Back to News Listing

Digital Colliers Daily Briefing — May 22, 2026

Digital Colliers Daily Briefing — May 22, 2026
Digital Colliers May 22, 2026 7 min read

Digital Colliers Daily Briefing — May 22, 2026

Three stories today reshape how Washington and Silicon Valley intersect: a White House retreat from pre-release AI model review, a $60 billion compute pact that ties Anthropic to Elon Musk's infrastructure, and a Commerce Department move to take equity in nine quantum computing firms. Together they describe an administration willing to deploy capital aggressively into strategic hardware while pulling back from binding constraints on frontier software — and an industry rapidly consolidating around a smaller set of compute providers.

1. White House shelves AI executive order after CEO pushback

Vintage photo of an executive making a persuasive phone call.

What happened. President Trump declined to sign a long-anticipated executive order that would have directed the Office of the National Cyber Director and other agencies to build a federal evaluation process for frontier AI models prior to public release. According to TechCrunch, the draft included a requirement — first reported by CNN — that AI companies share advanced models with the government 14 to 90 days before launch. Trump told the press pool he "didn't like certain aspects" of the language and was unwilling to "do anything that's going to get in the way" of US leadership over China. Reporting from the Washington Post, Politico, and Axios converges on the same proximate cause: last-minute calls from Elon Musk, Mark Zuckerberg, and venture capitalist and White House AI adviser David Sacks. Sacks reportedly told Trump that companies were already cooperating voluntarily and that mandatory pre-release reviews would slow innovation. Axios sources added a blunter explanation — Trump "just hates regulation" — and TechCrunch noted the practical wrinkle that too few CEOs could reach Washington in time for a signing photo op.

Why it matters. The shelved order was the most concrete federal attempt to date to gate frontier model deployment on a security review. Its drafting had been driven in part by the recent release of Anthropic's Mythos and OpenAI's GPT-5.5 Cyber, both of which have demonstrated rapid vulnerability discovery and exploitation. Killing — or even pausing — the EO leaves the US without a binding pre-release regime and cements voluntary commitments as the default posture.

Who is affected. Frontier labs (OpenAI, Anthropic, Google DeepMind, xAI/SpaceXAI, Meta) avoid a 14-to-90-day disclosure window that would have complicated launch cadences. Federal cyber and national security agencies lose the leverage the order would have granted. State legislatures, particularly California and New York, are likely to read the federal vacuum as an invitation.

What to watch next. Whether a revised order surfaces with softer language, whether Sacks's "they're already cooperating" framing is formalized into a voluntary MOU, and whether Anthropic — which has publicly supported pre-deployment testing — breaks with the rest of the industry on the record.

Sources:

2. Anthropic commits $15 billion a year to SpaceX's Colossus data centers

Vintage photo of an engineer beside a room-sized mainframe computer.

What happened. SpaceX's S-1 filing, reviewed by The Verge, disclosed that Anthropic will pay $1.25 billion per month — $15 billion annually — through May 2029 for capacity at SpaceX's Colossus I and Colossus II AI training facilities in Memphis. The contract runs roughly four years and totals approximately $60 billion. For context, the annual figure alone approaches SpaceX's full reported 2025 revenue of $18.7 billion. Ars Technica's read of the same filing surfaces the strategic frame: SpaceX now describes launch and satellite operations as supporting infrastructure for an AI business it estimates against a $26.5 trillion total addressable market. The repositioning follows SpaceX's earlier-2026 acquisition of xAI, now consolidated as the SpaceXAI division housing Grok.

Why it matters. The deal is among the largest single AI compute commitments ever disclosed and gives SpaceX an anchor tenant that effectively underwrites continued Colossus buildout. It also illustrates how thin the line between competitor and supplier has become: Anthropic is now financially dependent on a Musk-owned facility that powers a direct rival, Grok.

Who is affected. Anthropic locks in capacity but accepts concentration risk and counterparty exposure to Musk; SpaceX gains predictable AI revenue ahead of its IPO; Nvidia secures a clear demand signal at Colossus scale; and hyperscalers (AWS, Azure, GCP, Oracle) face a credible non-cloud alternative for frontier training. Ars Technica notes that Grok itself continues to struggle commercially against OpenAI and Anthropic models — a reminder that infrastructure leadership and product leadership are decoupling.

What to watch next. The exit and substitution clauses in the contract (The Verge flags an "either"-party provision whose terms remain partially redacted), Anthropic's diversification posture toward Amazon Trainium and Google TPU capacity, and how SpaceX prices the IPO against AI-multiple comparables rather than launch-business ones.

Sources:

3. Commerce Department takes equity in nine quantum firms in $2B grant package

Vintage photo of a scientist inspecting a small crystalline component with tweezers.

What happened. The Commerce Department signed letters of intent to disburse $2 billion in grants to nine quantum computing firms in exchange for equity stakes. According to Ars Technica, IBM is slated to receive $1 billion and GlobalFoundries $375 million; the awardee list also includes D-Wave Quantum and Rigetti Computing. Markets responded immediately. CNBC, via Techmeme, reported D-Wave closing up 33%, Rigetti up 30%, and IBM up 12% on the day. Ars Technica flags two politically sensitive details: one awardee has links to a firm tied to the Trump family, and D-Wave was taken public in 2022 by Emil Michael, now a senior Pentagon official.

Why it matters. Equity-for-grants is a sharp departure from traditional non-dilutive federal R&D funding and extends the government-as-investor model — previously seen in the Intel and MP Materials transactions — into quantum. It signals that Washington views quantum as strategic enough to take balance-sheet exposure, and it gives Treasury an upside if the sector matures.

Who is affected. Pure-play quantum equities saw immediate revaluation. Established players (IBM, GlobalFoundries) gain capital with limited dilution; smaller firms (Rigetti, D-Wave, IonQ-class peers if added) face validation but also political entanglement. Allied governments — particularly the UK, EU, and Japan — will read the move as a signal to match terms or lose talent and supply chain position.

What to watch next. The dilution mechanics and any board or veto rights attached to the stakes, congressional reaction to the family- and Pentagon-adjacent awardees, and whether the same equity structure migrates to other priority sectors such as advanced packaging or fusion.

Sources:


The throughline is a hardening US industrial doctrine: subsidize and co-own strategic hardware (quantum, and by extension chips), let private capital build out the compute substrate at SpaceX-Anthropic scale, and resist binding pre-release constraints on the software riding on top. That posture concentrates risk in two places — counterparty exposure between AI labs and a small set of infrastructure owners, and security exposure absent any federal review regime. Both will be tested well before the 2029 expiry of the Anthropic-Colossus contract.

Related Posts