06-reference

stratechery data center veto digest

2026-05-22·reference·source: Stratechery·by Ben Thompson
data-center-politicsagent-economicsagentic-webdeepmindgoogle-iocapexpower-thesis

"The Data Center Veto (This Week in Stratechery)" — @Ben Thompson

Why this is in the vault

This Week in Stratechery is a weekly hybrid that combines (a) Andrew Sharp's editorial picks of the week's three favorite items with (b) a flat link index of every piece published across the Stratechery bundle (Updates, Interviews, Dithering, Asianometry, Sharp China, GOAT, Sharp Tech, video). Three of the picks this week land directly on RDCO operating territory: data-center politics (power thesis), agent economics (RDCO core), and a DeepMind / Google AGI approach note (model-substrate radar). Worth filing the digest as a routing index even though the load-bearing argument (Data Center Discontent) already has its own deep-dive note.

The week's picks (editor framing)

  1. Data Center Discontent (Ben Thompson, Monday Update + Sharp Tech episode). Thesis: AI's economic impact lands digitally, but AI depends on physical data centers, and data centers need local permits. That gives normal people a veto power over AI infrastructure that they didn't have during globalization (when the disruption was elsewhere). Sharp's editor framing: "understanding this dynamic is more important than trying to correct misinformation, which is a symptom, not a cause, of data center opposition." Already filed as a standalone deep-dive — see [[2026-05-18-stratechery-data-center-discontent]].

  2. Agent Economics (Andrew Sharp pick — Stratechery Interview with Parag Agarwal). Parag Agarwal (former Twitter CEO) is now running Parallel, a company focused on the question: what does internet economics look like when ad-supported models break under agentic web traffic? The interview revisits Ben's summer 2025 piece The Agentic Web and Original Sin and gets into why ads make sense for humans, why incentivizing content for agents is structurally different, and what Parallel is building to solve it. Sharp's framing: "I learned a ton from this interview."

  3. Never Count Out the Slime Mold (Andrew Sharp pick — Ben's Google I/O Update). Sharp invokes the "ungovernable, poorly coordinated mold in Mountain View" framing of Google, noting the I/O firehose of "10 different types of AI spaghetti at the wall." But: Google is a nearly $5T company, the transformer architecture supercharged the AI era, and Ben specifically highlights a DeepMind world-models approach to AGI that is distinct from OpenAI and Anthropic's approaches. Sharp's reluctant tell: pay attention to DeepMind's bets even when Google looks chaotic.

Issue contents (link index)

Stratechery Articles and Updates

Dithering (Ben + John Gruber)

Asianometry (Jon Yu)

Sharp China (Andrew Sharp + Bill Bishop)

Greatest of All Talk

Sharp Tech (Andrew Sharp + Ben)

This week's Stratechery video: The Inference Shift (already filed as deep-dive — see [[2026-05-11-stratechery-inference-shift-agentic]] for prior coverage of the same arc).

Mapping against Ray Data Co

Data Center Discontent / Veto thesis → power thesis direct hit. This is the political-risk dimension of the Power Cycle v1 / v1.1 thesis ([[2026-05-17-power-cycle-v1]], [[2026-05-18-power-cycle-v1.1]]). The thesis already prices in physical bottlenecks (gas turbine lead times, interconnect queue, PJM capacity-auction prints). Ben's veto framing adds a durable political-friction layer on top of the physical bottlenecks — even after the supply chain catches up, local-permit veto is a structural rate-limiter on hyperscaler capex deployment. Strengthens the long-tail of the power thesis (slower buildout = longer cycle = longer runway for incumbents with already-permitted sites and existing IPP relationships). The deep-dive note on the Monday Update already captured this; this digest just reinforces that Ben himself is doubling down on the framing across both his Update and the Sharp Tech episode this week.

Agent Economics / Parallel interview → RDCO core operating territory. This is the same problem space that [[2026-05-21-every-after-automation]] (Dan Shipper's "frame vs framer" essay) maps from the deployer angle. Parallel's bet is on the supply side (content economics, incentivizing creators to produce for agents) where Shipper's bet is on the demand side (humans framing agentic work). Both are downstream of Ben's Agentic Web and Original Sin (2025) prediction. For RDCO this matters two ways: (1) Sanity Check newsletter publication-economics — if agent traffic eats ad-supported publishing within 24-36 months, the newsletter-as-loss-leader / paid-tier / paid-sponsorship model needs to be locked in before the transition window closes; (2) the Squarely / MAC info-product surface needs to be agent-discoverable / agent-purchasable, not ad-funnel-discoverable. Queue: read the Parallel interview directly (free article? confirm at deep-fetch time), then decide whether to spin up a [[RDCO agent-economics positioning]] concept article. Not done in this digest — the interview body wasn't deep-fetched, only Sharp's editor framing was captured.

DeepMind world-models / Google I/O note → model-substrate radar. Ben specifically calling out that DeepMind's approach to AGI is distinct from OpenAI and Anthropic is a model-substrate signal. RDCO is currently 100% Anthropic-substrate (Claude Code harness). A DeepMind/Gemini approach diverging architecturally enough that Ben — who normally treats the three frontier labs as a homogeneous capex bloc — calls it out separately is the kind of signal that should ping the "substrate diversification" backlog. Not actionable this week; tag for the next L5-substrate-strategy review.

Sharp Tech "much ado about data centers" episode → likely same-thesis evidence as the Update, no separate deep-dive needed unless founder asks.

Asianometry "Little Vertical Laser" (VCSEL pun?) and Intel Costa Rica — lighter signal, not on any active thesis. Skip unless founder pings them.

Sponsor disclosure

No sponsor block in the digest itself. The bundle is subscription-funded (no third-party ads), but the digest does heavy cross-promotion of every sister-podcast in the Stratechery bundle (Dithering, Asianometry, Sharp China, GOAT, Sharp Tech) — that's bundle-internal cross-promo, not third-party sponsorship, but worth flagging for the curation-self-promo invariant: every link in this digest's curation section is to a Stratechery-bundle property. Zero third-party curation.

Related