Two Frontier Bets Print This Month: Cerebras Goes Public, SpaceX "Project Apex" Lines Up

For four years I have been telling LPs the same thing: the late-stage names you actually want — SpaceX, Cerebras, OpenAI, xAI, Neuralink, Perplexity — never come back to the secondary market at the price you think they will. You buy the round you can buy, you size with discipline, and you wait for the tape to tell the story.

This month, two of those positions hit the tape at the same time. This is the post I have been waiting to write.

Why this matters for the syndicate:

• We hold Cerebras alongside SPV co-investors who came in across multiple late-stage rounds, with a meaningful markup since our entry.
• IPO lockups on insider-held Class A typically run 180 days. CEO Andrew Feldman is not selling a single share into the IPO and will hold 10.3M shares — a signal we read as constructive.
• Cerebras is the only real wafer-scale alternative to Nvidia in inference compute. With Nvidia's data-center multiples plus the AI buildout we have walked through in prior memos, the comparable multiple compression risk looks asymmetric to the upside on day one.

We are in. We have been in. The lockup will set the tone for the next leg.

SpaceX — "Project Apex," 21 Banks, $1.5T–$2T

This is the one I have been quietly accumulating into for years. The numbers are now public.

• December 2025: SpaceX completed an insider tender at $421 per share / approximately $800B valuation — nearly 2x the $212 / $400B mark from July 2025.
• January 2026: The Financial Times confirmed Goldman Sachs, JPMorgan, Bank of America, and Morgan Stanley as lead bookrunners for a 2026 listing.
• April 2026: Reuters and CNBC reported the syndicate has grown to 21 banks under the internal code name "Project Apex," one of the largest underwriting books ever assembled. ("Project Apex" is just the bankers' working codename for the deal — but it tells you how serious this process has gotten.)
• Target valuation at IPO: $1.5 trillion to $2 trillion — which would dethrone Saudi Aramco's 2019 listing as the largest IPO in history.

Polymarket traders currently price Morgan Stanley as lead bookrunner at 49%, with prospectus release expected late May and a roadshow in June. ETrade has reportedly been tapped to lead retail allocation of up to 30% of the deal — unprecedented for a megacap IPO and a direct response to retail demand for Starlink/Starship exposure.

Our position was built primarily through 2023–2025 secondary tenders and one direct allocation, with a meaningful markup since entry. I am not selling on day one. I am writing the IPO playbook with our LPs now: how much to ride past lockup, how to ladder out into the first earnings cycle, what to do if Starlink gets spun out separately (which is on the table).

Why This Window, Not the Last One

Q1 2026 was the largest VC quarter ever — $300B into roughly 6,000 startups. The IPO window has been frozen for the better part of three years. It is now cracking open with two of the highest-conviction names on the late-stage roster going first.

Pile on:

• The DOL just opened the $8T 401(k) market to alternatives. Pension and DC-plan demand for late-stage venture is structural, not cyclical.
• Power-law math means you need 15–20+ frontier positions, not 2–3. Two SPVs in a friend's syndicate is a hobby, not a portfolio.
• Physical AI is finally crossing the factory floor — Figure at BMW, Tesla Optimus, $6B robotics quarter — and most LPs still have zero physical-AI positions.

This is what I mean when I say boring infrastructure wins: subs docs ready, KYC done, capital calls clean, reporting on time.

The Wildcard No One Else Has — The Newchip Warrant Portfolio

While public-market eyes are on Cerebras and SpaceX, Angel Deal Syndicate, LLC won the court-approved auction for what I have called the largest court-ordered liquidation of startup warrants in history.

81+ portfolio companies that have already raised $186M+ post-issuance. Existing markups in the 4x–6x+ range before any of them go public. We expect at least three of those names to file S-1s in the next twenty-four months. None of our LPs paid for the warrants — we acquired the entire book at the auction floor.

This is not the kind of opportunity that comes back. Bankruptcy auctions of accelerator portfolios this size happen approximately never.

How to Actually Get on the Bus

I have built a 660-LP network from a cold start over four years. The first OpenAI allocation went in at $157B. It is now marked at $852B. That is what disciplined late-stage entry plus the right infrastructure produces over time.

Here is the unusual part most syndicates will not tell you:

Three doors are open right now:

1. Substack — free deal flow. advalorem.substack.com — every soft-interest form, every IPO update, every warrant marker, in your inbox.
2. Discord — real-time signal. discord.gg/gErgtp9gTS — LPs, founders, and co-investors trade signal in real time.
3. Talk to me directly. X for unfiltered takes, LinkedIn for longer-form, or schedule a call.

If you want to be sized into the next SpaceX style IPO secondary or the next Cerebras-tier name, start the conversation now, not the day the prospectus drops.

Investment Criteria

Founder execution. Large scalable markets. Stage-flexible with disciplined entry valuations. High conviction in AI, Deep Tech, Robotics, Quantum, Blockchain.