19th century English economist William Stanley Jevons, this phenomenon is observed when technological advancements make a resource cheaper to use. If demand for that resource is price-elastic, the lower cost leads to an overall increase in total consumption.

Jevons first noticed this with coal. As steam engines became more efficient and were able to use less coal, overall coal consumption still soared. Why? Cheaper steam power unlocked new uses: new factories, new ways of heating buildings, new kinds of locomotives. Coal wasn’t just used for the same purposes less expensively, it was used more and in more ways because it was no longer price-prohibitive.

We’ve seen this pattern throughout history with other price-elastic resources:

  • Electricity: As electrical power got cheaper and more widespread, energy use skyrocketed. We didn’t stop at using the same lamps more cheaply—we created new product categories. We developed new home appliances, better factories and public transportation, and created batteries to move power beyond the grid.
  • ATMs: ATMs seemed destined to make human bank tellers obsolete. Who would wait in line when you could just use the machine to complete your deposits? But because ATMs lowered the overall cost of operating branches by making them more efficient, banks responded by opening more branches and expanding their offerings and services, creating more demand for “higher-touch” customer service roles.

In each case, making a resource cheaper unleashed an even bigger wave of new demand. Paradoxically, advancements in AI won’t lead to less usage of compute, but to far more. As costs fall, we don’t conserve intelligence—we consume exponentially more.

Intelligence is the new abundant resource, unlocking unprecedented use cases

Human intelligence is slow and limited to our physical limitations. No matter how smart and efficient you are, you still need to occasionally eat and sleep, and you likely have other needs to attend to (social, societal). These limits make human intelligence comparatively expensive. Artificial intelligence, on the other hand, can keep going all day and all night. Not only does it not need to rest, it can work orders of magnitude faster and more efficiently, and be less prone to error.

As technological advancements make AI more reliable over time, it will make sense for businesses to use this “less expensive” source of intelligence to get work done. And having a superabundance of intelligence—artificial and human combined—will unlock unprecedented use cases. As Packy McCormick writes, “The increased supply of intelligence will create more demand for tasks that require intelligence, that we’ll turn gains in intelligence efficiency into ways of doing new things that weren’t previously feasible.”

Essentially, as AI makes thinking cheaper, we’ll find more things to apply that thinking to. Not only will we be able to enhance existing workflows, the intelligence superabundance also means entirely new businesses will emerge that were impossible to consider when human brainpower was the limiting factor.

AI will change how we work within organizations—and startups are primed for this shift

Not only will the intelligence superabundance lead to new use cases, it will also restructure how we work. Human intelligence will still be valuable; we’ll concentrate it on the tasks where our humanity adds value, and work alongside AI to create impact. There are a few key shifts on the horizon:

Converging roles: organizations were siloed to manage and simplify human work.  In an AI native world, we see the blurring of previously distinct roles, where one human can manage a widening scope with the help of AI. For example: 

Operations today is fragmented (BizOps, RevOps, PeopleOps, FinanceOps) and often coordination-heavy. AI removes the coordination tax: data pipelines unify reporting, so no one needs to reconcile metrics manually; forecasting and scenario planning become AI-driven, reducing need for specialized analysts; HR, finance, and ops tasks can be automated (policy drafting, onboarding workflows, compliance checks). One “ops partner” can oversee a broader scope (financial, people, GTM support) because AI manages execution, reporting, and compliance. Tonkean, Campfire, and Maximor are already helping operators do more in this space. 

R&D: Historically split into product (vision/requirements), design (user experience), and engineering (execution). AI allows three roles to converge. AI can generate prototypes, write code, run usability tests, and iterate. The human R&D role shifts to framing the problem, articulating goals, and refining outputs. One “product builder” works across the stack, with AI agents generating wireframes, writing production-grade code, and even simulating customer adoption. For example, LinkedIn just launched a program hiring AI-savvy product builders who do all of the above, and more. Our portfolio companies PlayerZero, OpenStudio, and Kombai are building the platforms that support this emerging role. 

We see GTM roles converging as well. Historically, sales, marketing, and customer success were distinct: marketing generated leads, sales closed them, CS managed retention, all with a focus on acquiring and servicing the customer. In an AI-native world, a single GTM leader with AI can: generate hyper-personalized campaigns (traditionally marketing), run automated outbound and qualify leads (traditionally SDR/AE work), monitor usage data in real time and deploy automated nudges or tailored upsell plays (traditionally customer success). Instead of three teams passing the baton, one person orchestrates the full lifecycle, with AI handling data analysis, personalization, and campaign execution. Our companies Jasper, Regie, Docket, and Arcade are building towards this future.

