The Great Acceleration: AI, Macroeconomics, and the Future of Innovation

The Great Acceleration and Macroeconomic Transformation

• AI is serving as an accelerant for every technology currently under study, leading to what is described as a "great acceleration."

• This acceleration is in its initial stages and will impact several domains:

  • Personal life and lifestyles.

  • Business growth trajectories.

  • Macroeconomic growth for the global economy.

• Underlying rates of macroeconomic growth are expected to transform over the course of the current and the next business cycle.

• Projections suggest that annual growth rates will exceed $7\%$.

• This transformation will play out in unexpected and interesting ways, altering the state of play technologically and economically.

Historic Investment Cycles and AI Software

• A visual analysis of capital investment cycles in the United States since the $\text{1800s}$ reveals that the current investment in AI software is of historic proportions.

• The combination of investment in AI software, specifically agents designed to assist enterprises and governments, is expected to collectively exceed the investment put into the railroads in the late $\text{1800s}$.

• Investment sectors catalyzed by AI growth include:

  • Robotaxis and autonomous transport.

  • Drones.

  • Genetics and health medicine (Multiomics).

• These technologies are viewed as global catalysts that will drive the growth of the entire global economy.

The Five Major Innovation Platforms

• The technological world is currently being transformed by five major innovation platforms, which are uniquely positioned for growth over the next $10$ to $20$ years:

  1. Artificial Intelligence (AI).

  2. Public Blockchains.

  3. Multiomics (Advanced genetics and health research).

  4. Robotics.

  5. Energy Storage.

• These platforms are highly interdependent and are currently converging.

• The scale of this convergence is modeled by examining how much various technologies depend on neural networks.

Infrastructure and the Learning Curve of Intelligence

• The usable market for advanced technology is expanding as critical developments occur in space and terrestrial data centers.

• In the space sector (e.g., SpaceX), the learning curve is measured on cumulative capacity rather than just the number of orbits.

• There appears to be no "decay" to the demand/utility curve for intelligence; the global economy is predicted to require trillions of dollars in intelligence to continue transforming.

• Data center investment has exploded since $2012$:

  • Spending on data center systems has moved from approximately $\$200,000,000$ per year per unit to a level that is currently "exploding."

  • The market valuation of major companies has grown from approximately $\$7,000,000,000,000$ to $\$28,000,000,000,000$ in a very short period (roughly four weeks for some recent valuation jumps).

• Current developments mirror the early stages of the Internet in $1996$, just as the dot-com era was beginning to take shape.

AI Agents and Productivity Gains

• AI agents are crossing a "commercially useful threshold," meaning they can capture meaningful tasks and operate independently for extended periods.

• Productivity metrics for AI agents:

  • The duration for which an AI can assist with human tasks has increased from approximately $0.5$ hours to between $12$ and $16$ hours over a period of less than five months.

  • This represents a $20 \times$ increase in the ability to successfully complete and return work.

• Economic impact of agents:

  • Agents are expected to facilitate $\$8,000,000,000,000$ in purchases by $2030$.

  • By $2030$, close to $\$1,000,000,000,000$ in revenue could flow through AI engines on the consumer-facing side.

Impact on Advertising and Business Models

• The landscape of advertising will change significantly due to "synthetic users" (bots).

• Traditional advertising wisdom states that $50\%$ of spend is wasted; in a world of synthetic users where only $1$ in $20$ users is human, $95\%$ of traditional advertising might be wasted, but the remaining $5\%$ becomes extremely valuable.

• Businesses will need to decide whether to advertise to specialized bots or the underlying human user.

• Customer acquisition channels, such as "free trials," are at risk of disruption. In the past, humans spent significant time switching providers for small credits (e.g., $\$100$ long-distance credit). With AI, the cost of posing as a new customer and managing hundreds of accounts collapses to near zero, potentially making traditional free trial channels ineffective and expensive.

Enterprise AI Spending and Labor Metrics

• There is a growing theory that "token spend" (spending on AI processing) may exceed "employee spend" in the near future.

• Among some venture portfolios, it is estimated that $60\%$ to $66\%$ of companies may move toward this model.

• A current effectiveness metric suggests businesses spend $10\%$ of the productive value gained from software on that software. For example, if software provides a $2 \times$ productivity boost to a worker, the business goal is to spend $10\%$ of that worker’s salary on the underlying agentic software.

Autonomous Transportation and Robotaxis

• The adoption of autonomous technology creates an "accelerated path of capital formation."

• Traditional cars are inefficient assets, used only approximately $5\%$ of the time. Robotaxis are expected to be utilized $40\%$ to $50\%$ of the time.

• The "clearing price" for mass adoption of robotaxis in the U.S. is estimated at less than $\$1$ per mile. In contrast, buying a traditional car is equivalent to buying a bundle of miles at approximately $\$0.80$ per mile.

• Key milestones and figures:

  • A meaningful rollout of robotaxis is modeled for $2027$.

  • Tesla production capacity (e.g., Model Y in Austin) is used as a benchmark for how quickly autonomous fleets could be produced compared to smaller-scale distributors like Waymo.

  • The lost utility or wages associated with traditional driving is estimated at $\$10,000,000,000,000$ against a US economy of approximately $\$30,000,000,000,000$ per year.

Humanoid Robotics Difficulty and Potential

• Creating humanoid robots is estimated to be a $20,000,000 \times$ harder problem than creating robotaxis due to the complexity of the environments.

• However, society is approximately $100 \times$ more error-tolerant of a robot (e.g., a robot stubbing its toe is less dangerous than a robotaxi hitting a child).

• When adjusted for error tolerance, the humanoid robot problem is roughly $200,000 \times$ harder than autonomous driving, indicating a need for significant software innovation.

Digital Financial Assets and Blockchain

• Financial assets are being truly digitized, making them digitally transferable.

• Projections suggest a shift from tens of billions in digitized assets to $\$11,000,000,000,000$ in digitized financial assets.

• The global financial ecosystem currently extracts approximately $3\%$ against global wealth annually for transaction and administrative costs (middlemen, lawyers, etc.).

• Digitization through public blockchains could lead to asset appreciation and entrepreneurship opportunities across cryptocurrencies, reaching a valuation just south of $\$1,000,000,000,000$ by $2030$.

Historical Economic Context and Step Changes

• Economic history shows periods of "step changes" in production rates.

• In $1400 \text{ AD}$, the rate of economic growth was very low; today it is around $3\%$.

• Previous transformations (internal combustion, electrification, telephone) changed the underlying growth rate.

• The current convergence of the five innovation platforms (AI, Multiomics, Robotics, Blockchain, Energy Storage) is expected to trigger a new transformation with a compounding real growth rate of $7\%$.

• This era is moving toward "local abundance," where the cost of manufactured goods and intelligence collapses.

Multiomics and Life Extension

• As basic needs are satisfied by abundant wealth and technology, capital will shift toward extending human life.

• The cost of generating genetic information is collapsing.

• Biological data is extremely rich; it is estimated that genetic testing will generate $200 \times 10^{12}$ (200 trillion) tokens of data by $2030$.

• By comparison, a frontier-class AI model is typically trained on approximately $15 \times 10^{12}$ (15 trillion) tokens.

• The market for advanced genetic testing and recurrence blood tests is expected to exceed $\$1,000,000,000$ in the US.

Market Valuation and Disruptive Innovation

• It is estimated that two-thirds of the current equity market valuation represents future growth expectations.

• This is historically consistent with periods like the railroad expansion.

• There is a "creative destruction" aspect to this cycle: nondisruptive, exposed businesses may be "eaten alive" by disruptive forces that enable the next class of business models.