Tech-Reader AI Digest

Thursday, May 7, 2026

#AI #TechNews #Digest

Story 1: Musk's Best Trial Day — The Sutskever Dossier and Three Hits on Altman's Honesty

What happened: As the trial between Elon Musk and OpenAI ended its second week, Musk's witnesses landed three solid punches in testimony about how Altman runs OpenAI — raising concerns about his dedication to AI safety, the nonprofit's mission, and his honesty as a leader.

Hit One — Rosie Campbell, former OpenAI safety researcher: Campbell testified that when she joined OpenAI in 2021, it had two teams dedicated to long-term AI safety — one ensuring AI aligned with human values, another preparing the world for superhuman AI. She left in 2024 after her AGI Readiness team was disbanded in early 2024 — a turning point for her trust in the mission — the same period the Super Alignment team was also shut down. "When I joined, it was very research-focused and common for people to talk about AGI and safety issues," she testified. By the time she left, she felt she had no home at OpenAI.

Hit Two — Tasha McCauley, former OpenAI board member: McCauley testified to a pattern of dishonesty and a culture of deceit under Altman that caused "crisis events" every few months. The most explosive moment: she received an email from Ilya Sutskever — now being called "The Sutskever Dossier" by legal teams — containing "dozens of pages of examples of different chaotic events that had occurred from Sam's behavior or lies that he had told." McCauley also testified that Altman lied to another board member about her intention to remove Helen Toner. And critically — the board was not informed prior to the public launch of ChatGPT. "We are a non-profit board and our mandate was to be able to oversee the for-profit underneath us. We did not have a high degree of confidence at all to trust that the information being conveyed to us allowed us to make decisions in an informed way."

Hit Three — David Schizer, former Columbia Law School dean, Musk's nonprofit law expert: Schizer testified that OpenAI under Altman launched products without board knowledge and that Microsoft tested a version of GPT-4 without going through the company's safety review process — directly echoing the Scott Memo from the March 2018 Microsoft email we covered last week. "The board and CEO need to be partnering, working together, to make sure the mission is being followed. If the CEO is withholding that information, it's a big problem."

No court Friday. Trial resumes Monday. (Source: Business Insider / TechCrunch / NBC Bay Area / ABC7)

Why it matters: Three independent witnesses — a safety researcher, a former board member, and a nonprofit law expert — all testified on the same day to the same pattern. The Sutskever Dossier is the governance smoking gun that explains why the board took action in November 2023. What the world dismissed as a palace coup was a board that had accumulated documented evidence — dozens of pages of it — of exactly what Musk is alleging.

Aaron's take — The trial is rewriting the history of November 2023 in real time. The board that fired Altman — a firing the world dismissed as a failed coup — is starting to look like a governance body that had been building a documented case for months. The Sutskever Dossier didn't appear in a day. It was compiled over time. By one of OpenAI's own co-founders.

Story 2: Anthropic Commits $200 Billion to Google Cloud — Vertical Extraction and the Circular Infrastructure Economy

What happened: Anthropic has committed to spend $200 billion with Google Cloud over five years starting in 2027, The Information reported, citing a person with knowledge of the matter. Reuters confirmed the contours. Neither company confirmed the figure on the record. The commitment suggests Anthropic accounts for more than 40% of Google Cloud's total revenue backlog.

The circular structure is worth understanding clearly. Google invested up to $40 billion in Anthropic. Anthropic commits $200 billion back to Google Cloud over five years. The investment paid for the chips. The chips run the model. The model generates the revenue that funds the investment.

Analysts are calling Google's position Vertical Extraction — by owning the chips (TPUs), the cloud infrastructure (GCP), and the investor stake (up to $40 billion in Anthropic), Google extracts value at every layer of the $200 billion commitment simultaneously. It's not just a cloud deal. It's a vertically integrated capture of Anthropic's entire compute spend.

Contracts involving Anthropic and OpenAI now account for more than 50% of the $2 trillion in backlogs across Amazon Web Services, Microsoft Azure, and Google Cloud Platform. Two AI startups — one founded in 2019, one in 2021 — are responsible for over $1 trillion in committed cloud spending driving the revenue backlogs of three of the world's largest technology companies.

Anthropic's broader compute stack: Amazon ($100B+ over 10 years), Google ($200B over 5 years from 2027), Microsoft/NVIDIA ($30B Azure), Fluidstack ($50B), SpaceX Colossus 1 (300MW, 220K GPUs). The company projecting $20 billion in compute outlays this year alone is betting that $30 billion in run rate today becomes the foundation for a much larger business by 2029. (Source: The Information / Reuters / Engadget / Sherwood News / Let's Data Science)

Why it matters: Google receives 40% of its entire cloud revenue backlog from a single customer it also invested $40 billion in. That's not a vendor relationship. That's a strategic fusion. Vertical Extraction is the architecture that has Alphabet closing in on Nvidia as the world's most valuable company — its market cap hitting $4.67 trillion versus Nvidia's $4.79 trillion — a gap driven by 63% Google Cloud growth and TPU-led vertical integration.

