⚠ MAY 2026 — INVESTOR BRIEFING

AI Is Getting Really Serious

In 30 days the AI industry restructured itself. Frontier models started training their own successors. The corporate map looks nothing like it did at the start of the year. Most portfolios haven't repriced.

If your AI investing thesis is still "buy NVDA and the Mag 7" — or its opposite, "avoid AI, it's a bubble"both views are out of date by about two product generations. Here's the structural shift, the strategic implication, and the math.

Section 1 · The Real Headline

AI Has Started Building AI — And The Lab Equity Math Changes

The most important thing that happened in early 2026 was not a new chatbot. It was that AI systems became good enough at coding, fine-tuning, and optimization to materially speed up the production of the next generation of AI systems.

📊 The Benchmark Mosaic

Four AI-doing-AI benchmarks that flipped in 24 months:

  • SWE-Bench (real-world coding): ~2% (late 2023) → 93.9% (2026). Solved.
  • CORE-Bench (reproducing scientific papers): 21.5% (Sep 2024) → 95.5% (Dec 2025). Solved.
  • MLE-Bench (building ML systems from scratch): 16.9% (Oct 2024) → 64.4% (Feb 2026).
  • PostTrainBench (fine-tuning other models): AI now at half the uplift of human experts at frontier labs.

Plus: AI systems can now optimize their own training code by 52x. Human researcher baseline: 4-8 hours to achieve 4x. AI is ~13x past human at making AI faster.

The investment thesis splits cleanly into two questions.

💰 Question 1: Lab Equity

If AI can train its own successor by 2027-2028 (and senior safety researchers are estimating 30-60% probability), then OpenAI / Anthropic / xAI / DeepMind equity becomes a pure option, not a DCF. The "lab can't recoup capex" argument becomes irrelevant — because in the RSI scenario, the lab IS the asset. In the no-RSI scenario, the lab is a bagholder bet.

Either way: don't size lab equity like a growth stock. Treat it like venture, expect 80%+ underperformance, accept binary outcomes.

💼 Question 2: AI Deployment

Independent of lab equity, AI deployment is happening — measurably, across industries. That's a different bet, and a much more pricing-able one. The matrix below scores 28 industries across 5 horizons on AI repricing. It is not waiting on RSI.

Section 2 · The Corporate Map Redrew

Five Alliances In 30 Days. None Were Predictable.

Between April 27 and May 7, 2026, the AI industry produced more strategic restructuring than the previous five years combined. Shared driver: compute pressure, capex pressure, and a race to define the industry standard before lockups expire.

🤝 #1 — Anthropic + Elon Musk (May 6)

Anthropic locked up all 220,000+ GPUs at SpaceX's Colossus 1 (300MW, Memphis). Musk had publicly accused Anthropic of "massive-scale training data theft" months earlier. May 7: Musk dissolved xAI as an independent company, folded everything into SpaceX. Anthropic now buys compute from its biggest critic. Vertical integration: rocket + compute + LLM under one roof.

🤝 #2 — Microsoft + OpenAI Restructured (April 27)

Microsoft lost: AI exclusivity, the AGI clause, Azure-only revenue. Microsoft kept: 20% revenue cut through 2030 + $250B OpenAI Azure spending commitment through 2032. Next day: OpenAI live on AWS Bedrock. The most exclusive partnership in AI is now multi-cloud. Microsoft AI-passthrough thesis softens substantially.

🤝 #3 — OpenAI + Private Equity (May 4)

$10 billion JV ("The Deployment Company") with TPG, Bain, Brookfield, Advent, SoftBank, Dragoneer + 13 others. Catch: OpenAI promised PE backers 17.5% guaranteed annual return for 5 years. The play: top-down AI mandates inside PE portfolio companies, bypassing the corporate immune system (MIT NANDA "95% of enterprise AI projects fail").

🤝 #4 — Anthropic + Wall Street (May 4)

$1.5 billion AI services firm with Blackstone, Goldman, Hellman & Friedman, Apollo, GIC, Sequoia. Target: compete with McKinsey/Bain/BCG for AI transformation revenue. AI lab as consulting firm, financed by Wall Street, eating the consulting industry.

