Meta CEO Mark Zuckerberg sets a tone of realism about investments in artificial intelligence, suggesting meaningful AI revenue is still a few years away. “Building the leading AI will also be a larger undertaking than the other experiences we've added to our apps and this is likely going to take several years,” said Zuckerberg.
Nor is that an unreasonable expectation, for Meta, other app suppliers or cloud computing hyperscalers who might literally double their compute capability over the next four years to support AI, as Synergy Research Group suggests will be the case.
As generally is the case, capacity has to be put into place before monetization can scale. And that arguably will prove the case for most AI-related investments: investment and cost will come first; monetization will follow, but not in a linear way.
"Capacity growth will be driven increasingly by the even larger scale of those newly opened data centers, with generative AI technology being a prime reason for that increased scale,” says Synergy.
Globally, Mordor Intelligence has suggested that AI hardware and software spending overall will reach about $310 billion by 2026, with a compound annual growth rate of 38 percent. Precisely how much will be spent by data centers is less clear, but is expected to be substantial.
Year | Processing CapEx (USD Billion) | Storage CapEx (USD Billion) | Source | Discussion |
---|
2021 | 50-70 | 20-30 | Synergy Research Group (2022) | Estimates based on overall data center CapEx growth and industry trends related to AI adoption. |
2022 | 55-75 | 25-35 | Gartner (2023) | Estimates based on data center equipment sales figures and analyst projections for AI hardware growth. |
2023 | 60-80 | 30-40 | IDC (2023) | Forecasts based on hyperscale data center spending surveys and analysis of enterprise AI deployments. |
2024 | 65-85 | 35-45 | Mordor Intelligence (2022) | Projections based on AI hardware market growth and anticipated increase in data center infrastructure spending. |
2025 | 70-90 | 40-50 | Cowen Research (2023) | Analyst estimates based on industry surveys and projections for continued growth in AI workloads and data volumes. |
Capital investments by the four large operators of hyperscale data centers might have a compound annual growth rate of 11 percent to 35 percent between 2021 and 2025, some estimate.
Year | Estimated Hyperscale Data Center CapEx (Processing & Storage) | Source | Discussion |
2021 | $80 Billion - $100 Billion | Synergy Research Group (2022) | This is an estimate for total CapEx on processing and storage in hyperscale data centers, not specifically for AI. |
2022 | $85 Billion - $105 Billion | Synergy Research Group (2023) | Similar to 2021, this represents total CapEx, but a portion will likely be directed towards AI needs. |
2023 | $90 Billion - $115 Billion | Gartner (2023) | Gartner predicts a 6.1% growth in data center IT spending in 2023, with a significant portion likely going towards processing and storage. |
2024 | $95 Billion - $125 Billion | IDC (2023) | IDC forecasts worldwide data center spending to reach $352 billion in 2024, with hyperscale CapEx on processing and storage being a major driver. |
2025 | $100 Billion - $135 Billion | Mordor Intelligence (2022) | Mordor Intelligence predicts a CAGR of 13.4% for the data center hardware market (2020-2027), suggesting continued growth in CapEx. |
Though Meta and others investing heavily in core models will have to manage investor expectations, there's a strong argument to be made that leadership in generative AI models could offer business advantages similar to leadership in established platforms like operating systems, search engines, social media, and e-commerce.
Just as dominant operating systems or search engines have conferred business advantages, leadership in generative AI could position a company as a gatekeeper for a crucial technology. Network effects also matter, as leadership brings usage, which generates more data, leading to better performance and attracting even more users. This creates a self-reinforcing cycle, similar to how dominant social media platforms gain traction.
Leading generative AI models can become platforms for further innovation, creating ecosystems of value as developers build applications and services on top of the AI, just like businesses build apps on dominant operating systems or e-commerce platforms.