Anthropic hired OpenAI's co-founder. Mercury raised $200M (profitable 4 years). Aschenbrenner shorts Nvidia.

Good morning!

Four stories. One theme: The AI war is reshaping everything: talent, capital, markets, and infrastructure.

The intel:

Andrej Karpathy, OpenAI co-founder and Tesla's former AI chief, joined Anthropic this week, a major coup in the escalating talent war. Mercury, the fintech serving one in three U.S. startups, raised $200M at $5.2B on the back of the AI founder boom. Polymarket partnered with Nasdaq to let retail investors bet on private company valuations (OpenAI, Anthropic, SpaceX). And Leopold Aschenbrenner, ex-OpenAI researcher, disclosed $7.5B in bearish bets against Nvidia and chip stocks, while going long on Bitcoin miners and AI infrastructure.

In today's private markets rundown:

• Karpathy joins Anthropic: The talent war heats up

• Mercury $200M Series D: Banking the AI boom

• Polymarket launches private company trading

• Aschenbrenner shorts chips, longs power infrastructure

• This week's funding & M&A recap

• The Big Picture: What this all means

☕

Jose, Alberto, and the Founderscrowd team.

🎁 We got you $100 in credits. Redeem now.

Attio is the AI CRM for high-growth teams.

Connect your email, calls, product data and more, and instantly builds your CRM with enriched data and complete context. Whether you’re running product-led growth or enterprise sales, Attio adapts to your unique GTM motion.

Then to plan your next move.

Run deep web research on prospects. Update your pipeline as you work. Find customers and draft outreach emails. Powered by Universal Context, Attio's intelligence layer, Attio searches, updates, and creates across your data to accelerate your workflow.

Ask more from your CRM.

🧠 ANDREJ KARPATHY

OpenAI Co-founder joins Anthropic in major talent coup

What happened:

Andrej Karpathy, one of the original 11 co-founders of OpenAI and former head of Tesla's AI division, announced Tuesday (May 19) that he's joining Anthropic.

Who is Karpathy:

• Age: 39, Slovak-Canadian AI researcher

• Background: PhD in computer science from Stanford

• Career:

  • Co-founded OpenAI (2015)

  • Left for Tesla as Director of AI (2017)

  • Returned to OpenAI briefly

  • Founded Eureka Labs (AI education startup)

  • Now: Anthropic (May 2026)

His role at Anthropic:

Joining the pre-training team under team lead Nicholas Joseph.

Pre-training = the large-scale training runs that give Claude its core knowledge and capabilities. It's the most expensive, compute-intensive phase of building a frontier model.

Karpathy will lead a new team focused on using Claude to accelerate pre-training research—essentially, teaching Claude to improve itself.

This is the path to recursive self-improvement (RSI), the holy grail of AI research.

The post:

"Personal update: I've joined Anthropic. I think the next few years at the frontier of LLMs will be especially formative. I am very excited to join the team here and get back to R&D. I remain deeply passionate about education and plan to resume my work on it in time."

The post drew nearly 3 million views within an hour.

Why this matters:

1. The talent war is intensifying.

OpenAI has suffered significant brain drain. Anthropic was founded by former OpenAI employees (Dario and Daniela Amodei). Now Karpathy—an original co-founder—has joined them.

This signals:

• Anthropic is winning the talent war

• OpenAI's internal culture may be deteriorating

• Researchers are choosing Anthropic's approach over OpenAI's

2. Anthropic is emerging as the magnet for elite AI minds.

Recent hires:

• Karpathy (OpenAI co-founder, Tesla AI lead)

• Ross Nordeen (xAI founding member, ex-Tesla, joined May 2026)

• Chris Rohlf (cybersecurity veteran, Yahoo's "The Paranoids," joined May 2026)

Anthropic is building a dream team.

3. This comes days after the Musk-Altman trial.

The trial (concluded May 18) aired OpenAI's dirty laundry. Elon Musk sued Sam Altman claiming OpenAI violated its nonprofit mission. Musk lost.

