June 5 Roundup: Anthropic Fights Pentagon Blacklist, House AI Bill Drops, Google Gemini Spark Goes Live, SpaceX AI Targets 100× Revenue Surge, and OpenAI Publishes Federal Blueprint
Thursday delivered one of the most consequential news days of the year for AI. The Pentagon drew a harder line on Anthropic's national security designation. Congress unveiled its first serious bipartisan AI bill. Google's Gemini Spark agent went live for US subscribers. Goldman Sachs put staggering numbers on SpaceX's AI future. And OpenAI formally asked Washington to govern frontier AI on federal terms — all while Anthropic's IPO preparations continue apace. Here's what you need to know.
1. Hegseth Doubles Down: Anthropic Stays on the Pentagon's Blacklist
In a significant blow to Anthropic's government contracting ambitions, Defense Secretary Pete Hegseth formally denied the AI company's request to reconsider its national security risk designation. In a June 3 decision letter obtained by Politico, Hegseth wrote that the "pre-deployment risks with the Covered Entity's products and services, the loss of trust, and other risk factors … were and remain sufficient to support the prior Determination."
The designation effectively bars Anthropic from certain Pentagon contracts and signals that the department is not backing down from its assessment of the company as a supply-chain security concern. The decision comes at a particularly awkward moment: Anthropic had just confidentially filed its S-1 with the SEC on June 1 and announced $47 billion in annualized revenue. The company is preparing for one of the largest AI IPOs in history, and a Pentagon blacklist doesn't help the roadshow narrative.
"Pre-deployment risks with the Covered Entity's products and services, the loss of trust, and other risk factors … were and remain sufficient to support the prior Determination."
— Defense Secretary Pete Hegseth, June 3, 2026 decision letter (via Politico)
The dispute reportedly stems from concerns around Anthropic's training data practices, model transparency, and its foreign investor relationships — including significant backing from Google and Saudi Arabia-linked funds. Anthropic has consistently pushed back on the designation, and the denial is expected to accelerate legal and political pressure from the company's allies on Capitol Hill.
This is the most concrete example yet of AI companies getting caught in the crossfire between national security policy and commercial AI expansion. If you're a federal contractor evaluating AI vendors, the Pentagon's stance should factor directly into your procurement risk assessment. Anthropic isn't going away — its enterprise business is massive — but the government sales channel just got harder. Meanwhile, competitors like OpenAI, which has actively courted federal alignment through its governance blueprints, are benefiting from the contrast.
Sources: Politico, TechCrunch
2. Anthropic's IPO Momentum: Daniela Amodei Shrugs Off the Doubters
Despite the Pentagon headwind, Anthropic's co-founder and President Daniela Amodei showed no signs of slowing down at the Bloomberg Tech conference on Thursday. Amodei addressed the decision to file confidentially for an IPO — announced June 1 — with characteristic directness: it's about capital access, full stop.
"It's a really big upfront cost to train the models and to serve inference on them. My guess is that over time, the sort of core set of companies that are working to advance the frontier are just going to need access to capital, and I think the public market is very well suited to that."
— Daniela Amodei, Bloomberg Tech Conference, June 4, 2026
The timing is deliberate. Anthropic announced annualized revenue crossed $47 billion in May — a dramatic leap from roughly $9 billion at the end of 2025. Multiple investors told TechCrunch the company's recent $65 billion fundraise at a $965 billion valuation was "greatly oversubscribed." With private demand still strong, the IPO is now a question of when, not if.
Amodei also pushed back on skeptics questioning whether enterprise AI spending is delivering real ROI — a concern that surfaced publicly when Uber acknowledged not all its AI investment had proven productive. Her view: companies are still early in figuring out deployment, and use cases across coding, financial services, legal, and healthcare will continue to expand as businesses become more fluent with the tools.
The filing means Anthropic could potentially hit public markets within months — racing alongside SpaceX, which is targeting a June 12 IPO debut, and OpenAI, which has been targeting a September listing.
