Peter Diamandis projects $700 billion

Peter Diamandis’ May 14 post on X took a set of already-public company forecasts and analyst estimates and put them into one headline number: roughly $700 billion of hyperscaler capital spending in 2026. That figure lines up with a Moody’s estimate for six U.S. hyperscalers and with reporting that Big Tech’s AI infrastructure buildout is approaching that level this year. (abc.xyz) The post matters because the underlying company guidance is unusually large even by recent cloud-investment standards. Alphabet told investors on February 4 that its 2026 capital expenditures were expected to be $175 billion to $185 billion, and Meta said on January 28 that its 2026 capital expenditures would be $115 billion to $135 billion. (ferc.gov) Amazon has separately projected about $200 billion in 2026 capital expenditures, according to company and earnings reporting. The second part of Diamandis’ thread tied that spending to the power system. Federal regulators are already working on rules for “large loads” — generally electricity demand above 20 megawatts — as data centers and other big customers push for faster grid connections. (datacenterdynamics.com) FERC said on April 16 that it would act by June 2026 on that rulemaking. ### Where does the $700 billion figure come from? Moody’s said U.S. hyperscaler capital expenditures are expected to rise to about $700 billion in 2026, up from $387 billion in 2025, across Amazon, Microsoft, Meta, Alphabet, Oracle and CoreWeave. That is the clearest published benchmark matching the number in Diamandis’ post. (abc.xyz) Alphabet’s own guidance accounts for a large share of that total. The company said on its February 4 earnings call that 2026 CapEx would be in the range of $175 billion to $185 billion. Meta’s January 28 earnings release gave a similarly explicit range. (ferc.gov) The company said 2026 capital expenditures, including principal payments on finance leases, would be $115 billion to $135 billion, driven by investment in its “Meta Superintelligence Labs efforts and core business.” Meta later raised that range to $125 billion to $145 billion in its April 29 results. (datacenterdynamics.com) Amazon has not posted the $200 billion figure in the snippets reviewed here in the same line-item format as Alphabet and Meta, but Amazon projected about $200 billion in 2026 capital spending, according to CNBC’s February reporting and subsequent coverage tied to its earnings. (abc.xyz) Amazon’s investor site shows its latest earnings call was held on April 29 and its annual shareholder meeting is scheduled for May 20. ### Why are these companies spending so much? Alphabet said the higher spending is meant “to meet customer demand and capitalize on the growing opportunities ahead of us.” The company tied the increase to AI products and cloud demand, with cloud backlog reaching $240 billion at the time of its February earnings call. (investor.atmeta.com) Meta said its spending growth was driven by “increased investment” for its superintelligence efforts and core business. CNBC reported in April that Meta raised its 2026 CapEx range because of higher component pricing and additional data-center costs for future capacity. (cnbc.com) Moody’s said the spending surge is accelerating revenue growth but eroding free cash flow and prompting higher borrowing. That framing came from the ratings agency, not from the companies themselves. ### How does the grid fit into the story? (abc.xyz) FERC’s large-load proceeding is aimed directly at the kind of demand growth now associated with AI data centers. The commission said the docket covers reforms for the “timely, orderly, and equitable integration” of significant electrical loads into transmission infrastructure. The agency defines large loads in that docket as demand greater than 20 megawatts. (investor.atmeta.com) It is seeking comment on whether flexible large-load customers should move through interconnection studies faster, whether they should bear the full cost of required grid upgrades, and how reliability should be assessed when generation is paired with a new large customer at the same site. (datacenterdynamics.com) FERC has already taken related steps in PJM and Southwest Power Pool. The commission said it required PJM in December 2025 to implement transparent rules for substantial loads co-located with generation and approved SPP’s High Impact Large Load initiative in January 2026. (ferc.gov) ### Can the “70% of interconnection requests” claim be verified? Diamandis’ specific claim that AI accounted for more than 70% of new U.S. grid interconnection requests this year could not be independently verified from the primary and authoritative sources reviewed here. FERC’s public docket materials describe the policy response to rising data-center demand, but they do not state that AI represents more than 70% of new U.S. interconnection requests. (ferc.gov) Grid and power researchers have published adjacent figures. Grid Strategies said data centers account for about 55% of demand growth in utility load forecasts over the next five years, while Lawrence Berkeley National Laboratory has documented long delays and high withdrawal rates in generation interconnection queues. (ferc.gov) Those are related indicators, but they are not the same as Diamandis’ “over 70% of new requests” formulation. ### What should readers watch next? (gridstrategiesllc.com) (ferc.gov)