Panel
Balderton, Altimeter, TVG, Coatue, Alpha & Harmonic: Investment in AI Who Gets Funded and Why
- The AI value distribution is expected to invert over an eight-year timeline to favor applications over semiconductors and infrastructure, though the speed and magnitude of this shift remain uncertain.
- Scarcity in semiconductor and infrastructure layers is projected to persist for the foreseeable future, as application spending continues until optimization enables lower power-density and cost configurations.
- Significant reductions in AI application costs through innovations like on-device inference are anticipated to take considerable time to materialize despite the application layer representing the largest total addressable market.
- Data center capacity constrained by power limitations and long transmission architecture queues in the US and Europe is identified as the primary bottleneck for AI build-outs.
- A secondary infrastructure bottleneck will emerge as high rack-density NVIDIA and AMD SKUs render older data centers obsolete, necessitating a new cycle of capital expenditure.
- Gross margins for AI companies are expected to converge toward SaaS-like levels, though likely falling short of the 80% to 90% efficiency of multi-tenant SaaS due to persistent high-end GPU costs.
- Margin improvements will be driven by the emergence of "Generation N-1 models" offering cheaper inference for products that do not require frontier model capabilities.
- The venture capital environment is shifting toward earlier investments with larger check sizes, while exit timelines are predicted to extend significantly due to a thinner IPO market.
- European companies are not expected to face capital access limitations, though growth funding will likely originate from a mix of US and European sources.
- While generative tools will reduce costs for prototypes, production-ready enterprise software costs will not approach zero due to ongoing requirements for compliance, architecture design, and system integration.
- The industry will prioritize the development of exceptional product and design managers as the cost of coding declines and consumer preference remains focused on product quality.
- Software engineering thresholds will rise for both entry-level and top-tier roles, increasing overall workforce capability and GDP while potentially elevating the relative value of exceptional engineers.
- The dispersion between exceptional and average engineers is forecast to widen as base-level software creation becomes cheaper while premiums for scalable, differentiated products increase.
- Deep domain expertise combined with technology skills will become a critical founder differentiator, as rapid market shifts prevent the ability to "learn the problem" during development.
- Being the first mover in greenfield markets is deemed essential for speed and barrier creation, whereas late entrants in brownfield markets will face significant difficulties in replacing incumbent platforms.
- Success in enterprise AI will require a combination of rapid technical execution and intensive change management to assist customers with use cases and workflow alterations.
- "AI-first" companies that build businesses rather than traditional products are expected to emerge as the dominant future enterprises.
- Apple faces vulnerability to disruption and capital poaching due to perceived delays in building AI infrastructure and a lack of recent product innovation.
- Google's core business model is at major risk as the traditional search paradigm ends and users increasingly rely on chat interfaces for information retrieval.
- Apple's vulnerability is further attributed to regulatory challenges and stagnating innovation in legacy products, though this view is noted with uncertainty.