“History is being made on two fronts today.” That is not a press release. That is the lead of the most sober AI news tracker in the market, writing about June 12th 2026. SpaceX began trading. OpenAI filed its S-1. WWDC closed. And the Magnificent Seven collectively lost more wealth in a month than most countries produce in a year. That is not noise. That is signal.”
Ground Truth · Episode 07 · June 2026Three days ago, on June 12th 2026, SpaceX began trading on the Nasdaq under the ticker SPCX at $135 per share. The company’s opening market capitalisation made it the most valuable company ever to list on a public exchange at debut — larger than Apple was at its IPO, larger than Saudi Aramco, larger than anything the public markets have absorbed in a single listing. The Goldman Sachs syndicate that managed the offering spent three weeks taking institutional orders across New York, Boston, and San Francisco. When the books closed, demand had exceeded supply by a factor that Goldman described as “extraordinary.” The stock opened. The history books updated. And simultaneously, the Magnificent Seven — Apple, Microsoft, Nvidia, Amazon, Alphabet, Meta, and Tesla — collectively shed approximately $2 trillion in market capitalisation over the course of June as institutional investors rotated out of established tech positions and into AI IPO allocations. Welcome to Episode 07 of Ground Truth. Today we map what is actually moving and why.
The Largest IPO in History and the Index Battle That Follows
What traded on June 12th. SPCX represents SpaceX consolidated with Elon Musk’s xAI, which was absorbed into SpaceX in February 2026 at a combined valuation of $1.25 trillion. The listed entity therefore includes Starlink — now reaching over 10 million subscribers across 160 countries — the Falcon and Starship launch systems, the Grok AI model family, and the xAI research operation. It is not a pure AI play. It is a diversified technology infrastructure company with a significant AI division, which is precisely what institutional investors wanted: AI exposure with cash-flowing infrastructure underneath it.
The index battle. The S&P 500 committee reviewed SpaceX for fast-track inclusion and declined — citing profitability requirements that the company meets but also concerns about concentration and the pace of market cap addition relative to existing index constituents. The Nasdaq, which amended its inclusion rules in May 2026 specifically to accommodate megacap IPOs, has opened a 15-trading-day fast-entry window. Based on the June 12th listing date, SPCX becomes eligible for Nasdaq-100 inclusion on approximately July 7th. The Nasdaq-100 is tracked by over 200 investment products with more than $600 billion in assets under management, including the widely held QQQ ETF. When SPCX enters the index, every product tracking the Nasdaq-100 is required to buy it — a forced purchase event that will move the stock regardless of investor preference. This is the mechanical force underneath the July 7th date.
What Polymarket says. The June listing probability for SpaceX was priced at 65.5% on Polymarket going into the week of June 12th. It happened. The next Polymarket question the AI market is watching: the probability of Anthropic listing before year-end, currently priced at 71%, and OpenAI listing in 2026, priced at 58%. Both companies filed confidential S-1s within days of each other. The IPO race is not a metaphor. It is a dated calendar item.
SpaceX (SPCX): Listed June 12 at $135/share. Largest IPO in recorded history. Nasdaq-100 eligible ~July 7. $600B+ in index-tracking assets must buy on inclusion. Anthropic: Confidential S-1 filed. $965B valuation. First operating profit Q2 2026 (~$559M). IPO probability before year-end: 71% (Polymarket). OpenAI: Confidential S-1 filed June 8. Targeting up to $1T valuation. $25B annualised revenue. Loss of $14B projected in 2026. IPO probability 2026: 58% (Polymarket). The Magnificent Seven: down approximately $2 trillion combined in June. Goldman Sachs projection: 2026 IPO proceeds could reach $160 billion total, 4x 2025 levels.
Where the Money Is Going and What It Is Leaving Behind
The $2 trillion that left the Magnificent Seven in June did not evaporate. It rotated — from established AI-adjacent tech into direct AI exposure through SpaceX, Anthropic, and OpenAI IPO allocations, and into the secondary market positions that investors took in anticipation of those listings. This is a classic late-cycle rotation: investors who rode the Magnificent Seven to extraordinary returns since 2020 are now taking profits and redeploying into the companies they believe represent the next chapter of the AI trade.
The rotation has winners and losers that are not evenly distributed across the seven. Nvidia, which has been the purest beneficiary of AI infrastructure spending, has held up better than the consumer-facing companies like Apple and Meta, whose AI offerings have been slower to generate direct revenue. Microsoft, whose Azure AI revenue is growing rapidly but whose relationship with OpenAI has become structurally complicated, sits in the middle. The Magnificent Seven is not a monolith — it is seven companies with different AI exposure profiles, and the rotation is revealing those differences.
The more uncomfortable question, which institutional investors are asking privately and which the financial press is beginning to address, is whether the $2 trillion rotation is rational or whether it represents exactly the kind of speculative reallocation from proven cash flows to future-value stories that precedes corrections. The companies being sold — Apple, Microsoft, Nvidia — are profitable, cash-generating businesses with real AI integration. The companies being bought into — Anthropic, OpenAI — are loss-making at different stages, dependent on continued capital investment, and subject to competitive pressure from every direction simultaneously. The rotation from one to the other is a bet that the future-value story is worth more than the current cash flow. History has a view on that bet. It is not consistently positive.