But the trial revealed internal conflicts, Musk's attempts to control OpenAI, and questions about OpenAI's governance.

Karpathy's move—announced May 19, one day after the verdict—feels deliberate.

4. Karpathy bridges theory and practice.

He's one of the few researchers who can bridge the gap between LLM theory and large-scale training practice. That's rare.

Anthropic just hired the best person in the world to make Claude smarter faster.

Bottom line:

Karpathy joining Anthropic is a massive win for the Claude team and a blow to OpenAI. The talent war is no longer theoretical. Anthropic is winning.

And if Karpathy can successfully build Claude to improve itself? That's the end game.

💳 MERCURY

$200M Series D at $5.2B: Banking the AI founder boom

What happened:

Mercury, the fintech providing banking for startups, raised $200 million Series D at a $5.2 billion valuation on May 20.

The round:

• $200M raised

• $5.2B valuation (up 48.6% from $3.5B Series C in March 2025)

• Led by: TCV

• Investors: Andreessen Horowitz, Coatue, CRV, Sapphire Ventures, Sequoia Capital, Spark Capital

• Total funding: ~$700M (primary + secondary)

The numbers:

• $650M annualized revenue (Q3 2025)

• 300,000+ customers (including 1 in 3 U.S. startups)

• 2.5x increase in applications (Q1 2026 vs Q1 2025)

• 4 consecutive years of GAAP net income and EBITDA profitability

The thesis:

CEO Immad Akhund: "AI is collapsing the friction between an idea and a company faster than anything I have seen in my career. We are going to see more founders in the next five years than in the last twenty."

Translation: AI is creating the largest wave of entrepreneurs in a generation. Mercury wants to be their bank.

Why the growth:

1. Entrepreneurship is surging.

Q1 2026 saw an 18% increase in new U.S. business applications vs Q1 2025 (U.S. Census Bureau).

AI is making it easier to start companies. Lower barriers to entry = more founders = more Mercury customers.

2. Mercury owns the startup banking market.

One in three U.S. startups banks with Mercury. That's dominant market share.

Customers include: Supabase, ElevenLabs, Lovable, Linear, Phantom, Tempo.

3. They're expanding beyond banking.

New products launched:

• Mercury Insights: In-product AI tool for real-time financial health (no spreadsheets)

• Model Context Protocol (MCP): Secure banking access for AI developers

• Direct API actions: AI agents can take banking actions programmatically

Mercury is positioning itself as AI-native banking infrastructure.

4. They got a bank charter.

April 2026: Mercury received conditional approval for a U.S. national bank charter from the Office of the Comptroller of the Currency (OCC).

This is huge. It means Mercury can operate as a full bank, not just a fintech partner.

More control. More products. More revenue.

Why this matters:

1. The AI founder boom is real.

Mercury's 2.5x growth in applications proves it. AI is creating more founders. Those founders need banking.

2. Fintech funding is back.

After a brutal 2023-2024, fintech is hot again. Mercury, Stripe, and others are raising at premium valuations.

Investors believe the AI wave will create massive demand for financial infrastructure.

3. Profitability matters.

Mercury has been profitable for 4 consecutive years (GAAP net income + EBITDA).

In a market that punished unprofitable growth companies, Mercury's profitability is a competitive advantage.

4. Mercury is building for AI-first companies.

MCP integration. AI Insights. API-driven banking. Mercury isn't just banking startups—it's banking AI-native startups.

That's the future.

Bottom line:

Mercury raised $200M at $5.2B because they're the default bank for the AI founder boom. One in three U.S. startups banks with them. They're profitable. And they're building AI-first infrastructure.

This is what winning looks like.

🎲 POLYMARKET

Retail can now bet on private company valuations

What happened:

Polymarket launched prediction markets tied to private company milestones on May 19, in partnership with Nasdaq Private Market.

What you can bet on:

• Valuations: Will OpenAI hit $1T by 2027?

• IPO timing: Will Anthropic IPO before OpenAI?