A changing role for human intelligence: What can’t be done by AI will be owned by humans. Human intelligence will still be important for discerning the big vision and for building trust. We’ll see growth in roles and companies related to accountability and legal responsibility, relationship management, strategic planning and architecture. AI will still be a tool here, but with humans very much steering the ship.

Unsurprisingly, startups have an advantage as we head into the age of companies powered by superintelligence. While incumbent organizations are likely to struggle with such foundational shifts in their structure and processes, new companies can build this way from the ground up. AI-first organizations can move faster, be more nimble, and reap the benefits of AI sooner—leading to a barbell effect in the market.

The barbell effect: an increase in extremes

As many smaller, AI-native startups gain outsize impact in the market, we’ll start to see the “barbelling” of business scale. The distribution of outcomes will stretch to extremes. Fewer companies will capture disproportionate value (and many of them will do it with far fewer people), while mid-sized incumbents will struggle to adapt. The companies in the middle will start to consolidate, driving more volume to both ends of the barbell.

This means we’ll start seeing new trajectories of growth, and we’ll also need to set new markers for success. Instead of the traditional SaaS benchmark of T2D3 growth, going from $1M to $3M to $6M to $9M over a startup’s first few years, these AI-native companies will start around $3M ARR in year one and scale to $12M in year two, $40M in year three, and $100M+ by year four. We believe this Q2T3 trajectory—highlighted in Bessemer’s 2025 analysis of 10 high-growth AI “shooting stars”—will become the new benchmark for startups in the age of AI. What once took a decade in the cloud-SaaS era can now happen in just a few years as intelligence becomes superabundant.

Go big, go AI-native, go now

For a founder, this new era will be liberating: you can scale faster and aim higher with a lean team. But it’s also daunting: if you cling to old growth playbooks or modest goals, you risk underplaying a huge opportunity. The key is to harness the shifts driven by cheap intelligence. Here are five principles to keep in mind:

  1. Design your org chart around AI-first assumptions: Don’t hire traditional roles by default; start with a blank slate. Which functions can be powered by AI from the start? Instead of a big QA team, think about what a team built around automated testing bots might look like. Instead of separate product and engineering (and even product design) divisions, form one “product building” team that does it all. Embrace organizational minimalism: every human you hire should be doing what AI can’t, and coordinating AI to do the rest. This makes your company fast, adaptable, and cost-efficient from day one.
  2. Hire polymaths and architects, not narrow specialists: Optimize for versatile, high-agency talent. A small team of “Swiss army knife” collaborators will out-innovate a larger team of folks who stay in their lane. The people on your team should have a breadth of skills, the ability to wear many hats, and an eagerness to learn and implement new tools. Look for system architects who can envision and build complex systems of AI workflows on their own to achieve feats that used to require whole departments.
  3. Embrace collapsing functions: Tear down the walls between building and selling: blend product, engineering, and GTM through automated product-led growth. Encourage your team to own outcomes, not functions. A feature team might handle its feature’s growth analytics and even write the launch copy with AI partners, while engineers can directly observe user behavior and iterate growth experiments. This collapse speeds up the feedback loop and keeps you hyper-aligned on the mission. Long term, it also makes work more fulfilling. Team members see the direct impact on users rather than tossing tasks over a wall. 
  1. Anticipate future economics and build for tomorrow’s costs: Keep the future in mind when planning your business model and pricing. AI compute costs are on a declining curve. Don’t constrain your vision to current unit economics if you see a path to cost drops that could drastically impact your margins. Conversely, remember that competitors will also have cheaper AI, so future-proof your strategy. Assume commoditization of basic AI capabilities and build revenue streams around network, brand, or proprietary data—things that won’t be commoditized. Run scenarios for if your cost of prediction or text generation fell by 90%. How would you redesign your product or pricing? Plan now to get ahead of the competition, while remaining cautious with burn.

Think in decades: The intelligence superabundance is just beginning. This era rewards audacious goals and a long-term vision; incremental thinking might lead you to sell early or plateau. Make decisions that prioritize scale over short-term profit, choose investors who share your big vision, and hire employees who are excited about rewriting the rules together. Be willing to face the unknown with the conviction you’re on the right path. Tomorrow’s overnight successes will be the teams that started playing the long game today. Will your company be one of the first 10-person $100B companies? Consider this an invitation.

Call for startups

At Foundation, we’re actively seeking founders who are fundamentally reimagining how AI will work alongside humans over the next decade. If you’re building the next $100B company, we’d love to learn more about your vision. Reach out at jchen@foundationcap.com and llu@foundationcap.com.

Published on September 4, 2025
Written by Leo Lu and Joanne Chen

Related Stories