Private-Market Diligence · Accredited Use Only

Spawn-Point Private-Market Map

Autonomous science · physical AI · energy primitives · bio & quantum — August 2026

What this is, and what it cannot be

There is no EquityZen investor API and no live feed into your portal. Nothing here reads your account. This page reconstructs two things an outsider can assemble: the last priced round (funding truth from Harmonic) and a secondary implied mark (Caplight / Forge / Hiive / Nasdaq Private Market and press). Those give a directional premium-or-discount to the last round.

The one number that decides a trade — the live ask on a specific listing — appears only when you log in. So does the fee schedule for your account, the exact share class, and the preference stack.

How to use the page. Read the reconstructed premium/discount in the table below. Then open the listing in your own portal and check the single number only you can see — the live ask — against it. If the live ask sits below both the last round and the data-provider consensus, you may have a real discount. If it sits above, a fund is marking it up to you. The mechanics section spells out the full check.

The blunt read

None of these twenty is a clean secondary discount today. The hot names re-rated 2–5x in the last year and trade at or above their last round. The cheap names are too new to have a secondary market — they are primary or SPV only. The few showing an apparent discount are either data-lag artifacts (stale quotes against a fresh up-round) or discounts you would earn by sitting behind a large preference stack.

One name is worth watching for a genuine motivated-seller gap: Emerald Cloud Lab — an aging private with no priced round since 2019, real lots on Forge/EquityZen. Verify a live bid/ask below the implied mark and clear the pref stack first.

The diligence table

Discount — worth diligence At mark — watch Premium — pass on valuation Too new / now public

Premium/Discount is secondary implied mark vs. last-round post-money. “UNKNOWN” means no public secondary print exists and the direction needs your portal. Marks are directional and dated ~Aug 2026.

