Frontier signal stream

Anthropic signed a 20-year lease at TeraWulf's Justified Data campus in Hawesville, Kentucky, taking roughly 401 MW of IT load, with first power delivery expected in phases. The deal is expected to generate about $19 billion in contracted revenue for TeraWulf and marks a major AI-infrastructure capacity commitment by Anthropic ahead of its planned IPO. TeraWulf separately sold its Abernathy stake for ~$530M. Announced July 6, 2026.

Reflection AI gains immediate access to Nvidia GB300 chips at SpaceX's Colossus 2 data centre, agreeing to pay SpaceX $150 million a month from 1 July 2026 through 2029 — about $6.3 billion if the agreement runs its full term. CRITICAL CAVEAT: the $6.3 billion is contingent on the full term and is NOT a committed sum — either party can terminate on 90 days' notice after the first three months. CNBC reported the terms from materials it reviewed, and a Reflection spokesperson commented on the record. Reflection has not yet released a public frontier open-source model. For SpaceX this extends Colossus from an internal Grok training asset into a commercial compute platform, following deals with Anthropic, Google and Cursor.

Per a regulatory filing Friday, Google will pay SpaceX $920M/month for 32 months (~$30B) to lease ~110,000 Nvidia GPUs plus CPUs/memory housed in SpaceX (xAI) data centers, ramping from October 2026 through June 2029. It is SpaceX's second mega compute-leasing deal after the Anthropic Colossus-1 agreement in May, bolstering its AI-revenue story ahead of its planned ~$1.75T IPO targeted around June 12.

CNBC reported on 21 May 2026 that Microsoft and Anthropic were in talks about Anthropic using Microsoft's in-house Maia 200 AI accelerator, extending a relationship set out in November 2025 when Microsoft agreed to invest up to $5bn in Anthropic (alongside up to $10bn from Nvidia) and Anthropic committed to $30bn of Azure capacity. The report describes negotiations rather than a concluded agreement; no terms, volumes or timing were confirmed by either company.

Thinking Machines Lab signed a new agreement to expand its use of Google Cloud AI infrastructure, valued in the single-digit billions of dollars according to a source familiar with the matter. It includes access to Google systems built on NVIDIA's GB300 chips plus infrastructure services for model training and deployment, and makes Thinking Machines among the first Google Cloud customers on GB300-powered systems. It is the lab's first deal with a cloud services provider, and is non-exclusive. Google noted in a press release that it can support the startup's reinforcement learning workloads, which the architecture of its Tinker product relies on. Founding researcher Myle Ott said Google Cloud 'got us running at record speed with the reliability we demand.'