STRATEGYMonths to result88% confidence

The Generative AI Winners Window

In exponential platform waves, slope locks in winners before observers notice the gap

Problem it solves

Identifying why the current AI allocation period is structurally irreversible

Best for

Investors and operators timing infrastructure bets on AI platforms during the current exponential growth window

Not ideal for

Picking individual AI application winners — Schmidt is explicit that which apps solve which problems remains unknown

Overview

Why this framework exists

Schmidt articulates a race-to-slope model he observed across internet, mobile, and now generative AI: once a platform system achieves a quadrupling growth rate (approximately every six months), the capital required for any challenger to match it at any future point grows exponentially. The gap widens even as absolute spending increases — meaning late entrants face a structurally impossible catch-up problem regardless of resources.

The core insight is that slope, not size, determines competitive moats in platform markets. Investors and strategists who wait for the winner to be obvious are waiting for the moment the window has already closed. Schmidt places the generative AI winner-determination window at six to twelve months from his late-2024 recording — meaning the slope-locking dynamic is active in the current period.

Schmidt grounds this in Google's own experience: Google Video was technically capable and better-resourced than YouTube, but YouTube operated without Google's internal constraints and moved faster. In platform transitions, speed and freedom from incumbent rules matter more than resource advantage. This has direct implications for unconstrained fast-movers competing against regulated incumbents.

Core principles

5 total
  1. Slope (growth rate) determines competitive moats in platform markets, not absolute size or resource advantage
  2. Once a platform achieves quadrupling velocity, the catch-up cost for challengers grows exponentially — the window to enter closes permanently
  3. At platform transition moments, freedom from incumbent constraints outweighs resource advantage — unconstrained fast-movers beat better-funded incumbents
  4. Every successful future company will embed AI compute in its stack — the platforms that establish inference relationships earliest accumulate structural switching costs
  5. The winner-determination window in generative AI is narrow and observable in real time — it is not a distant future event

Steps

4 steps
  1. Identify the current growth slope of the target platform
    Measure the platform's growth rate over the last two to three periods. Schmidt's threshold is approximately 4x every six months. If a platform is near or above this slope, the winner-determination window is active.
    Pro tipFocus on inference volume and API dependency growth, not headline revenue — infrastructure usage locks in before financials reflect it.
  2. Map constraint asymmetries between incumbents and challengers
    Identify which competitors operate under regulatory, reputational, or organisational constraints that slow their iteration speed. Unconstrained fast-movers in the same market as constrained incumbents have structural advantages at transition moments, even with smaller resource bases.
    WarningDo not conflate resource size with competitive position — YouTube had fewer resources than Google Video and still won.
  3. Assess switching cost accumulation for early infrastructure entrants
    Determine whether early platform entrants are accumulating switching costs: user trust, fine-tuning data, API dependency. These compound over time and become structural moats before they are visible to external observers.
    Pro tipInfrastructure switching costs (compute relationships, model fine-tuning, workflow integration) are stickier than consumer app switching costs.
  4. Set a decision threshold tied to the observable window
    Schmidt's framework implies a time-bounded decision: positions established before the slope locks in compound longest. Establish a clear trigger (e.g., the agent concatenation wave arriving, or inference volume milestones) that signals the window is closing rather than waiting for consensus confirmation.
    WarningWaiting for the winner to be obvious is waiting for the window to close — the framework specifically predicts that late entrants face impossible catch-up economics.

Checklist

Saved in your browser

Examples

2 cases
YouTube vs. Google Video

Google Video was technically capable, better resourced, and operated by the dominant internet company of the era. YouTube was an externally built, under-resourced product that operated without Google's internal rules and constraints. Schmidt uses this to demonstrate that at platform transition moments, constraint-freedom and iteration speed beat resource advantage.

OutcomeGoogle acquired YouTube for $1.65B in 2006 — effectively buying the platform winner that its own better-resourced product had failed to become.
Generative AI slope-lock in 2024–2025

Schmidt records in late 2024 that generative AI platform winners are being determined in the next six to twelve months. He identifies this as the decisive window in which slope (not size) is locking in — after which the capital required for challengers to catch up becomes exponentially prohibitive.

OutcomeInfrastructure positions established during this window (compute platforms, inference APIs, model fine-tuning relationships) will compound longest — the framework predicts these early entrants accumulate structural moats before they are visible to consensus observers.

Common mistakes

3 traps
Treating AI as a 'wait and see' opportunity
Schmidt explicitly identifies this as the structural error. Investors who wait for the winner to be obvious are waiting for the moment the slope has already locked — after which catch-up costs are exponential, not linear.
Conflating resource advantage with competitive position at transition moments
Google Video had more resources than YouTube and still lost. At platform transition points, freedom from incumbent constraints and iteration speed outweigh funding. Defaulting to the best-resourced player as the likely winner is wrong at transition moments.
Focusing on application winners rather than infrastructure positions
Schmidt is explicit that which applications solve which problems is unknown. The winners who can be identified are infrastructure platforms — the layer every successful company will need regardless of which app wins.

Origin story

How this framework came to be

Schmidt developed this framework from direct operational experience scaling Google through multiple platform transitions and from observing the internet, mobile, and now AI waves as both operator and investor. The YouTube-versus-Google-Video example is drawn from his tenure as Google CEO, where he watched a less-resourced, externally-built competitor outmanoeuvre Google's own product in the same market.

He refined the slope-locking concept through his post-Google work advising the US Department of Defense and co-authoring books on AI with Henry Kissinger, where platform dynamics intersect with national security timelines. The framework is explicitly predictive: Schmidt uses it to argue that the current six-to-twelve month window (from late 2024) represents the last period in which infrastructure positions can be established before the growth slope becomes self-reinforcing.

Source

Traced to primary
Source · PODCAST
Ex Google CEO: AI Can Create Deadly Viruses! If We See This, We Must Turn Off AI!
Eric Schmidt · 2024
Open source →

Related frameworks

Browse all Strategy →