AI & ComputeLOT 261122 · 3 min read

Loeb’s $3 Trillion Test for AI’s $5T–$7T Buildout

Third Point says AI’s 2025-2030 buildout needs $2T-$3T in added annual revenue as data-center debt yields rise and Meta’s projected 2026 capex reaches $145 billion.

By Rebecca Stone622 words

Key takeaways

  1. Third Point cites $5T-$7T in forecast hyperscaler and AI-developer spending for 2025-2030.
  2. Its bear-math model requires $2T-$3T in additional annual AI revenue, equal to 33%-50% of the roughly $6T IT market.
  3. Third Point estimates OpenAI and Anthropic generate a little more than $100B annualized combined, nearly 20 times below the lower $2T target.
  4. Goldman Sachs projects more than $400B of hyperscaler debt issuance in 2027 and about $300B of global AI project financing.
  5. Meta's 2026 capital spending is projected at $135B-$145B, with quarterly free cash flow turning negative.
Billionaire Daniel Loeb’s ‘Bear Math’: AI’s $7 Trillion Spending Boom Faces a $3 Trillion Reality Check
Billionaire Daniel Loeb’s ‘Bear Math’: AI’s $7 Trillion Spending Boom Faces a $3 Trillion Reality Check — blickpixel / pixabay

By Rebecca Stone

Industry forecasts put hyperscaler and leading AI-developer spending at $5 trillion to $7 trillion from 2025 through 2030; Third Point estimates that buildout will require $2 trillion to $3 trillion in additional annual revenue.

Billionaire hedge-fund manager Daniel Loeb’s calculation, outlined in Third Point’s third-quarter investor letter, assumes investors demand annual returns in the low teens and that equipment depreciates over time. The spending envelope covers advanced chips, data centers and the electricity infrastructure supporting them.

The lower and upper figures equal 33% and 50%, respectively, of today’s roughly $6 trillion global IT market. Third Point did not present the gap as money already lost.

“The bear math is hard to dismiss,” Loeb wrote.

What does the $3 trillion figure actually mean?

Loeb’s $3 trillion figure is not an existing revenue shortfall. It is the upper end of the additional annual revenue needed to justify projected spending, subject to the required returns and equipment depreciation in his model.

For scale, OpenAI and Anthropic together generate a little more than $100 billion in annualized revenue, according to Third Point. The lower $2 trillion target is nearly 20 times that combined run rate.

Reported figures show the scale of the buildout:

  • U.S. AI spending has crossed $1 trillion.
  • Dell Technologies has a $95 billion AI backlog.
  • Broadcom’s AI chip revenue growth reached 221%, although the source specifies no reporting period for that increase.

Those figures sit against fortunes created for chipmakers and a multiyear technology-stock rally, but neither establishes returns on the next dollar of infrastructure spending.

Why are financing costs changing the AI equation?

Third Point said investment-grade data-center securities that yielded below 5% in January approached 7% by November. Some riskier data-center financings carried double-digit yields, according to the letter.

Goldman Sachs projects that hyperscaler debt issuance will exceed $400 billion in 2027, alongside approximately $300 billion in global AI project financing. Higher borrowing costs can turn otherwise profitable projects into disappointing investments.

Meta’s AI data-center expansion is consuming cash: 2026 capital spending is projected at $135 billion to $145 billion, and quarterly free cash flow is turning negative.

Nvidia benefits immediately when customers purchase GPUs. Meta, Microsoft and Alphabet must eventually produce sufficient revenue or operating savings from those purchases. One company’s revenue becomes another company’s capital expense; both cannot claim the same dollar as profit.

Is Loeb calling for an AI collapse?

Loeb is not calling AI another dot-com bubble. Third Point acknowledges that infrastructure demand remains strong, cloud revenue growth is accelerating, and the addressable market extends well beyond traditional IT budgets.

AI could eventually transform portions of the $60 trillion knowledge-worker economy. That opportunity does not guarantee every infrastructure investment will earn attractive returns, prompting Third Point to reduce its exposure to semiconductor, hyperscaler and data-center-related investments.

The gap between infrastructure suppliers and the companies financing expansion is becoming more important. Broadcom’s $42 billion financing arrangement with Anthropic shows how chip suppliers can take on financial exposure to support customers’ AI ambitions.

Where do returns shift next?\n

The investment test is no longer whether AI works, but whether companies can earn enough from it to justify the cost of the infrastructure supporting it.

Third Point’s investor takeaway favors businesses demonstrating rising free cash flow and measurable returns on AI investments over companies relying on ever-larger spending commitments. Its $3 trillion figure remains a calculated revenue requirement, not a confirmed loss.

As financing costs and customers’ capital burdens rise, competitive advantage should shift from selling more infrastructure to producing measurable free cash flow, even if AI demand continues to expand.

247wallst.com (Source)

Filed under

Share this article:

More from Rebecca Stone

Rebecca Stone

Show full bio

Correspondent covering technology and product news in Semiconductors, chip manufacturing, hardware, AI compute at Chip Dispatch.

1,245 articles

Read also

« Previous article