Aaron's take — Google invested $40 billion in Anthropic. Anthropic commits $200 billion back to Google. That's a 5x return on the investment before Anthropic ships a single additional product. The circular economy of AI infrastructure is the most interesting financial structure since the railroad bond markets of the 1860s. The buildout is real. The demand is real. The question is whether the timeline matches the commitments — and at $200 billion over five years starting 2027, Anthropic needs to scale substantially from its current $30 billion run rate for the math to work cleanly.

Story 3: The $2 Trillion Backlog — How Two AI Startups Are Funding the Cloud Giants

What happened: The $200 billion Google commitment is the headline but the context is the real story. Contracts involving Anthropic and OpenAI now account for more than half of the $2 trillion in backlogs at Amazon Web Services, Microsoft Azure, and Google Cloud Platform.

To put that in perspective: two AI startups — one founded in 2019, one in 2021 — are responsible for over $1 trillion in committed cloud spending that is driving the revenue backlogs of three of the world's largest technology companies. Amazon's backlog. Google's backlog. Microsoft's backlog. All of them substantially inflated by commitments from companies that didn't exist seven years ago.

Previous projections estimated server costs in 2026 could reach $45 billion for OpenAI and $20 billion for Anthropic. Similar circular structures have also emerged at chipmakers — Nvidia has made its own investments into OpenAI, creating the same invest-and-commit loop at the silicon layer.

The honest question: what happens if AI revenue growth decelerates before these commitments are fulfilled? The $200 billion Anthropic owes Google over five years starting 2027 requires Anthropic's revenue to scale substantially from its current $30 billion run rate. The math works at $100+ billion in annual revenue. It gets complicated at $40 billion. (Source: The Information / Reuters / Engadget / Let's Data Science / Motley Fool)

Why it matters: The $2 trillion backlog is the single most important number in AI infrastructure right now and it's barely being discussed. It represents the financial commitment that makes the infrastructure war real — not aspirational. And it means the cloud giants' revenue growth for the next five years is substantially dependent on two AI startups continuing to scale at historic rates.

Aaron's take — The railroad bond markets of the 1860s were built on the assumption that westward expansion would generate the traffic to service the debt. Most of those bonds eventually paid off — but not before a spectacular bubble and crash in between. The AI infrastructure circular economy may follow the same arc. The question isn't whether the buildout is real. It is. The question is timing.

Quick Hits — The Rest of Today's AI World

Anthropic / Claude

Beyond Stories 2 and 3 — Claude Code rate limits doubled Wednesday remain in effect. $900B valuation funding round in final stages. October IPO trajectory unchanged. (Source: Anthropic)

Gemini (Google)

Alphabet market cap at $4.8 trillion — closing in on Nvidia's $5.08 trillion. 63% Google Cloud growth and Vertical Extraction strategy driving the surge. No new model announcements today. (Source: Reuters / Motley Fool)

VS Code / GitHub Copilot

No new announcements. Token-based billing June 1 — 25 days remaining. (Source: GitHub)

Replit

No new announcements. (Source: Replit)

Perplexity

No new announcements today. (Source: Perplexity)

Microsoft Copilot

Copilot Cowork mobile from Wednesday remains standing news. No new announcements today. (Source: Microsoft)

xAI / SpaceXAI

Formally rebranding as SpaceXAI confirmed. Colossus 1 / Anthropic deal remains standing news. No court Friday — trial resumes Monday. (Source: Musk X post)

OpenAI

Trial Day 8 closes — Musk's best day yet. No court Friday. Trial resumes Monday. No new product announcements. (Source: TechCrunch / Business Insider)

Palantir

No new announcements today. Q1 blowout earnings remain standing news. (Source: CNBC)

Reflection AI

Pentagon Impact Level 7 Asimov rollout proceeding. No new announcements today. (Source: Breaking Defense)

Ollama

No new announcements today. (Source: Ollama)

DeepSeek

V4-Pro and V4-Flash live since April 24. Cost-efficiency leader at $3.48/M output tokens. No new announcements today. (Source: DeepSeek)

Alibaba / Qwen / Z.ai

No new announcements today.

Inflection Pi / Mistral

No major news today.

That's your AI world for Thursday, May 7. No court Friday. Trial resumes Monday. Back then. — Aaron

Aaron Rose is a software engineer and technology writer at tech-reader.blog.

Catch up on the latest explainer videos, podcasts, and industry discussions below.

Search This Blog

Tech-Reader.blog