Three of the most valuable AI companies in the world could go public within the same 90-day window. That's an extraordinary moment for the industry. For enterprise buyers, it also means these companies will face intense quarterly scrutiny of their revenue growth and margins — which typically accelerates their push to convert enterprise pilots into committed annual contracts. If you're in active evaluation of Claude, expect the sales motion to become significantly more aggressive post-IPO.
Sources: TechCrunch, Anthropic Newsroom
3. House AI Bill Lands: Federal Framework Would Preempt State Model Laws
The most significant AI governance development in months dropped Thursday, as a bipartisan pair of House lawmakers released draft legislation that would create a federal framework for AI — and bar states from passing laws that regulate the development of AI models themselves.
The draft, authored by Democrat Lori Trahan of Massachusetts and Republican Jay Obernolte of California, would prohibit states from requiring models to undergo pre-release testing or other development-stage oversight. Crucially, the bill would not preempt state laws that regulate how AI is actually used — a distinction that matters enormously for employment law, consumer protection, and sector-specific applications.
"We are releasing this draft to hear from stakeholders, experts, and the public so we can strengthen the legislation before it is formally introduced."
— Rep. Jay Obernolte (R-CA), joint statement, June 4, 2026
The bill aims to create "a national framework that protects Americans, supports innovation, and ensures the U.S. leads the world in shaping this technology." But the reception was predictably polarized. Consumer advocacy group Public Citizen slammed the bill, saying it would leave "oversight largely to a federal government that has repeatedly failed to pass meaningful AI protections" — and notably fails to address algorithmic discrimination, deepfake exploitation, and youth mental health harms. State legislators in Massachusetts and New York warned Rep. Trahan against preempting their own AI regulatory authority.
The industry applauded. The Information Technology Industry Council and major tech lobbying groups have long pushed for a federal preemption approach, arguing that a patchwork of 50 state AI laws would create unworkable compliance burdens for developers.
The draft lands just days after OpenAI published its own federal blueprint on June 3, calling on Congress to enact a single federal framework for frontier AI and preempt state laws that "regulate the same frontier safety risks." The convergence of industry lobbying and bipartisan legislative action suggests a federal AI law is closer than it has ever been — though the path through Congress remains treacherous.
The Trahan-Obernolte bill is a watershed moment even in draft form. For the first time, there's a credible bipartisan vehicle for federal AI legislation — and the model/use distinction is cleverly designed to thread the needle between industry and consumer advocates. If this passes in anything like its current form, it fundamentally reshapes the compliance landscape: companies building AI tools would face one federal standard for model development, while downstream use cases (employment, housing, healthcare) remain governed at the state and federal sector level. Boards and general counsel should be tracking this closely.
Sources: Reuters, Axios, Politico
4. Google Gemini Spark Is Now Live — And It Runs While You Sleep
Google's most ambitious consumer AI product yet began rolling out Thursday. Gemini Spark — the 24/7 personal AI agent announced at Google I/O 2026 — is now available to Google AI Ultra subscribers in the United States. It's not just another chatbot upgrade. Spark is explicitly designed to operate in the background, running on Google Cloud infrastructure even when your phone or computer is off.
Built on Gemini Flash 3.5 and Google's Antigravity agentic platform, Spark can book flights, manage calendars, draft emails, gather documents for status updates, and take actions on your behalf across Google's product suite — all without requiring you to be present. PCMag described it as going "beyond what Google Gemini alone offers," noting its native integration with Gmail, Google Docs, Google Calendar, and Google Search as a key differentiator over standalone agentic tools.
"At its recent I/O developer conference, Google introduced Gemini Spark as an always-on agent that connects to your personal data, completes online tasks, and automates aspects of your daily interactions."
— Wired, hands-on review
Access currently requires a Google AI Ultra subscription and is limited to users aged 18 and over in the US. Enterprise access is available to select business users via a separate program. The rollout marks a significant escalation in the agentic AI arms race: Anthropic's Claude agents, OpenAI's Codex, and now Google's Spark are all competing to become the default background worker in professional and personal workflows.
