Where do you see the greatest opportunities for value creation in the AI stack beyond foundational models, and why?
Most see the greatest value in the application layer and tooling above foundational models, though some highlight infrastructure or data-homogenizing layers as key opportunities.
7 answers
Answers


Benedict Evans
Independent analyst
Applications and Tooling
Benedict Evans says the value will move up the stack to applications and tooling, because the lab companies cannot build all the use cases, products and tooling themselves.
Antler Global · Watch on YouTube
Keith Rabois
Managing Director · Khosla Ventures
Own the Whole Workflow
Keith Rabois says the greatest opportunity is to own the whole workflow, and advises building a company the frontier labs can't swallow rather than a product they can.
Linear: A Vertical Software Newsletter · Linear: A Vertical Software Newsletter · Read

David Haber
Board Member · Partnership For New York City
Application Layer
David Haber says he is more bullish on the application layer after going deep in building software, arguing the labs won't eat every vertical and build their own applications.
Fintech Leaders · Fintech Leaders · Read

Matthew Hartman
Managing Partner · Factorial Capital
Infrastructure
Matthew Hartman says much of the opportunity is still in infrastructure, citing developer tools and maintenance, and non-language models like time-series prediction and robotics.
Every · Watch on YouTube

Emily Man
Partner · Primary Venture Partners
Applications Layer
Emily Man says she focuses on the applications layer, the last level of the stack, after chips, infrastructure and models; she notes past tech shifts lay the pipes first before building on top.
Run the Numbers with CJ Gustafson · Watch on YouTube

Itamar Novick
Founder & General Partner · Recursive Ventures
Application Layer
Itamar Novick points to the application layer, saying AI infrastructure is in a bubble and application companies delivering end solutions to customers are the ones that emerge, not LLM firms.
The answer starts at 24:20.
The Tech Trek · Listen on Apple Podcasts

Julien-David Nitlech
Managing Partner · Iris
Data-Homogenizing Layers
Julien-David Nitlech points to data-homogenizing layers, saying everyone will use them to make data homogeneous and operate automations, since current stacks aren’t fit for automation.
The European VC · Watch on YouTube
More Venture Capital questions
View all →- What qualities matter when investing in deep tech startups?11 answers
- What qualities do you seek in founders before investing?15 answers
- What practice helps an investor build trust with founders?7 answers
- How has your firm adapted fund strategy to larger VC funds?10 answers
- How do you keep investments grounded during market exuberance?6 answers















