01 Jun, 2026

Log 015 - Changing of the Guard

Anthropic raised $65 billion at a $965 billion valuation, then filed for an IPO four days later. Michael Saylor's Strategy sold 32 BTC, its first Bitcoin sale in years. IBM put $10 billion and a 2029 date behind quantum. And NVIDIA picked China's Unitree as the body for its first open humanoid robot system. Four leaders changed position in one week.

In This Log

1. Anthropic Takes the Lead (AI)

On May 28, Anthropic raised $65 billion at a $965 billion valuation. The point was simple: Anthropic needs much more money for much more compute. 1

The revenue number is what makes the valuation less absurd. Anthropic said run-rate revenue crossed $47 billion in May, up from about $14 billion in February. The company was valued at $183 billion in September 2025, $380 billion in February, and $965 billion now. That is a 5x jump in nine months.

The money is tied to compute. About $15 billion of the round was already committed by cloud partners, including $5 billion from Amazon. Anthropic also has five-gigawatt compute commitments with AWS and Google. Then, on June 1, Anthropic confidentially filed a draft S-1 for an IPO. 2 OpenAI was last valued at $852 billion. Anthropic is now above it.

Why it matters

Anthropic is no longer just the safety-focused OpenAI rival. It is now the most valuable private AI company in the world, with real revenue, giant compute contracts, and an IPO filing. The fight with OpenAI is no longer only about who has the best model. It is about who controls customers, compute, developers, and the story investors believe.

Reality check

A private valuation is not a public stock price. The S-1 is still confidential, so margins, compute costs, customer concentration, and debt are hidden. The $65 billion headline also includes money already committed by cloud partners. This is a huge shift, but the IPO is the real test.

2. Saylor Blinks (Crypto)

Strategy disclosed on June 1 that it sold 32 BTC between May 26 and May 31. It raised about $2.5 million at an average price of $77,135 per Bitcoin. 3 The reason was boring: pay preferred-stock distributions. The signal was not boring. Strategy spent years telling the market one thing: buy Bitcoin and do not sell.

The sale was tiny. Strategy still held 843,706 BTC afterward. But Strategy had made itself the buyer of last resort during crypto panic. Miners were selling BTC to fund AI data centers. Strategy was buying and aiming for one million Bitcoin. Now even the loudest never-sell vehicle has shown it can sell when obligations come due.

The timing also broke a prediction market. A Polymarket contract worth more than $50 million asked whether Strategy sold before May 31. The sale happened before the deadline, but the filing arrived on June 1. 4 That is exactly the kind of edge case prediction markets are supposed to settle cleanly. They did not.

The rest of the market was already weak. Digital-asset funds saw $1.67 billion of weekly outflows, the second-largest pullback of 2026. Bitcoin products took the biggest hit. 5

Why it matters

The amount does not matter. The posture does. Strategy gave corporate Bitcoin treasuries their operating myth: permanent accumulation. A 32 BTC sale does not kill the thesis, but it proves there is a constraint. "Never sell" was not a mechanism. It was a promise.

Reality check

Thirty-two Bitcoin is rounding error for Strategy. This was not a bet against Bitcoin. The company still holds one of the largest BTC positions on Earth. The Polymarket dispute is also a market-structure story, not a Bitcoin story.

3. IBM Puts $10B on Quantum (Quantum)

On May 28, IBM told investors it plans to spend more than $10 billion on quantum computing over the next five years. The money goes into research, equipment, manufacturing, partnerships, and acquisitions. The target is clear: a large-scale fault-tolerant quantum computer by 2029. 6

The important part is not another lab milestone. It is IBM putting a dollar amount and a date into an SEC filing. Quantum has lived on roadmaps for years. This turns one roadmap into a capital commitment investors can judge.

IBM is also trying to make quantum look less like research and more like infrastructure. The company is betting that the winner will not only build better qubits. It will control the systems, software, and manufacturing path around them.

Why it matters

A 2029 date gives quantum a clock. If IBM gets close, the sector moves from promise to product faster than most investors expected. If it misses, the field has to explain why another expensive roadmap slipped.

Reality check

Ten billion dollars over five years is spread across many buckets. It is not one clean check for one machine. Fault tolerance is still hard, useful quantum workloads are still narrow, and 2029 is a target, not a delivery receipt.

4. NVIDIA Picks the Humanoid Body (Robotics)

At GTC Taipei, NVIDIA announced Isaac GR00T N1.5 and its first open humanoid robot design. The body is not NVIDIA's. It is Unitree's H2 Plus. The hands come from Sharpa. The brain is NVIDIA: GR00T models, Isaac robotics software, and two Jetson Thor T5000 modules built on Blackwell. 8

The first buyers are research labs, including Ai2, ETH Zurich, Stanford, and UC San Diego. Unitree is expected to sell the system in late 2026.

The choice is the story. NVIDIA is not trying to win the robot-body race. It wants to own the brain, simulator, training pipeline, and edge computer that everyone else builds around.

Unitree gives NVIDIA the fastest path into labs. The company already ships cheap robots at a scale Western rivals mostly do not. It is also moving toward public markets: Chinese state media says its Shanghai STAR Market IPO review has passed. 9 That makes the politics awkward. Western researchers may end up using NVIDIA software on a Chinese body.

Why it matters

NVIDIA won AI by owning the chips and the software around them. This is the same play in robotics. Let Tesla, Figure, Boston Dynamics, and Unitree fight over bodies. NVIDIA wants to be the default brain.

Reality check

A research robot is not a factory worker. Unitree's strength is cheap hardware and lab sales, not proven industrial labor. NVIDIA's design does not solve dexterity, safety, maintenance, uptime, or useful work in messy places.

Signals

Binance Starts Selling Stocks

Binance launched trading for more than 7,000 U.S.-listed stocks and ETFs for users in approved countries. It also previewed bStocks: token versions of some U.S. stocks and ETFs. 7 They are not shares, do not give direct ownership, and are not available to U.S. users.

Claude Opus 4.8 Ships

Anthropic released Claude Opus 4.8 on May 28. Better coding, better agent work, stronger professional tasks, and high-effort mode as the default in Claude Code and the API. 10 The valuation story got louder, but the product kept moving.

NVIDIA and Microsoft Put AI Agents on PCs

NVIDIA and Microsoft unveiled RTX Spark for Windows PCs on June 1: an Arm-based chip for personal AI agents with up to one petaflop of AI performance and 128GB of unified memory. 11 Same play as robotics: own the local compute.

Crypto Funds Bleed $1.67B

Digital-asset funds saw $1.67 billion of weekly outflows, the second-largest pullback of 2026. Bitcoin products took the biggest hit. 5 Strategy's sale landed in a week when institutions were already backing away.

Meta-Thread

The guard changed in four places. Anthropic passed OpenAI on private valuation and filed for an IPO. Strategy, the loudest never-sell Bitcoin company, sold. IBM put real money and a date behind quantum. NVIDIA moved into humanoids by choosing the brain layer instead of building the whole body.

None of these are final victories. Anthropic still has to face public markets. Strategy still holds a mountain of Bitcoin. IBM still has to build the machine. NVIDIA still has to prove humanoids can do useful work outside labs. But the direction changed. The money, the filings, and the obligations started forcing leaders to show their real shape.

Next Log drops next week.

© 2026 AELIUM // Nothing here is advice // readable by humans and agents