Red Queen Effect vs Exponentials

Comparison

The Red Queen Effect and Exponentials are two of the most important mental models for understanding the modern technology economy — and they are locked in a fascinating tension. One describes a world where furious effort yields merely survival; the other describes a world where compounding gains produce transformative abundance. In 2026, as AI labs spend billions per training run and inference costs fall 90%+ year over year, both dynamics are playing out simultaneously across the same industries.

The distinction matters because the two frameworks lead to opposite strategic conclusions. Exponential thinking says invest early, ride the curve, and reap compounding returns. Red Queen thinking says those returns are illusory — your competitors are on the same curve, so your exponential investment buys you nothing more than continued relevance. Understanding when each model applies is the difference between building durable advantage and pouring capital into a treadmill.

The AI arms race of 2025–2026 — with OpenAI, Anthropic, Google DeepMind, and DeepSeek all racing to release frontier models — is perhaps the purest real-world example of both forces operating at once. Capabilities are growing exponentially. Competitive position is a Red Queen treadmill. The question is which lens to apply, and when.

Feature Comparison

DimensionThe Red Queen EffectExponentials
Core claimContinuous effort is required just to maintain relative positionConsistent percentage growth compounds into transformative absolute gains
OriginEvolutionary biology (Leigh Van Valen, 1973); Lewis Carroll's Through the Looking-GlassMathematics and technology forecasting (Moore's Law, Wright's Law, Diamandis's Six Ds)
Frame of referenceRelative — your position depends on what competitors doAbsolute — the curve compounds regardless of rivals
Strategic implicationInvest to survive; standing still means falling behindInvest early to ride compounding returns before disruption hits
Attitude toward investmentSkeptical — investment may buy only parity, not advantageOptimistic — investment compounds and creates asymmetric advantage
Predicts winners?No — predicts that most participants merely survive or are eliminatedYes — predicts that early movers on the curve capture disproportionate value
AI lab spending (2025–2026)Explains why labs spend billions per training run yet gain only temporary leadsExplains why each model generation is dramatically more capable than the last
Inference cost dynamicsCost reductions are competed away as all providers pass savings to usersCost per token falls ~92% since 2023, following demonetization curves
Timescale focusShort to medium term — competitive cycles of months to yearsMedium to long term — compounding over years to decades
Failure modeExhaustion — participants burn out or run out of capitalDeception — the flat early part of the curve fools observers into dismissing the technology
Applicability beyond techBiology, arms races, regulatory compliance, any zero-sum or relative-status competitionAny information-based process: energy, genomics, computing, communications

Detailed Analysis

Relative vs. Absolute Gains: The Core Tension

The fundamental difference between these two models is the frame of reference. Exponentials measure absolute progress — the cost of sequencing a genome, the FLOPS per dollar of compute, the capability of a foundation model. By any absolute measure, AI in 2026 is staggeringly more capable than AI in 2023. The Red Queen Effect measures relative position — whether your model is better than the competitor's model that launched last month. By this relative measure, most AI labs are roughly where they started: scrambling to keep up.

Zhang and Zhang's 2026 paper on Digital Intelligence Capital formalizes this tension as "endogenous depreciation." A foundation model's value depreciates not on a fixed schedule but on a schedule set by competitors' R&D velocity. The model improves exponentially in absolute terms while depreciating in relative terms — both statements are simultaneously true. The strategic question is which one matters more for your particular decision.

The AI Arms Race: Both Models in Action

The foundation model race between OpenAI, Anthropic, Google DeepMind, Meta, and DeepSeek is the clearest case study of both dynamics operating simultaneously. On the exponential side: model capabilities have been doubling successful task lengths every five months as of mid-2025, generative AI revenue grew roughly 230% in 2025 to $60 billion, and the AI agents market is growing at a 44% CAGR. The technology is on a classic exponential trajectory through the disruption phase of Diamandis's Six Ds.

On the Red Queen side: no lab has sustained a durable lead for more than a few months. Each frontier model release — GPT-5, Claude 4, Gemini 2.5 — resets the competitive landscape. Labs are spending billions per training run, and 85% of enterprises increased AI investment in 2025 with 91% planning to increase again in 2026. This is classic Red Queen dynamics: exponentially increasing investment to maintain roughly the same competitive position.

The U.S.–China AI rivalry intensifies this further. With Nvidia's market share in China's AI GPU sector collapsing from 95% to near zero under export restrictions, Chinese firms like Biren Technology are racing to build alternatives. Both nations are running the Red Queen's race at the geopolitical level while the underlying technology follows exponential curves.

When Exponentials Win: Platform Layers and Infrastructure

Exponential thinking is the correct model when you are building infrastructure that benefits from compounding network effects or cost curves — and where competitive dynamics are muted by high switching costs, standards lock-in, or natural monopoly characteristics. Huang's Law (GPU performance doubling faster than Moore's Law) benefits Nvidia more than its competitors precisely because the CUDA ecosystem creates switching costs that dampen Red Queen dynamics.