• Secondary market activity: Will Anthropic be worth more than OpenAI in 2026?

Companies covered:

OpenAI, Anthropic, SpaceX, Stripe, Databricks, Anduril, Kraken, Ripple.

How it works:

• Polymarket: Platform for prediction markets

• Nasdaq Private Market (NPM): Provides valuation data and resolution data

• Event contracts: Binary bets (Yes/No) on specific outcomes

You're not buying equity. You're betting on outcomes.

Current markets:

• 88% chance Anthropic hits $1 trillion valuation by year-end

• 94% chance Anthropic will be worth more than OpenAI in 2026

• 69% chance Anthropic IPOs before OpenAI

Why this is significant:

1. Retail gets exposure to private markets.

For decades, private companies were off-limits to retail investors. Only accredited investors, institutions, or the well-connected could access them.

Now, anyone can bet on private company outcomes.

2. Nasdaq Private Market makes it legit.

NPM provides the data. That's not some random oracle—it's Nasdaq.

NPM already runs secondary share trading and tender offers for late-stage startups. They have the data. They know the valuations.

3. This creates price discovery.

Polymarket odds = market consensus on valuations, IPO timing, and competitive dynamics.

Example: 94% chance Anthropic is worth more than OpenAI in 2026 = the market believes Anthropic is winning.

4. It's a signal of where private markets are headed.

More transparency. More access. More liquidity.

If Polymarket succeeds, expect competitors (Kalshi, TradeXYZ) to copy the model.

The controversies:

1. Is this gambling or investing?

You're betting on outcomes. You don't own equity. You don't get dividends. You don't vote.

Critics: This is gambling dressed up as finance.

Defenders: This is price discovery. Markets are betting mechanisms. This is no different.

2. Market manipulation risks.

With $39B in U.S. volume YTD (2026), Polymarket has liquidity. But it's still vulnerable to manipulation.

Example: A whale could bet heavily to move odds, then use those odds to influence fundraising or secondary market prices.

3. Regulatory scrutiny.

The SEC hasn't weighed in yet. But betting on private company outcomes via crypto-based prediction markets? That's going to attract attention.

Why this matters:

1. Democratization of private markets.

This is what Founderscrowd is all about. Giving retail investors access to deals they couldn't touch before.

Polymarket isn't equity, but it's exposure. And exposure = participation.

2. It validates the narrative.

Private companies staying private longer. More than 1,600 unicorns valued at $5T+ combined.

Retail wants in. Polymarket gives them a door.

3. It creates a new asset class.

Private company event contracts = new financial instrument.

If this scales, it becomes a permanent feature of capital markets.

Bottom line:

Polymarket launched private company trading, letting retail bet on OpenAI, Anthropic, and SpaceX valuations and IPO timing. It's not equity. But it's exposure. And it's democratizing access to private markets.

The walls are coming down.

💎 WHAT FOUNDERSCROWD PREMIUM MEMBERS SAW

MARCH 2025: WE PROFILED CEREBRAS

What we told Premium members:

• Cerebras was trading on secondaries at $102-$107

• IPO expected in Q2 2026 at $150-160 range

• $510M revenue, $10B OpenAI contract

• Strong buy for investors who want AI infrastructure exposure

What happened:

• IPO'd at $185 (23% above high-end estimate)

• Popped to $311 on day one (68% gain)

• Currently trading at ~$280

If you'd bought at $105 in March 2025:

You're up 166% in 14 months. Annualized return: 142%.

If you'd bought at IPO ($185):

You're up 51% in 8 days.

📉 LEOPOLD ASCHENBRENNER

Former OpenAI researcher shorts Nvidia, longs Bitcoin miners

What happened:

Leopold Aschenbrenner, a former OpenAI alignment researcher, disclosed that his hedge fund Situational Awareness LP has placed $7.5 billion in bearish bets against AI chip makers.

