Three years ago, Sequoia partner David Cahn was one of the first to put a hard number on Silicon Valley’s enormous AI infrastructure bet. In 2023, he calculated that $200 billion in revenue would be needed to justify Nvidia’s $50 billion in GPU revenue and the associated data center costs. Today, that figure has ballooned. Cahn now estimates AI infrastructure spending will hit $1.5 trillion by 2026, requiring the industry to earn $3 trillion to justify the investment.
Cahn’s updated analysis accounts for the hyperscaling of the past three years and notes that his estimate is likely conservative. Rising costs for memory, exotic chips, and inference-specific hardware are pushing the required revenue per gigawatt of capital expenditure sharply higher.
Also read: Elon Musk calls Anthropic the ‘leader in AI’ after signing $40 billion SpaceX deal
The revenue gap: OpenAI and Anthropic aren’t enough
On the revenue side of the ledger, the numbers remain far from closing the gap. Anthropic is thought to have reached $60 billion in annualized revenue, while OpenAI reportedly earned $13 billion in 2025 and has since claimed a $20 billion annualized run rate. Even combined, these figures represent a fraction of the $3 trillion target.
The core question remains: Will AI products and services generate enough demand to fill the gap, or will a significant portion of the infrastructure investment fail to pay off?
Also read: OpenAI unveils GPT-5.6 family: Sol, Terra, Luna target coding, cybersecurity, and cost savings
Apollo’s warning: A market risk that goes beyond tech
Torsten Slok, chief economist at Apollo Global Management, has flagged a specific risk in a recent note. The hyperscalers—Google, Meta, Microsoft, and Amazon—are all projecting massive accelerations in their free cash flow by 2028, banking on a payback from their chip purchases. Slok questions what happens if those projections fall short.
He points to two emerging trends that could undermine the hyperscalers’ financial models. First, more organizations are turning to cheaper open-weight models, often from Chinese developers, rather than those built by frontier labs like OpenAI. Second, token prices are falling. OpenAI’s latest model, per CEO Sam Altman, is 54% more token-efficient on coding tasks. While that benefits users concerned about agent costs, it may hurt companies building token factories if overall token usage doesn’t increase proportionally.
“With so much riding on so few names,” Slok writes, “a slower payoff wouldn’t just be a sector problem, it would risk tipping the economy into recession and the S&P 500 into a correction.”
What this means for the broader market
The concentration of AI spending among a handful of hyperscalers means that any shortfall in their expected returns could have outsized effects on the broader economy. Slok’s analysis suggests that the current infrastructure buildout is not just a tech-sector story but a macroeconomic one with potential systemic implications.
For now, the industry is watching to see whether the demand for AI services can grow fast enough to justify the spending. As Cahn’s challenge to entrepreneurs suggests, the onus is on product builders to create the applications that will drive usage—and revenue—at the scale required.
Frequently Asked Questions
How did David Cahn arrive at the $3 trillion revenue figure?
Cahn started with Nvidia’s GPU revenue and added the costs of operating data centers and operator margins. He now calculates that the AI industry must earn $3 trillion to pay back the $1.5 trillion in infrastructure spending projected for 2026.
What is Torsten Slok’s main concern about AI spending?
Slok warns that if hyperscalers like Google, Meta, Microsoft, and Amazon don’t achieve their projected free-cash flow growth by 2028, the market reaction could be severe, potentially tipping the economy into recession and the S&P 500 into a correction.
Why might AI companies struggle to generate the required revenue?
Token prices are falling due to competition from cheaper open-weight models and improved efficiency from frontier labs like OpenAI, which means users may not increase their overall token usage enough to offset the price drops.
What is the ‘hyperscaler’ bet on future cash flow?
Google, Meta, Microsoft, and Amazon are all projecting massive accelerations in their free cash flow by 2028, expecting the payback from their massive chip and data center investments to materialize by then.
Be the first to comment