Gemini Spark represents the most aggressive deployment yet of the "AI that works while you don't" premise. The integration advantage Google holds here is real — if Spark can reliably act across Gmail, Calendar, Drive, and Search without requiring you to context-switch into a separate interface, that's a genuine workflow transformation, not just a feature upgrade. The critical questions are: what does its error rate look like on real-world tasks, and how does it handle ambiguity? Wired's headline about it "friend-zoning her boyfriend" is cute — but enterprise buyers need to understand the judgment boundaries before handing it calendar access and email send permissions.
5. Goldman Sachs: SpaceX's AI Revenue Will Hit $322 Billion by 2030 — a 100× Surge
Goldman Sachs has shared projections with potential SpaceX IPO investors that are almost difficult to process: the bank expects SpaceX's AI division to grow revenue from $3.2 billion in 2025 to $322 billion in 2030, a roughly 100-fold increase over five years. The figures, first reported by the Financial Times and confirmed by Reuters, underpin SpaceX's reported $1.78 trillion IPO valuation — one of the largest ever attempted.
Goldman sees SpaceX's AI segment growing 388% year-over-year in 2026 alone, reaching $15.6 billion, then $34.5 billion in 2027, before accelerating sharply through the end of the decade. Total SpaceX revenue is projected to reach $474 billion by 2030, with the AI division driving the largest share of growth. Adjusted EBITDA across the company is expected to reach $352 billion — up from $6.6 billion in 2025.
"Goldman Sachs expects revenue from SpaceX's AI division to surge to $322 billion by 2030, up from $3.2 billion in 2025."
— Reuters, citing Financial Times reporting on Goldman Sachs investor projections
The projections are largely powered by SpaceX's acquisition of Elon Musk's xAI earlier this year in a deal that valued xAI at $250 billion. That transaction folded Grok models, xAI's inference infrastructure, and its enterprise AI products into SpaceX's business — creating a combined entity with both the distribution advantages of Starlink's global connectivity and the frontier model capabilities of one of the leading AI labs.
SpaceX is reportedly targeting a June 12 IPO debut, making this week's projections part of the final investor pitch push. Whether the market will accept Goldman's growth assumptions — which require an extraordinary compounding of AI adoption across Starlink, enterprise software, and autonomous systems — remains to be seen.
Take the $322 billion figure with appropriate skepticism — these are IPO pitch projections, not audited forecasts. But the underlying thesis isn't crazy: Starlink already has millions of enterprise and government subscribers globally, and layering AI services on top of an existing connectivity business is a genuinely different go-to-market than OpenAI or Anthropic's pure software plays. If even 20% of the projected AI revenue materializes, SpaceX/xAI becomes one of the largest AI revenue generators on the planet. That changes competitive dynamics significantly for everyone else in the space.
Sources: Reuters, Motley Fool, Financial Times
6. OpenAI Publishes Federal AI Governance Blueprint — and Asks Congress to Preempt State Laws
OpenAI released a nine-page blueprint on June 3 calling on Congress to establish a single federal framework for governing frontier AI — and explicitly asking lawmakers to preempt state laws that "regulate the same frontier safety risks." The document, titled "Democratic Governance of Frontier AI: A Blueprint for a Federal Framework," is the company's most direct engagement with federal policy to date and was released alongside a separate public policy agenda document.
The blueprint lays out a three-part structure: federal safety standards for frontier model development, a licensing or notification regime for the most powerful models, and a preemption of state-level laws that would otherwise create a fragmented compliance patchwork. OpenAI argues that "building on state laws and White House actions" while establishing federal primacy is the right approach for maintaining U.S. leadership in AI.
"OpenAI published a nine-page blueprint asking Congress to enact a single federal framework for frontier AI and to preempt state laws that regulate 'the same frontier safety risks.'"
— Implicator.ai, analysis of the June 3 OpenAI policy release
The timing is notable: OpenAI's blueprint landed the day before Rep. Trahan and Rep. Obernolte released the House draft bill — suggesting a well-coordinated industry-to-Congress influence pipeline. Critics immediately noted that OpenAI's proposed framework would primarily benefit OpenAI by creating standards that frontier labs are better positioned to meet than smaller competitors, and by neutralizing the more aggressive state-level oversight regimes emerging in California, Illinois, and New York.