Who is Aschenbrenner:

• Age: 24 (as of May 2026)

• Background:

  • Graduated valedictorian from Columbia at 19 (dual major: economics and mathematics)

  • Joined OpenAI (2023) on the Superalignment team

  • Fired from OpenAI for leaking safety concerns

  • Published "Situational Awareness: The Decade Ahead" (June 2024, 165-page manifesto on AGI)

  • Launched Situational Awareness LP (hedge fund, 2024)

The portfolio (Q1 2026 13F filing):

Total disclosed equity exposure: $13.67 billion (up from $5.5B at end of 2025)

Short positions (bearish puts):

• $2B+ in bearish positions on VanEck Semiconductor ETF (SMH)

• $1.56B in put options on Nvidia (NVDA)

• Large puts on: Oracle (ORCL), Broadcom (AVGO), AMD, Taiwan Semiconductor (TSM), ASML

Long positions:

• Bitcoin miners: IREN, Core Scientific, Riot Platforms, CleanSpark, Bitfarms, Bitdeer, Hive Digital

• AI infrastructure: Bloom Energy, CoreWeave

• Power and capacity: Companies supplying electricity and data center capacity for AI

The thesis:

Aschenbrenner believes:

1. Chips are overvalued. Nvidia, AMD, Broadcom = too much hype.

2. The real bottleneck is power. AGI requires massive electricity. The constraint isn't chips—it's electrons.

3. Bitcoin miners own the infrastructure. They control electricity, data centers, and compute capacity. That's the next layer of value.

From his manifesto: "AGI could arrive as early as 2027, with the real constraint being physical infrastructure, above all, electricity."

Why this is contrarian:

1. He's shorting the stocks everyone loves.

Nvidia is trading at 65% odds to finish 2026 as the world's largest company by market cap (per Polymarket).

Wall Street analysts are bullish. Nvidia earnings (May 20) were strong.

Aschenbrenner is betting against the consensus.

2. He's longing Bitcoin miners.

Bitcoin miners aren't sexy. They're viewed as commodity businesses tied to Bitcoin price.

But Aschenbrenner sees them differently: They own power infrastructure and data centers. That's what AGI needs.

3. He's betting on a shift in value.

The AI trade has been: Buy Nvidia, AMD, Broadcom (the chip layer).

Aschenbrenner is saying: The value is moving downstream—to the power and capacity layer.

The evidence supporting his thesis:

1. TeraWulf (Bitcoin miner):

Q1 2026: AI and high-performance computing hosting revenue ($21M) surpassed Bitcoin mining revenue ($13M) for the first time.

2. Core Scientific:

Announced plans to convert its Pecos site into a 1.5-gigawatt AI data-center campus, repurposing 300 megawatts of existing mining capacity.

3. Power demand is exploding.

AI training runs require massive electricity. Data centers are constrained by power availability, not chip availability.

The market's response:

Wall Street is skeptical.

Nvidia's stock is up. AMD is up. The semiconductor trade is still hot.

Polymarket: Only 23% chance of "AI bubble burst by end of 2026." (One trigger: Nvidia falling 50% off all-time high.)

But Aschenbrenner has a track record.

His fund grew from a few hundred million to $13.67B in under 18 months. He's raised capital from Patrick Collison (Stripe), Nat Friedman, and Daniel Gross.

He's not a random bear. He's a former OpenAI insider with a thesis.

Why this matters:

1. It's a barbell strategy.

Long on power/infrastructure (the bottleneck). Short on chips (overvalued). Hedge against a chip correction while capturing the infrastructure boom.

2. It signals where the AI trade is headed.

If Aschenbrenner is right, the next wave of AI value creation isn't in chips—it's in power, data centers, and compute capacity.

3. It's a bet on AGI timelines.

Aschenbrenner believes AGI arrives by 2027. If he's right, power becomes the constraint. Bitcoin miners own the power. That's the trade.

4. It creates a short-term vs long-term tension.

Short-term: Nvidia and chip stocks are hot.
Long-term: Power infrastructure is the bottleneck.

Aschenbrenner is betting on the long-term shift.

