Superpower Daily

Superpower Daily

Your reading space

Sign in
Google Cloud Opens Singapore Engineering Center for Global AI P...Apple Releases macOS 27 With a Siri That Can Read Your ScreenBolt’s New Policies Permit Training Beyond Its Forge-Only PitchEUCLYD Raises More Than €200 Million for Non-GPU AI Inference S...Anthropic Publishes Misuse Report, Says Safeguards Falter When...WIRED Reports a Date Filmed and Posted a Woman Without Discussi...Google Cloud Opens Singapore Engineering Center for Global AI P...Apple Releases macOS 27 With a Siri That Can Read Your ScreenBolt’s New Policies Permit Training Beyond Its Forge-Only PitchEUCLYD Raises More Than €200 Million for Non-GPU AI Inference S...Anthropic Publishes Misuse Report, Says Safeguards Falter When...WIRED Reports a Date Filmed and Posted a Woman Without Discussi...
Swipe

Wall Street Sees AI Spending Rolling On Despite Slowdown Calls

The safety appeal now faces an expansion cycle that analysts expect major cloud companies to sustain, even if they build more capacity than they need.

By 2 min read
Wall Street Sees AI Spending Rolling On Despite Slowdown Calls
Wall Street Sees AI Spending Rolling On Despite Slowdown Calls

Listen to this story

The audio brief

About 1:36
0:001:36
Read transcript
Alphabet, Amazon, Meta, and Microsoft spent $293 billion on capital expenditures in just the first half of 2026—and they were reportedly on pace for nearly $600 billion in AI infrastructure spending for the full year. That makes Wall Street’s message clear: calls to slow advanced AI development are not, at least for now, stopping the build-out behind it. The appeal for restraint came from Anthropic CEO Dario Amodei, with OpenAI CEO Sam Altman and Elon Musk echoing calls for a slower pace of frontier-model development. But D.A. Davidson analyst Gil Luria says the major hyperscalers are not slowing either development or spending. The distinction matters. The safety warnings target how quickly increasingly capable models are developed. The spending figures measure data centers and other infrastructure needed to support that work. They also describe a pace or forecast—not a final 2026 total—and they do not prove that every model project will pay off. Spending is still accelerating in the cited data. Meta’s capital expenditures rose 58 percent from one quarter to the next, while Amazon and Google also increased spending. Luria says hyperscalers could absorb overbuilding during the next one to two years, potentially with higher cash flow. There is political resistance, too. President Donald Trump rejected AI-model guardrails and framed continued development as a U.S.-China competition. The key question now is whether voluntary restraint can overcome the financial and strategic pressure to keep expanding capacity—even if companies end up building more than they need.

Story brief

3 key points

Investors and hyperscalers are not treating public calls to slow frontier-model development as a near-term brake on infrastructure spending. Alphabet, Amazon, Meta, and Microsoft spent $293 billion in the first half of 2026, with reported full-year AI infrastructure outlays approaching $600 billion. That figure is a pace or forecast, not a final total, and infrastructure commitments do not prove every model project...

  1. 01

    D.A. Davidson analyst Gil Luria said major hyperscalers are not slowing AI development or spending.

  2. 02

    Meta’s capital expenditures rose 58% quarter over quarter; Amazon and Google also increased spending.

  3. 03

    Luria said hyperscalers could absorb overbuilding over the next one to two years, potentially with higher cash flow.

Wall Street is treating calls to slow frontier AI development as unlikely to interrupt the infrastructure build-out behind it. Alphabet, Amazon, Meta and Microsoft recorded $293 billion in combined capital expenditures in the first two quarters of 2026 and were reportedly on pace to spend nearly $600 billion on AI infrastructure for the year.

The spending outlook follows Anthropic CEO Dario Amodei’s public appeal for the industry to slow down. OpenAI CEO Sam Altman and Elon Musk echoed that call. But D.A. Davidson technology analyst Gil Luria said major hyperscalers are not slowing AI development or spending.

The disagreement is not just about whether companies acknowledge safety risks. Amodei and the leaders backing him want a slower pace for advanced-model development. The capital-expenditure figures instead show spending on the data-center infrastructure that supports AI work, and the annual figure is a forecast rather than a completed total.

The current spending pace
$293 billionFirst-half capital expenditures

Alphabet, Amazon, Meta and Microsoft recorded this total across the first two quarters of 2026, according to AlphaSpace data cited by Yahoo Finance.

Nearly $600 billionReported 2026 AI infrastructure pace

The figure is a reported pace for the full year, not a completed 2026 spending total.

The spending was accelerating in the cited data: Meta’s capital expenditures rose 58% quarter over quarter, while Amazon and Google also increased spending. Luria argued that even if hyperscalers overbuild over the next year or two, they could absorb the extra capacity; he said their cash flow could rise in that scenario.

The slowdown appeal also faces political resistance. President Donald Trump rejected calls for AI-model guardrails and said the United States is winning the AI race against China. That framing puts continued development inside a competitive contest, rather than treating voluntary restraint as the immediate priority.

The immediate tension is between a request to slow capability gains and a spending cycle investors expect to continue. The spending pace does not establish that every project will succeed, and it does not show that model developers will reject restraint. It does show why a voluntary slowdown will confront powerful expectations for continued infrastructure expansion.

Sources

  1. finance.yahoo.comWall Street pushes back on Anthropic

Loading discussion...

Related coverage

Continue reading