Estimated Global Compute Distribution and Capacity

This appendix documents the global distribution of computing power as of Q4 2024, with particular focus on AI accelerators and general-purpose computing hardware. All computing power is measured in floating point operations per second (FLOP/s), enabling direct comparison between different types of computing devices.

Methodology

Our analysis tracks two primary categories of computing power:

1. Specialized AI Hardware: Data center GPUs (e.g., NVIDIA H100s, AMD MI300X), TPUs, and custom ASICs (e.g., Microsoft MAIA, Meta MTIA)
2. General Purpose Computing: Consumer devices including smartphones, personal computers, and gaming consoles

The analysis covers 2023-2024, with several key methodological considerations:

Time Period Normalization

Many source datasets span 2022-2024. For these three-year totals, we estimate the 2023-2024 portion as 80% of the total, reflecting:

• Accelerating deployment rates through the period
• Increased manufacturing capacity in later years
• Reported quarterly growth patterns (Mann 2024)

Infrastructure Accounting

To account for infrastructure deployed before our analysis period that remains in active use, we apply a 1.3× multiplier to new deployments. This multiplier derives from:

• Historical hardware retirement rates (Epoch AI 2024)
• Typical infrastructure lifecycle lengths
• Reported infrastructure retention patterns (Mann 2024)

Consumer Device Calculations

For devices like smartphones, we calculate:

$$\text{Total Capacity} = \text{Quarterly Sales} \times \text{Number of Quarters} \times \text{Device Performance}$$

Key assumptions include:

• 275M average quarterly smartphone sales (Ming 2024)
• 8 quarters (covering 2023-2024)
• 80/20 split between iPhone/Android market share
• Device performance from manufacturer specifications

Results

Hardware Specifications

We begin by documenting the performance characteristics of key computing hardware. Table 6.1 shows peak theoretical performance for major AI accelerators and consumer devices.

Table 6.1: Hardware Specifications (2023-2024)

Estimated New Hardware Deployments

Table 6.2 documents NVIDIA’s GPU shipments, which represent the majority of new AI computing capacity deployed during this period.

These shipments are distributed across major cloud providers and AI companies. Table 6.3 details how this computing capacity is allocated, including both NVIDIA GPUs and custom AI accelerators

Table 6.2: NVIDIA GPU Shipments (2023-2024)

Estimated Base Computing Capacity

To establish total worldwide computing capacity, we first document the base capacity that existed at the start of our analysis period (Q1 2023).

Estimated Global Computing Distribution

Finally, Table 6.6 presents the complete worldwide distribution of computing capacity as of Q4 2024, combining base capacity, new deployments, and applying our temporal adjustment factors.

Analysis and Implications

Distribution of Computing Power

The results reveal several key insights about global computing power distribution:

1. Cloud Concentration: The four largest cloud/AI providers (Meta, Microsoft/OpenAI, Google/DeepMind, and Amazon/Anthropic) collectively control 15.9% of global computing power. This represents a significant concentration of high-performance computing resources.
2. Consumer Device Significance: Despite lower per-unit performance, consumer devices collectively represent 32.83% of global computing power, with smartphones alone accounting for 23.23%. This highlights the massive scale of deployed consumer hardware.
3. Legacy Infrastructure: Pre-2023 and other cloud infrastructure remains significant, representing 51.28% of total capacity. This suggests substantial inertia in computing infrastructure deployment.

Growth Patterns

The data reveals several notable trends in computing power growth:

• Accelerating Deployment: New capacity added in 2023-2024 ($2.82 \times 10^{22}$ FLOP/s) is approximately 7 times larger than the Q1 2023 base capacity ($3.98 \times 10^{21}$ FLOP/s).

Table 6.3: Estimated Major Cloud Provider Deployments (2022-2024)

Table 6.4: Base Computing Capacity (Q1 2023)

Table 6.5: Estimated New Computing Capacity (Q1 2023 - Q4 2024)

Table 6.6: Estimated Total Worldwide Computing Capacity (Q4 2024). Note that these calculations exclude cloud CPU compute.

• Custom Hardware: Major cloud providers are increasingly deploying custom AI accelerators (MAIA, MTIA, Inferentia) alongside NVIDIA GPUs, though these still represent a minority of total capacity.
• Consumer Updates: The rapid replacement cycle of consumer devices (particularly smartphones) means this sector maintains a significant share of total computing power despite lower per-unit performance.

Methodological Limitations

Several factors affect the precision of these estimates:

1. Utilization Rates: Our analysis uses peak theoretical performance. Actual achieved performance varies significantly based on:
   • Workload characteristics
   • Cooling and power constraints
   • Network and memory bottlenecks
2. Temporal Distribution: The 80% factor used for 2023-2024 portion of 2022-2024 deployments is a rough estimate based on the degree to which NVIDIA sales and chip FLOP capacity have greatly increased in the past two years (Macrotrends LLC 2024).
3. Consumer Device Usage: Not all consumer devices are actively used for computation at any given time, potentially overestimating their contribution to global computing power.