Bottom line:

Leopold Aschenbrenner, ex-OpenAI researcher, is shorting $7.5B worth of Nvidia and chip stocks while longing Bitcoin miners and power infrastructure. His thesis: The AI trade is moving from chips to power. If AGI arrives by 2027, electricity is the constraint.

Contrarian? Yes. Crazy?

Maybe. But he's put $13.67B behind it.

📊 THIS WEEK'S FUNDING & M&A

The deals that shaped private markets (May 18-24, 2026)

Big raises this week:

Fintech:

• Mercury: $200M Series D at $5.2B (TCV) — Banking for AI-first startups, 1 in 3 U.S. startups

AI/Infrastructure:

• Anthropic: Hired Andrej Karpathy (OpenAI co-founder) — Major talent acquisition

• No new funding disclosed this week for Anthropic, but secondary market prices rising

Notable hires (talent as capital):

• Anthropic: Andrej Karpathy (OpenAI co-founder, Tesla AI lead)

• Anthropic: Ross Nordeen (xAI founding member)

• Anthropic: Chris Rohlf (cybersecurity veteran)

New markets/infrastructure:

• Polymarket: Launched private company prediction markets (Nasdaq partnership)

Total raised this week: ~$200M disclosed (Mercury only major round this week)

M&A highlights:

None disclosed this week.

Notable secondary market activity:

• SpaceX: IPO filing confirmed, June 12 listing at $1.75T

• Oura: Filed confidentially for IPO at $11B

• Polymarket: Now offering contracts on OpenAI, Anthropic, SpaceX, Stripe valuations

Portfolio moves:

• Leopold Aschenbrenner's Situational Awareness LP: $7.5B in bearish bets against Nvidia, AMD, chip stocks; Long on Bitcoin miners and AI power infrastructure

🔥 THE BOTTOM LINE

Andrej Karpathy joined Anthropic. Mercury raised $200M at $5.2B. Polymarket launched private company trading. And Leopold Aschenbrenner is shorting Nvidia while longing Bitcoin miners.

This week showed:

The talent war is real (Anthropic is winning).
The AI founder boom is real (Mercury is banking it).
Democratization of private markets is real (Polymarket is leading it).
The power infrastructure thesis is real (Aschenbrenner is betting billions on it).

The pattern:

AI is reshaping everything—talent, capital, markets, infrastructure. The winners aren't decided yet. But the battle lines are being drawn.

The takeaway:

If you're investing in private markets, pay attention to:

1. Talent flows (where are the best researchers going?)

2. Infrastructure plays (who's building picks-and-shovels?)

3. New access platforms (how can retail get exposure?)

4. Power infrastructure (is this the next bottleneck?)

Because right now? The AI war is the only game in town. And it's just getting started.

See you Tuesday.

Until then, keep hunting. ☕

Jose, Alberto, and the FC team.

P.S. Premium members: We're publishing the complete Anthropic vs OpenAI talent analysis Monday. Who else is likely to leave? Who's likely to join? And how does this affect IPO valuations? Plus: Our full breakdown of Aschenbrenner's power infrastructure thesis with specific stock picks.

DISCLAIMER

Founderscrowd is an independent financial media company. We are not a registered investment adviser, broker-dealer, or FINRA/SEC-regulated entity. Nothing in this newsletter constitutes investment advice, a recommendation, or an offer to buy or sell securities.

Private market and pre-IPO investments carry substantial risk. You may lose some or all of your invested capital. These investments are illiquid and may not be suitable for all investors. Only invest what you can afford to lose entirely.

Deal opportunities are sourced from third-party platforms. Founderscrowd does not act as a broker-dealer or conduct independent due diligence. We may receive compensation for featuring certain opportunities, which creates a conflict of interest.

All Founderscrowd Premium memberships are non-refundable. The value is delivered upon access to information, which cannot be returned once shared.

Past performance is not indicative of future results. Conduct your own research and consult qualified advisers before investing.

© 2026 Founderscrowd. All rights reserved.