Company	Thesis	Last round (date · raise · post)	Total raised	Secondary implied mark	Prem / Disc vs last round	Liq-pref / structure	Runway	Availability	Verdict
Emerald Cloud Lab	Cloud lab — remote design/run of life-science experiments across 200+ instruments; the original self-driving-lab-as-a-service.	Series B · Apr 2019 $42.7M · post n/d	~$150M (Harmonic undercounts $69.7M)	No public mark. Aging private, no up-round since 2019 — pattern favors a discount to any stale internal mark.	DISCOUNT (inferred)	Unknown; 2010 vintage, stacked prefs possible. Diligence the stack.	No raise ~7 yr at ~200 heads — revenue self-sufficient or insider bridges. Not fresh cash.	EquityZen (as emeraldtherapeutics) + Forge. Likely secondary lots.	Discount — diligence
CuspAI	Generative-AI + molecular-sim materials search engine; lead use case porous materials for carbon capture. Cambridge UK.	Series B · Jul 2026 $450M · $2.6B post	~$650M (Harmonic $980M inflated)	Listed UpMarket, “not currently traded.” Round 3 weeks old, ~5x in 9 mo — trades at/above $2.6B if at all.	PREMIUM	Assume 1x non-part (KP/NEA-led). N/d.	$450M just in at ~63 heads — multi-year, no seller pressure.	Primary-only in practice; scarce premium sliver on pre-IPO desks.	Premium — pass
Radical AI	Physical self-driving materials labs (robotics + AI, hypothesis→synthesis→characterization). Brooklyn Navy Yard.	Seed · Jul 2025 $55M · ~$275M est	$55M ($60M A rumored, unconfirmed)	On Hiive / Notice, no public mark. Hot NVIDIA/RTX seed typically asks above the ~$275M est.	PREMIUM (likely)	Assume 1x non-part (seed). N/d.	$55M at ~40 heads but capital-heavy (labs, buildout); already raising again.	Hiive (radical-ai-stock) + Notice; thin early lots.	At mark — watch
Polaron	Imperial spinout; AI reads microstructure images to predict material performance. Won £1M Manchester Prize. Cheapest name in the basket.	Seed · Feb 2026 $8M · ~$40M est	$8M	No secondary market exists — too early/small. Next raise is a priced Series A (primary).	N/A — primary only	Assume 1x non-part (EU seed). N/d.	$8M at 18 heads, software burn — ~18–24 mo.	Not listed. Direct-primary only — the one Daxos could lead.	Too new — primary/SPV
Lila Sciences	Flagship-hatched “scientific superintelligence” + autonomous Science Factory labs across therapeutics, materials, energy.	Series A · ~Oct 2025 ~$350M · ~$1.3B post	$550M–$785M (web vs Harmonic split)	Listed EquityZen/Forge/NPM, no live clearing price found. Hot-sector name usually a premium on these venues.	UNKNOWN → premium-lean	Assume 1x non-part (Flagship/GC-led A).	~$550M+ vs heavy wet-lab burn, 485 heads — 2–3 yr, will need more.	EquityZen + Forge + NPM. Most tradeable name in the batch.	At mark — watch
Periodic Labs	AI scientists + autonomous labs for physical sciences (superconductors, semis). Ex-DeepMind + ex-OpenAI founders.	Seed · ~Sep 2025 $300M · ~$1.3B post	$500M (Harmonic fields stale)	Pending round talked at ~$7B (Bloomberg/Forbes) = ~5x the seed in under a year. No marketplace listing.	PREMIUM (~5x re-rate)	Assume 1x non-part (seed).	$300M seed on ~73 heads — long; ~$500M A would extend years.	None — primary/SPV only, tightly held.	Premium — pass
Discovery Loop	Brand-new (Aug 5 2026) autonomous-science PBC; automate the full research loop. Jeff Dean + Sanjay Ghemawat + Vinyals + Le.	Round not closed valuation n/d	—	Not yet secondary-tradeable. No mark disclosed. Google a founding investor + year-one compute.	N/A — too new	Unknown. PBC structure — note for downstream terms.	Unknown; Google compute subsidy lowers early burn.	None — primary only, will be heavily oversubscribed.	Too new — primary/SPV
Chai Discovery	AI molecular design / drug discovery (Chai-2, antibody + structure prediction); used by Lilly, Novartis, Pfizer.	Series C · Jul 2026 $400M · $3.8B post	$630M	Hiive ~$39.88/sh, 3 live orders (07/31/26). Mark tripled off Dec-2025 B — a freshly-stepped-up premium.	PREMIUM	Assume 1x non-part (multi-tier stack).	$630M on ~34 heads — least capital-constrained per head; long.	Hiive (quoted); not confirmed EZ/Forge.	Premium — pass
Physical Intelligence	Robot foundation models (VLA / π-series). Pure software “brain” layer any robot maker licenses. Cleanest software bet.	~Mar 2026 round ~$1B · ~$11.5B post (prior $5.6B, Nov 2025)	$1.07B (pre-2026 round)	~2x step-up in ~4 mo — a premium primary re-rate. Secondary would clear near $11.5B.	PREMIUM (2x, 4 mo)	Assume 1x non-part.	Long — $600M + ~$1B, 283 heads, software burn.	Tightly held, top-tier cap table; little retail secondary.	Premium — pass
Apptronik	Apollo general-purpose humanoid — hardware OEM (bodies). Mercedes, GXO, Google DeepMind partners. Capital-heavy.	Series A ext · Feb 2026 $520M · $5B post	$1.01B	Hiive indications (Nov 2025) ~$5B — roughly AT the last primary mark. No clear gap.	≈ AT MARK	Assume 1x non-part.	Moderate — $520M fresh but 443 heads, humanoid hardware burns fast; watch 12–18 mo.	Best secondary liquidity of the physical-AI four — live Hiive indications near $5B.	At mark — watch
