"Shifting Alliances in a Changing World (This Week in Stratechery)" — @Ben Thompson
Why this is in the vault
Weekly Stratechery digest (issue 2026.20). Pointer-only document — every essay highlighted here was already filed individually earlier in the week (Inference Shift on 5/11, SpaceX-Anthropic-xAI on 5/12, DeployCo on 5/13, MoffettNathanson interview on 5/14). Filed for completeness of the Stratechery weekly trail and so that the cross-issue narrative arc ("agentic inference + xAI compute realignment + US-China posture shift") is captured as a single Thompson-curated framing.
Issue contents
Thompson's three "favorites of the week" callouts:
- A New Kind of Computing. Pointer to The Inference Shift. Two kinds of inference: today's "answer inference" (humans in loop, speed matters) vs the future-dominant "agentic inference" (no humans, speed mostly doesn't matter). Different architecture trade-offs, good news for China and space-based compute, possibly bad for Nvidia.
- All About Elon. Pointer to the xAI-Anthropic compute deal coverage. Anthropic side reads as "markets work." xAI side raises whether Musk listens to market signal about who SpaceXAI should serve. Plus Andrew Sharp on the OpenAI lawsuit — boring and insulting on the merits but already a strategic win for Musk regardless of verdict.
- 360 Degrees of US-China Relations. Pointer to Sharp China episode framing the Trump-Xi Beijing summit. "Underwhelming deliverables" predicted and confirmed. Useful frame: "upper hand" analysis is overblown, both sides incentivized to play for time and stability, China's posture has shifted from the 90s/2000s.
Full bundle table of contents:
- Stratechery articles: The Inference Shift; SpaceX and Anthropic, xAI's Two Companies; The Deployment Company, Back to the 70s, Apple and Intel; MoffettNathanson Conference Interview.
- Sharp Text: Elon's OpenAI Lawsuit Is Boring and Insulting.
- Dithering: Apple's Supply Squeeze; Apple's AI Land Grab.
- Asianometry: General Motors Dreamt of Robots.
- Sharp China: 10 Questions on the Trump-Xi summit.
- Sharp Tech: Inference in the Agentic Future, xAI Q&A, Intel, Apple.
- Stratechery video: Amazon's Durability.
- GOAT (basketball pod): NBA lottery + Cavs/Spurs/Knicks coverage.
Mapping against Ray Data Co
This digest doesn't extend the existing positions — it consolidates them. Three already-filed Stratechery essays now have a Thompson-curated through-line:
- Agentic-inference architecture shift is the load-bearing one for RDCO. It connects to the [[2026-05-14-treybig-how-agents-use-systems-differently]] thesis (agents have different latency/throughput needs than humans), the [[2026-05-14-technically-package-managers-ai-labs-acquisitions]] note on labs acquiring infrastructure-runtime companies (Anthropic-Bun, OpenAI-Astral) precisely because agentic compute shape is different, and the [[2026-05-14-innermostloop-singularity-compounds]] infra-thesis. The Thompson framing — "speed won't matter when humans aren't involved" — is the cleanest one-line statement of why agent infra is a separate stack from interactive AI infra. Worth quoting in any future Sanity Check piece on that shift.
- xAI-Anthropic deal reinforces the bet that compute is the bottleneck and that lab-to-lab compute trades are now strategy-grade events, not corner cases. No new RDCO action.
- US-China summit framing is useful background on chip-export and DC policy environment but doesn't change any RDCO bet.
No new positions. No re-frame. The week's individual essays already carry the load; this digest is the index.
Related
- [[2026-05-11-stratechery-inference-shift-agentic]]
- [[2026-05-12-stratechery-spacex-anthropic-xai-musks-two-companies]]
- [[2026-05-13-stratechery-deployco-70s-apple-intel]]
- [[2026-05-14-stratechery-thompson-moffettnathanson-compute-aggregation-consumer-ai]]
- [[2026-05-14-treybig-how-agents-use-systems-differently]]
- [[2026-05-14-technically-package-managers-ai-labs-acquisitions]]
- [[2026-05-14-innermostloop-singularity-compounds]]
- [[2026-05-08-stratechery-earning-spending-weekly]]