🚫 The Anti-Bedfellow — Pentagon Banned Anthropic (May 1)

The Pentagon awarded frontier-AI contracts to 8 companies (AWS, Google, Microsoft, OpenAI, SpaceX, NVIDIA, Reflection, Oracle). Anthropic was excluded — labeled a "supply chain risk" because they refused autonomous-weapons and mass-surveillance terms. Anthropic sued. A federal judge blocked the exclusion. Same week, Anthropic locked up Colossus and grew revenue past OpenAI. Worst-case treatment, best-case operational result.

Each alliance is a leading indicator. Watch who allies with whom over the next 6 months — that tells you where the compute and capital are actually flowing, before the earnings reports do.

Section 3 · The Tell

OpenAI's Own CFO Wants To Delay The IPO To 2027

If frontier AI were "all working" on the financial math, you'd expect the lab in the lead to charge straight at an IPO. Instead, OpenAI's CFO Sarah Friar has privately argued for delaying from 2026 to 2027. The reasons, per multiple reports:

$600B
OpenAI future compute commitments
900M
WAU vs 1B target (missed)
$25B
OpenAI ARR vs Anthropic $30B

Plus: the CFO says the company "doesn't yet meet rigorous reporting standards" for public corporations. Sam Altman is still publicly pushing 2026. Wall Street banks are reportedly telling both labs "first to market defines the industry" — creating pressure to rush anyway.

💡 Why It Matters For Your Portfolio

Our AI IPO Cliff dashboard models the supply flood from SpaceX, OpenAI, and Anthropic IPOs hitting lockup expiry. If OpenAI slips to 2027, the wave timing shifts — but the wave still hits. The bagholder-distribution thesis gets reinforced, not weakened.

When the CFO — the person whose actual job is making the IPO happen — applies the brakes, that's a signal worth pricing in.

Section 4 · The Operational Winner

Google Quietly Posted The Strongest AI Quarter Of The Mag 7

While OpenAI was renegotiating its plumbing and Anthropic was buying Musk's GPUs, Google was just printing money.

$109.9B
Alphabet Q1 2026 revenue (+22% YoY)
+63%
Google Cloud growth (beats AWS/Azure)
$460B
Cloud backlog (doubled QoQ)
32.9%
Cloud op margin (from 9.4% one year ago)

11th consecutive quarter of double-digit growth. AI-targeting turning every model improvement into ad revenue. DeepMind publicly admitted Google is internally compute-constrained — Search, Cloud, and DeepMind are fighting each other for new GPU capacity. That's a demand signal, not a capex problem.

The "Google as under-priced AI winner" thesis is no longer hypothesis. It's reported earnings.

Section 5 · The Industry Signal Inside The Labor Data

The Displacement Curve Is Asymmetric

The May 2026 NPWF study breaks the aggregate-labor framing apart:

47%
Women's share of US workforce overall
83%
Women's share of workers in 15 most AI-vulnerable jobs
-14%
Drop in under-25 hiring in AI-exposed roles (Anthropic Economic Index)

For investors, the labor story matters because industries with high clerical/administrative cost share are the ones where AI margin expansion shows up first. The displacement is concentrated where automation is cheapest and least visible. That's the signal under the social cost — and it's where your matrix scoring identifies repricing first.

Get The Scores, Not The Vibes

The AI Stocks engine scores 28 industries across 5 horizons (1yr, 2yr, 3yr, 5yr, 10yr) using 8 analytical dimensions and 167 cross-industry effects. It tells you which industries get repriced upward, which get repriced downward, and by how much — independent of whether OpenAI's equity holds.

See the Matrix Scores Free Portfolio Scan

Related Free Reports

The Trillion-Dollar Question Wall Street Is Asking Backwards — why the bear case (lab equity) and the bull case (AI deployment) are both correct, and how sophisticated capital is actually pricing the split.

The Bagholder Sanity Check — six cognitive biases that are picking your AI stocks for you, with a self-diagnostic for each.

The Controversy Premium — when moral panic becomes a moat, and why the "evil" stocks tend to outperform.

The AI IPO Cliff — interactive dashboard modeling the SpaceX/OpenAI/Anthropic IPO supply flood and lockup expiries.