PwC projects $30tn data-centre spend by 2050; Anthropic plans $518bn outlay, prospectus shows
PwC forecasts cumulative global data‑centre investment could reach $30 trillion by 2050; Anthropic’s IPO prospectus reveals plans for $518 billion in spending—over 100× its projected 2025 revenue—while tech leaders met President Trump at the White House.
AI-assisted synthesis · Based on 1 independent report from 2 outletsHow it's made
What happened
Key facts
- PwC projects cumulative global spending on data centres could top $30 trillion by 2050.
- Anthropic's planned $518 billion spend is described in the prospectus as more than 100 times its 2025 revenue.
- PwC said that this spending "dwarfs" what was spent in the railroad or dotcom booms after adjusting for inflation.
- The article says AI investment currently exceeds historical capital flows into railways and the internet.
- The article was written by Stephen Eisenhammer and datelined London on Oct 3, 2026.
The brief
PwC projects cumulative global spending on data centres could top $30 trillion by 2050, and Anthropic's IPO prospectus says the company plans to spend $518 billion, over 100 times its 2025 revenue. The article reports backers say AI will be more transformational than the steam engine and notes tech executives gathered to meet U.S. President Donald Trump at the White House.
Why it matters
- If PwC’s $30 trillion data‑centre forecast and Anthropic’s disclosed $518 billion build‑out materialize, they will redirect huge amounts of capital into power, cooling, servers and chips, driving a sustained surge in demand for semiconductors, renewable energy, and specialized construction — raising costs and bottlenecks for other tech sectors and manufacturers. Large cloud and AI firms, plus the countries that host their facilities, will gain outsized strategic leverage through control of compute capacity and data flows, while local communities face strains on electricity grids, land use and jobs tied to construction and operations. Regulators and investors will need to grapple with concentration risks, environmental impacts from massive energy consumption, and the geopolitical implications of where that infrastructure is built.
What to watch next
- Watch for Anthropic’s next SEC filings and investor roadshow disclosures: expect updated capital-raise plans, revised capex schedules or anchoring anchor-investor commitments to appear in the company’s quarterly filings or an amended S-1 within the next 30–60 days.
- Track capital-market moves by major data-centre owners (Equinix, Digital Realty) for new debt or equity offerings and announced land/site acquisitions — these companies typically announce financing or large campus purchases; look for press releases or 8-Ks over the next 4–12 weeks.
- Monitor announcements from hyperscalers (Amazon/AWS, Microsoft Azure, Google Cloud) about new buildouts, long‑term power purchase agreements, or multi‑GW procurement deals; expect contract awards or region expansion plans to be revealed in quarterly earnings calls or sustainability reports over the next 1–3 quarters.
- Follow semiconductor and power-equipment firms (NVIDIA, TSMC, Siemens Energy) for capacity-expansion capex plans and fab/power-plant timelines tied to data‑centre demand — watch upcoming investor days and capital‑expenditure guidance updates in the next 3–6 months.
- Watch regulators and major institutional backers (US SEC, UK FCA, large sovereign wealth funds) for scrutiny or conditions tied to massive AI/data‑centre investing—look for policy statements, consultation papers, or investment-committee disclosures over the coming 60–90 days.
How it's being covered
Coverage analysisShow details2 source articles · 2 unique outlets · 1 distinct reportCoverage mix: Center Only2 Center Where the outlets are based: Asia 100%
- Coverage is currently limited to 2 outlets.
One source is from a wire service (AP, Reuters, AFP or Bloomberg), whose reporting other outlets often republish.
How outlets' accounts compare
Original sources
2 articles from 2 outletsEvery story links to the reporting it was built from. How we choose sources
Spotted a factual error, a broken link or a misleading summary? Tell us.