Figure AI	Full-stack humanoid (Figure 02/03) + in-house Helix VLA. Body and brain. Most vertically integrated, most richly priced.	Series C · Sep 2025 >$1B · $39B post	$1.854B	Chatter implies compression below $39B (a discount to last primary) but no priced round confirms. C&D letters to brokers.	PREMIUM ($39B pre-rev)	Unknown — a $39B pre-rev mark may carry ratchet/senior pref. Do not assume clean 1x.	Long on cash ($1B+); 687 heads, heavy hardware + manufacturing burn.	Effectively blocked — C&D posture; anything offered likely unauthorized/forward. Avoid.	Premium — pass
Skild AI	“Skild Brain” — body-agnostic robot foundation model. CMU spinout. Same lane as Physical Intelligence.	Series C · Jan 2026 $1.4B · >$14B post	$2.213B	$14B, ~3x in 7 mo. No clean secondary print — vertical re-rate means any secondary clears at a premium.	PREMIUM (3x, 7 mo)	Unknown — SoftBank mega-rounds often negotiate structure (senior pref). Real diligence item.	Very long — $1.4B on just 102 heads; lowest burn, biggest war chest.	Tightly held, small team, elite cap table; little/no retail secondary.	Premium — pass
Quaise Energy	Superhot geothermal via millimeter-wave (gyrotron) drilling to ~20 km / ~500°C; runs superhot steam through fossil-plant turbines.	Series B · Jul 2026 $134M first close · post n/d	$225M	Harmonic estimates ~$737M (modeled, not a mark). Caplight profile, no reliable live price. No honest prem/disc.	UNKNOWN	Assume 1x non-part.	Capital-heavy, pre-rev. First close funds first pilot only; more equity+debt to follow. Multi-year.	Not on EquityZen; tracked by Caplight. Effectively no liquid secondary.	At mark — watch
Commonwealth Fusion (CFS)	Compact tokamak with HTS magnets (SPARC→ARC). Best-funded private fusion company in the world.	Later-stage · Jul 2026 $1.0B · post n/d	$3.99B	Hiive ~$245.98/sh (08/08/26) but company not approving transfers — indicated, not executable. Post n/d.	UNKNOWN (gated)	Unknown — SWF/institutional later-stage can carry structure.	1,074 heads, SPARC+ARC hardware — very high capex; fresh $1B, but power years out.	EquityZen (fund) + Hiive/NPM — but transfers blocked / ROFR-gated. Access is the constraint.	At mark — watch
Helion Energy	Pulsed FRC fusion with direct electricity recovery; D-He3 fuel. Microsoft PPA. Altman chairman. Stale-quote trap.	Series G · Jun 2026 $465M · $15.5B post (~3x prior ~$3.0B)	$1.51B	Estimates ~$80–85/sh reflect the OLD ~$3B mark; new post is $15.5B (~5x). Apparent huge discount is a data-lag artifact, not a live offer.	UNKNOWN (stale quotes)	Unknown — a 3x up-round to $15.5B is where ratchets/seniority hide. Diligence hard.	567 heads, building Polaris/Orion; $465M fresh; pre-rev bar the Microsoft PPA.	EquityZen + Forge + Hiive. Forge min ~$10K; EZ fund-based.	At mark — watch
Fervo Energy	Enhanced geothermal — horizontal drilling + fiber sensing (O&G playbook), ORC to power. Now public.	IPO · May 2026 $27/sh · ~$7.65B implied NASDAQ: FRVO	$3.76B	Trades publicly now — price discovery is the open market (FRVO), not a secondary desk.	N/A — public	N/A post-IPO — common stock.	275 heads; capital-heavy but public-equity + project-finance access.	No longer a pre-IPO secondary name. Old EZ/Hiive pages are historical.	Now public — a stock
Retro Biosciences	Longevity — cellular reprogramming, autophagy, in-vivo gene/cell therapy for ~10 healthy years. First Alzheimer’s trial running. Altman-anchored.	May 2026 size n/d · ~$1.8B post	$1.36B	EquityZen + Hiive ~$12.68/sh. No PPS for the round. Signal: chased ~$5B in Dec ’25, closed ~$1.8B — flat-to-soft.	UNKNOWN (reads soft)	Assume 1x non-part.	Long — ~$1.36B vs 88 heads, one early clinical asset; modest burn.	EquityZen + Hiive; small lots realistic.	At mark — watch
Altos Labs	Cellular rejuvenation (partial reprogramming). Long-horizon research, no disclosed clinical pipeline. Largest headcount of the four.	Later event · Mar 2024 anchor: ~$3B seed (2022) no official post-money	$5.62B	NPM ~$14.77, Hiive ~$16.00, Notice ~$20.96; ~$2.7B implied vs $5.6B raised = steep implied discount.	DISCOUNT (but earned)	Assume 1x non-part; large pref stack ahead of common — material. Diligence the waterfall.	Long on paper ($5.6B) but 557 heads, no revenue — very heavy burn.	Broad — NPM, Hiive, Forge, UpMarket, Notice. Liquid by private-market standards.	At mark — watch
PsiQuantum	Fault-tolerant quantum on silicon-photonic qubits in a standard fab; targeting million-qubit machines. Fault-tolerant-compute primitive.	May 2026 ~$1.5B · ~$10.5B post ($6B→$7B→$10.5B <1 yr)	$2.75B (Harmonic lags; true ~$4.3B+)	Marked up round over round. Any secondary clears at/above the $10.5B primary.	PREMIUM	Assume 1x non-part; late-stage crossover-heavy.	Long — $1.5B May 2026 + gov grants vs hardware buildout.	Thin/premium — institutional demand soaks supply; retail lots scarce, priced up.	Premium — pass
Xaira Therapeutics	AI-first drug discovery (generative protein/antibody design, ex-Baker/RFdiffusion roots). CEO Marc Tessier-Lavigne.	Launch · Apr 2024 >$1B committed · post n/d ARCH’s largest-ever check	$1.0B (PitchBook ~$1.3B)	PremierAlts ~$3.5B implied (Aug ’25) — already a premium to capital in. Platform estimate, no closed print. Soft.	PREMIUM (soft)	Assume 1x non-part.	Long — ~$1B+ vs 210 heads, no disclosed late-stage burn.	Hiive/UpMarket/Forge watch; tradable supply thin, tightly held.	At mark — watch