OpenAI also pointed to its GPT-Rosalind life sciences model — which received an update this week bringing stronger agentic coding capabilities and drug-discovery performance — and GPT-5.4 and GPT-5.5's availability on Amazon Bedrock in AWS GovCloud as evidence of its growing federal footprint.
OpenAI is doing something sophisticated here: simultaneously publishing safety-focused governance documents and lobbying for a federal framework that would neutralize its most aggressive regulatory threats. It's a classic "shape the standard before the standard shapes you" strategy. For enterprises, the practical implication is that a federal AI law — if and when it passes — will likely reflect OpenAI's preferred architecture: relatively permissive for model development, with governance focused on testing and disclosure rather than hard capability limits. That's a dramatically different world than a California SB-1047-style regime.
Sources: OpenAI, Blueprint PDF, Implicator.ai
7. Hyperscaler AI Capex: Goldman Sachs Raises Combined Estimate to $5.3 Trillion Through 2030
Separate from its SpaceX projections, Goldman Sachs's strategist Amanda Lynam published fresh hyperscaler capex figures that underscore just how much capital is flowing into AI infrastructure. Goldman now expects Meta, Microsoft, Amazon, and Alphabet to collectively spend $5.3 trillion on AI-related capex from fiscal year 2025 through fiscal year 2030 — up from a prior estimate of $4.5 trillion made before Q1 2026 earnings.
The baseline aggregate capex estimate across compute, data centers, and power infrastructure now stands at $7.6 trillion between 2026 and 2031. These numbers represent a breathtaking commitment of capital to AI buildout — and they explain why AI chip and power infrastructure stocks have remained resilient even as broader market volatility has increased.
"Goldman now expects a combined $5.3 trillion of capex spending for the four largest hyperscalers — Meta, Microsoft, Amazon, and Alphabet — from fiscal year 2025 to fiscal year 2030."
— Goldman Sachs, via Yahoo Finance / Amanda Lynam analysis
The revised figures come as Broadcom's latest earnings caused some turbulence in AI chip stocks, with investors parsing the difference between continued infrastructure investment and actual AI application revenue. The Goldman analysis suggests the hyperscalers have made a structural decision to build at scale regardless of near-term ROI uncertainty — treating AI infrastructure as a strategic asset rather than an incremental IT investment.
$5.3 trillion in hyperscaler AI capex through 2030 is not a bet being made on current AI ROI — it's a bet on a future where AI becomes as fundamental to enterprise operations as internet connectivity. For smaller enterprises still running AI pilots, this investment wave has a direct implication: the cost of AI inference will continue to fall as capacity grows, and the performance ceiling will continue to rise. The companies that start building AI-native workflows now will have a significant head start when that infrastructure comes fully online. Sitting on the sidelines waiting for "proof of ROI" is itself a strategic risk.
Sources: Yahoo Finance / Goldman Sachs analysis
Why It All Matters
Thursday, June 4, 2026 will likely be remembered as a day when the AI industry's governance, financial, and product stories all converged at once. The Anthropic Pentagon saga shows that even the most well-funded AI labs can face existential political risk. The House AI bill and OpenAI's blueprint together signal that federal AI regulation is no longer hypothetical — it's being actively shaped right now, and industry is in the driver's seat. Google Gemini Spark's launch marks the real beginning of the ambient AI agent era for consumers. Goldman's SpaceX and hyperscaler projections are a reminder that the capital commitments underlying all of this are almost incomprehensibly large. For business leaders: the regulatory environment is shifting fast, the product capabilities are accelerating, and the investment signals suggest this is not a bubble — it's a structural transformation. The question isn't whether to engage with AI. It's whether you're building the internal capability to use it before your competitors do.
Need help navigating AI for your business?
Our team turns these developments into actionable strategy — from AI policy compliance to agentic workflow deployment.
Contact SEN-X →