CNBC Scott Wapner with Brad Gerstner at Milken Global - May 4th, 2026

Key points

• **Mag5 revenue 8 quarters ago = $150B. Now = $350B.** Same number Brad cited on All-In this week but with the consolidated quarter-over-quarter delta. Mega-cap premium to S&P now only ~13% (Goldman) — multiples have not run away despite the growth. **The thesis: 'these are the most undervalued large caps in the history of the market relative to forward earnings power.'**
• **'Nvidia will be the first $10 trillion company' — Brad's explicit date-stamped call.** As of May 2026 valuation at ~$195/share, **13-14x fully-taxed GAAP earnings on $1 trillion in forward orders**. Trading *lower* than 6 months ago despite the order book. 'Terribly under-owned, terribly undervalued.' Vera Rubin LPX clusters coming later this year — comes from the Groq acquisition late last year. 'Another huge step forward for the models.'
• **Memory at 5x earnings = the lowest-conviction-needed trade Brad's pitching.** 'Samsung will do more in profits this year than Google.' Why the 5x multiple? 'Boom-and-bust fear.' If you believe earnings are durable through 2028-29, these are 'ridiculous values.' **Direct echo of his All-In disclosure this week (25% of portfolio in SKH/Samsung/Micron) and cross-references the Apple Mac Mini stealth-inflation story on 20VC.**
• **'Token per watt per dollar' as Nvidia's true moat metric.** 'Even if they're giving up a few share points to Trainium or TPUs, Nvidia is still the best in the business.' GPT-5.5 trained on new Blackwell chips. Vera Rubin shipping later this year. **Useful framing for evaluating chip-architecture moats going forward — focus on $/watt of inference rather than peak FLOPS.**
• **Anthropic vs OpenAI verdict: 'Team America wins, multi-horse race.'** Anthropic took lead Jan-Apr via Claude Code + Opus 4.6 (early Dec release driving enterprise adoption). **But Codex 'incredible in last 3-4 weeks' based on X sentiment.** OpenAI 'distributing all of their models across AWS' which will accelerate them. 'On enterprise side just as cloud (Amazon/Google/Microsoft), profits will split. On consumer side, OpenAI not yet at 1B WAU but >900M.' Confirms Issue 03 Sacks framing.
• **SpaceX + X IPOs at end of June 2026** — direct date stamp from Brad. 'First one out of the gates. Elon is an end-of-one. Most people talking about Anthropic and OpenAI but Elon is the first.' Cross-references this week's All-In on Elon-Colossus-Anthropic deal and the $4-5B incremental SpaceX revenue from Elon Web Services.
• **Anthropic + OpenAI already had 'private IPOs.'** OpenAI raised **$110-120B in most recent round** with cap table that 'looks like a day-1 IPO cap table' (Fidelity, Capital Group, Altimeter, Coatue). 'They're not capital-constrained.' Implication: they don't *have* to go public this year — and **demand for Anthropic shares is 'some of the highest I've seen in 25 years.'** Direct echo of Lemkin's framing this week that the IPO timeline could slip to 2027.
• **Reality check on retail IPO returns.** 'Buying a company at $1 trillion is not a get-rich-quick scheme. You're hoping to compound at 20-30% for a long time. If hoping to double or triple, it's unlikely.' Honest caveat from the public-markets veteran — 'companies are staying private longer; more value capture is occurring in private markets; we need to address this as a country.'
• **Trump Accounts launch July 4 2026 — single largest social contract transformation since Social Security.** $1,000 stapled to Social Security number at birth, invested in S&P 500. **Math: $1k + $50/month = $50k at 18, $200k at 30, $1M at 55.** Update from Brad: **5M kids signed up already, 10M expected by July 4, automatic at birth from 2027** (3.7M kids/year). Many large employers committed to 401k-like contributions to employees' kids (Uber, Nvidia, Salesforce, AMD). $6.25B already donated by Michael Dell + Susan + others. **'60-70% of people in this country just don't participate in compounding. Now we're going to change all of that.' This is the single highest-conviction policy claim from any guest this week.**
• **Mark Zuckerberg + Susan Lee at Meta still in 'flatter is faster' execution mode.** 'AI is bringing incredible efficiencies. The S&P 500 earnings increase is coming substantially from margin expansion. We'll see this for next 3 years.' Cross-references this week's 20VC Mag7 episode on Meta capex anxiety.
• **Iran / Strait of Hormuz as the tail risk.** $120 oil 'would take its toll if not solved in 3-4 months.' Only explicit geopolitical risk Brad flags as material.
• **Profit-taking advice for retail.** When a stock goes up 100% in a month (AMD +57%, Broadcom +31% recently): 'You shouldn't feel bad about taking a little profitability. This game is either hold-it-forever-and-let-it-compound OR pull a little profit off when the market is getting ahead of itself.' Implicit endorsement of dollar-cost-averaging out of overshot positions while keeping the core long.

Notable quotes

Themes

• Nvidia as the first $10 trillion company — Brad's explicit date-stamped call
• Memory at 5x earnings as the lowest-conviction-needed bet
• Anthropic-OpenAI as a multi-horse race not a winner-take-all
• SpaceX + X IPO at end of June 2026 (first out of the gates)
• Trump Accounts as the biggest social-contract transformation since Social Security