$2 trillion left the Magnificent Seven in June. It did not disappear. It rotated into the IPO pipeline for companies that are, by conventional accounting, losing money. The investors making that rotation are not irrational — the growth trajectories justify extraordinary valuations. They are making a specific bet about which kind of value creation wins over the next five years. The history of that bet in previous technology cycles is mixed.Neal Lloyd · Ground Truth, Episode 07
Anthropic Is Paying Elon Musk $15 Billion a Year to Run Its AI
Buried in the footnotes of the SpaceX IPO coverage is a number that deserves considerably more prominence than it has received. Anthropic — the AI safety company founded partly in response to concerns about concentrated power in AI — has committed to paying SpaceX $1.25 billion per month for compute infrastructure through May 2029. That is $15 billion per year. Over the remaining term of the contract, it is approximately $45 billion going to Elon Musk’s company.
The compute in question is provided through SpaceX’s Colossus supercomputer facility in Memphis, Tennessee — the same facility that xAI uses for Grok model training. Anthropic’s models run on hardware owned and operated by a company whose CEO is simultaneously building competing AI models, operating a competing AI platform, and now publicly traded in a way that makes Anthropic’s compute dependency a disclosed material risk to any investor in SPCX who also reads Anthropic’s S-1. The dependency is not secret. It is, however, striking.
To contextualise the figure: $15 billion per year in compute costs represents approximately 32% of Anthropic’s $47 billion revenue run rate. A third of revenue flowing to a single compute provider who is also a direct competitor is a concentration risk that any institutional investor reading an S-1 will flag immediately. Anthropic has been diversifying its compute across AWS and other providers, but the SpaceX contract represents a commitment that cannot be quickly unwound. The Anthropic IPO prospectus will need to address this dependency with considerable care.
While the market focuses on the IPO race between American frontier labs, Alibaba’s Qwen 3.7 Max is quietly attracting serious developer attention. On the Artificial Analysis Intelligence Index, Qwen 3.7 Max scores within striking distance of Claude Opus 4.7 on reasoning tasks — at roughly half the input cost and a quarter of the output cost. For developers building AI-native products, a model that approaches frontier capability at dramatically lower cost is not a footnote. It is a pricing pressure event. If Qwen continues this trajectory, it forces American frontier labs into a price war they were not planning to fight while simultaneously preparing for public listings at valuations that assume premium pricing power. The Chinese model threat is not theoretical. It is in production, it is benchmarked, and it is cheaper.
June 2026 Is the Most Active Model Release Month of the Year. Here Is the State of Play.
Claude Opus 4.8 shipped May 28th with a 1 million token default context window, Dynamic Workflows for complex multi-step agentic tasks, a 3x cheaper Fast Mode, and reclaimed the number one position on the coding benchmark leaderboard. For Anthropic heading into its IPO, Opus 4.8 is a critical product statement: capability leadership maintained, cost structure improving, enterprise positioning strengthened.
GPT-5.5 shipped April 24th with Instant, Thinking, and Pro modes covering the full range of speed-cost-capability trade-offs. Now generally available in Microsoft Foundry as of June 3rd — the same day Microsoft announced its own MAI models, making Azure simultaneously the distribution channel and the competitor for GPT-5.5. The competitive geometry is genuinely unusual and the Microsoft-OpenAI S-1 disclosures will need to address it directly.
Gemini 3.5 Flash shipped at Google I/O in late May with 284 tokens per second throughput and pricing of $1.50 per million input tokens — aggressive on speed and price, designed to capture the high-volume production workload market that OpenAI and Anthropic have been dominating. Google’s strategy is increasingly clear: compete on efficiency and price rather than raw capability, where its cloud infrastructure advantages are most relevant.
Qwen 3.7 Max from Alibaba is the sleeper story. Frontier-adjacent capability, half the input cost of Claude Sonnet 4.6, a quarter of the output cost. For developers building production applications where token costs matter — and at scale they always matter — Qwen 3.7 Max is a real alternative. The developer attention it is attracting in June 2026 is the early signal of a pricing pressure that will arrive in earnest over the next twelve months.
Anthropic is heading to a near-trillion-dollar IPO while paying $15 billion a year to a competitor for compute, racing to maintain model leadership against a Chinese alternative at half the price, and operating in a regulatory environment that could force material changes to its most valuable features before it lists. That is not a simple story. It is a complicated one, and the S-1 will need to tell it honestly.Neal Lloyd · Ground Truth, Episode 07
Ground Truth, Episode 07 · June 15 2026
Neal Lloyd covers the real-world impact of AI — money, power, geopolitics, and the stories behind the headlines. Ground Truth is his daily AI news and analysis series on emdexter.blogspot.com.
- Episode 01The Gold Rush$3.6T to Wall Street
- Episode 02ChatGPT Knows EverythingDreaming V3 and the 53-day deadline
- Episode 03Siri Is Now GoogleWWDC 2026 and the EU lockout
- Episode 04America’s AI Law Is a MessThe 50-state patchwork crisis
- Episode 05Is AI Taking Your Job?The real data on displacement
- Episode 06Microsoft vs EveryoneMAI models and the alliance that became a rivalry
- Episode 07SpaceX Is TradingYou are here