Similarly, Wright's Law cost curves in solar energy and battery storage have created durable advantages for scale leaders because cumulative production volume — not just R&D spending — drives cost reduction. The exponential model dominates when advantage accrues to the curve itself, not just to the latest innovation cycle.

When the Red Queen Wins: Application and Model Layers

Red Queen thinking is the correct model at the application and model layers — anywhere the competitive frontier resets frequently and switching costs are low. The foundation model layer is the starkest example: a state-of-the-art model from six months ago is now merely adequate. But the same dynamic applies to AI-powered applications, autonomous agents, and enterprise AI tooling, where the pace of innovation means that any feature advantage is temporary.

The agentic AI wave of 2026 amplifies Red Queen pressure. As multi-agent systems become the dominant paradigm for enterprise AI, the rate of competitive iteration accelerates further — agents can be updated, retrained, and redeployed far faster than traditional software, compressing the cycle time of the Red Queen's race.

Demonetization: Where the Models Converge

The most interesting intersection of these two frameworks is demonetization — the exponential collapse in cost that strips revenue from incumbent business models. Demonetization is an exponential phenomenon (costs fall on predictable curves), but it triggers Red Queen dynamics in the markets it disrupts. When AI inference costs fall 92%, every company that built a margin structure around expensive inference must race to restructure — not because they got worse, but because the cost floor moved beneath them.

This convergence is visible across the virtual economy. Content creation, customer service, code generation, and data analysis are all being demonetized by AI. The companies providing these services are caught in a Red Queen race driven by exponential cost reduction in the underlying technology. Understanding both models simultaneously — the exponential curve driving demonetization and the Red Queen race it triggers — is essential for navigating the 2026 economy.

Strategic Synthesis: Running Smart on the Treadmill

The most sophisticated strategic thinkers in 2026 are those who recognize that both models are always operating and who choose their investments accordingly. The playbook: invest in exponential infrastructure layers where compounding advantages are durable (data flywheels, ecosystem lock-in, cumulative cost curves), while treating Red Queen layers as arenas where you must invest to survive but should not expect to build lasting competitive moats.

This is why the smartest AI companies are increasingly competing not on model quality alone (Red Queen) but on distribution, ecosystem, and developer tooling (exponential). Anthropic's focus on safety and enterprise trust, OpenAI's platform play, and Meta's open-source ecosystem strategy are all attempts to escape the Red Queen at the model layer by building exponential advantages at adjacent layers.

Best For

Forecasting AI capability growth

Exponentials

Absolute capability growth follows exponential curves. Use exponential models (doubling times, S-curves) to predict what AI will be able to do in 2–5 years.

Predicting which AI lab will lead

The Red Queen Effect

No lab has sustained a durable lead. Red Queen dynamics explain why competitive position at the model layer resets every few months.

Enterprise AI investment planning

Both — depends on layer

Use exponential models for infrastructure investments with compounding returns. Use Red Queen thinking for application-layer bets where advantage is temporary.

Understanding AI cost trajectories

Exponentials

Inference costs, training costs, and hardware costs all follow predictable exponential decline curves. Wright's Law and Huang's Law provide quantitative frameworks.

Competitive strategy for AI startups

The Red Queen Effect

Startups building on commodity model APIs face constant competitive reset. Red Queen thinking correctly predicts that feature advantages erode quickly and continuous innovation is table stakes.

Deciding when to adopt a new technology

Exponentials

The Six Ds framework — especially the deception and disruption phases — is the best model for timing technology adoption. Wait too long in the deceptive phase and disruption catches you off guard.

Explaining industry consolidation and burnout

The Red Queen Effect

When companies burn through capital and talent without gaining durable advantage, Red Queen dynamics are the explanation. The treadmill exhausts participants even as absolute capability grows.

Long-term macroeconomic modeling

Exponentials

GDP growth, productivity gains, and technological deflation operate on exponential timescales where absolute progress dominates over competitive position.

The Bottom Line

These are not competing theories — they are complementary lenses that apply at different layers and timescales. Exponentials describe what is happening to the technology: capabilities compound, costs collapse, and industries are disrupted on predictable curves. The Red Queen Effect describes what is happening to the competitors riding those curves: they run faster and faster just to maintain position, burning capital and talent in an arms race with no permanent winner.

For strategic decision-making in 2026, the key insight is layer separation. At the infrastructure layer — chips, cloud, energy, data — exponential advantages compound and can be durable. Invest here with exponential conviction. At the model and application layer — where competitive resets happen quarterly and switching costs are low — Red Queen dynamics dominate. Invest here for survival, not for moats. The companies winning the current AI era (Nvidia in chips, hyperscalers in cloud) are those whose advantages sit on exponential infrastructure layers rather than the Red Queen treadmill of the model frontier.

If you take away one thing: exponentials tell you where the world is going; the Red Queen tells you how painful the journey will be for the companies getting it there. Both are essential. Neither alone is sufficient.

Further Reading