Yahoo Finance

Bret Taylor & Clay Bavor on renting intelligence, owning context

Bret Taylor & Clay Bavor· Co-founders at Sierra
·~17 min·English·Yahoo Finance
AgentsAI CompanyInferenceBusiness StrategyAI Infrastructure
TL;DR

Sierra's Bret Taylor and Clay Bavor argue that as intelligence turns cheap and abundant, every durable advantage — owning context, selling outcomes, earning public consent, matching the right model tier to each job — lives in what you build around the model, not in the model itself.

01The Core Bet

Rent the intelligence, own the context

When super-intelligence is a utility anyone can buy, the only durable advantage is the context a company already owns — its customers, its trust, its history.

we have this saying at Sierra rent the intelligence own context you know and if everyone has access to super intelligence how do you differentiate your business?

— Bret Taylor, Yahoo Finance
Key Insight
It flips the usual fear on its head: the companies that look most exposed to AI disruption — banks and alarm firms with customer relationships built over generations — may benefit most, because that relationship is the irreplaceable asset, and AI only lets them act on it faster.
02The Business Model

Sell outcomes, not tokens

Sierra charges for a resolved case or a closed sale, not for tokens, so the customer buys a result while the messy economics of model choice and inference stay hidden underneath.

companies thinking about buying outcomes as opposed to tokens and in fact in the AI ISP space we we built our whole business model around this notion of selling business outcomes.

— Clay Bavor, Yahoo Finance
Key Insight
Outcome pricing quietly solves the CFO's token-spend panic: if the vendor is paid per result, it — not the customer — now carries the incentive to drive inference cost down, turning a budgeting headache into the seller's optimization problem.
03Why Spend Keeps Rising

Cheaper intelligence, more demand

Falling cost per token does not shrink the AI bill — like cheaper energy, every efficiency gain gets absorbed by new uses, so total compute demand keeps climbing.

every time we've made methods of producing more energy that is cleaner and cheaper we use it all and it creates more demand. Uh it's hard to know when this ends but I think the same will probably be true of intelligence.

— Bret Taylor, Yahoo Finance
Key Insight
This reframes the market's central worry. Rising spend is read as a warning that outcomes aren't keeping up; the founders read the same curve as proof the technology is valuable enough to keep pulling demand — the spend is the signal, not the symptom.
04The Compute Buildout

Think crazy: even data centers in space

Because AI keeps surprising even its builders with how fast it moves and how big the demand is, ideas that sounded absurd two years ago — like putting solar and data centers in space — now look worth exploring.

I have to say it was like putting data centers and solar panels in space. Uh, 24 months ago, you know, I raised an eyebrow.

— Clay Bavor, Yahoo Finance
Key Insight
The willingness to entertain orbital compute is less a forecast than a tell: builders only reach for the absurd option when the ordinary ones start to feel insufficient for the demand they are already seeing.
05The Political Economy

The industry hasn't made its case

As the public pushes back on data centers, the founders admit the AI industry has failed to show ordinary people how the buildout could lower their costs and benefit their communities.

the AI industry has failed at that. Uh you know it's not a popular issue and people aren't saying hey this will lower my tax bill benefit my community and by the way the solution being created will benefit me as an individual.

— Bret Taylor, Yahoo Finance
Key Insight
It is a rare concession from inside the industry that the real bottleneck on compute may be political, not physical: if the public will not grant the buildout, capital and silicon do not settle it — public consent becomes part of the supply chain.
06Constraints as Fuel

Scarcity is the mother of invention

When GPUs, power, and budget become the rate-limiter, that pressure is exactly what produces distilled small models, new architectures, and efficiency gains all the way down to the silicon.

Scarcity is the mother of invention. Necessity is the mother of invention and I think that's right

— Clay Bavor, Yahoo Finance
Key Insight
This is the optimistic mirror of the demand story: unbounded demand pulls, and scarcity pushes. The same shortage that worries investors is what the founders expect to force the next round of cheaper, smaller, more capable models.
07The Market Shape

Intelligence won't be one thing

In a few years intelligence will come in tiers — like storage, trading off price, speed, and capability — and each workload will be matched to the tier it actually needs instead of one model for everything.

I think when we fast forward two or three years, intelligence won't be one thing. It's not going to be just one model. It's going to be something about latency, performance, cost

— Bret Taylor, Yahoo Finance
Key Insight
Because the labs own both the models and the compute, Bret bets they can run more efficiently than a company stitching open weights together on its own — so their edge may be less the single smartest model than the ability to serve the whole price-performance range.
08The Honest Caveat

Be humble about the path

The one thing the founders claim to know for certain is that everyone will be surprised again and again, so the honest stance for an investor is conviction on direction and humility on the path.

if there is one thing that I know for sure, it is that we we are very likely to be surprised again and again and again by the pace the the flavor of breakthroughs in this space.

— Clay Bavor, Yahoo Finance
Key Insight
The advice resolves the whole interview into one trade: bet on the thing that recurs in every section — unbounded demand for intelligence — and refuse to bet on the exact form the next breakthrough takes, because the last three years embarrassed everyone who tried.