The shortlist, grouped honestly

Lead takeaway: none of these is a clean secondary discount today. The groups below sort by why, not by hype.

Watch for a stale / motivated-seller discount

Emerald Cloud Lab — the one credible secondary-discount candidate. Aging private, no priced round since 2019, real lots on Forge and EquityZen. A 2010-founded company with no up-round in ~7 years usually trades below any stale internal mark. This is inference until your portal shows a live bid/ask below the ~$150M-raised implied mark and the pref stack is clean.
Helion Energy — the stale-quote gap. Trackers quote ~$80–85/share off the OLD ~$3B valuation; the June 2026 Series G reset the post-money to $15.5B (~5x). A naive read shows a huge discount — it is a data-lag artifact, not an executable offer. Only real if you find a live post-Series-G offer materially below the new mark. Verify it is not stale pre-round pricing before treating it as a discount.
Apptronik — trades near its own $5B last round on live Hiive indications. No dislocation to exploit today; interesting only if a motivated seller prints a real discount. The most accessible secondary of the physical-AI names.
Altos Labs — the implied discount (~$2.7B vs $5.6B raised) is real but earned: no clinical assets, top-decile burn, and a large preferred stack ahead of common. Diligence the pref waterfall before calling the discount value.

Premium — pass on valuation

CuspAI — $2.6B, ~5x in 9 months, tier-1 cap table locked. Even a discount would be a rich absolute price.
Physical Intelligence — ~2x re-rate to ~$11.5B in four months; anchor investor reported in distress, so the mark itself is shaky. Watch for a secondary print that resets below $11.5B.
Skild AI — $14B on 102 people, pre-revenue, 3x in 7 months. Steepest re-rate-per-head in the cohort. No dislocation to buy.
Figure AI — $39B pre-revenue froth; even a secondary “discount” is still rich, structure may carry a ratchet, and access is hostile (cease-and-desist to brokers). Avoid.
PsiQuantum — marking up round over round ($6B→$7B→$10.5B in under a year); pre-revenue on a decade-plus horizon. Nothing trades at a discount for a Daxos-size buyer.
Chai Discovery — $3.8B, tripled off the Dec-2025 round; Hiive quote is a freshly-stepped-up premium.
Periodic Labs — pending round talked at ~$7B, a ~5x re-rate off the $1.3B seed. Also not secondary-tradeable (see next group).

Too new — primary / SPV only, not yet secondary-tradeable

Discovery Loop — announced Aug 5 2026, round not closed, no mark. Google-anchored founder round; no outside checks or SPVs at this stage.
Periodic Labs — seed-stage, no marketplace listing; only route would be an SPV into the pending round if invited (unlikely).
Polaron — $8M Feb-2026 seed, 18 heads, no lots anywhere. The one name Daxos could actually lead — but as a primary at the Series A, direct to the company, not via EquityZen.
Quaise Energy — early Series B, no visible marketplace listing, no disclosed post-money. Revisit once the round fully closes and a real quote appears.
Radical AI — on Hiive/Notice but thin early lots at a likely premium; watch, don’t chase.
Commonwealth Fusion — listed but the company is blocking transfers / ROFR-gated. Access, not price, is the binding constraint.

Now public — evaluate as a stock

Fervo Energy (NASDAQ: FRVO) — IPO’d May 2026 at $27, ~$7.65B implied, ~3.5x the Dec-2025 private round. Nothing to source on the secondary side; any thesis here is a public-equity call, buy it on the open market at any size.
Note: Cerebras converted to a public path earlier on the same pattern — a private name whose diligence question becomes a listed-stock question, not a secondary one. Treat any newly-public spawn-point the same way.

Mechanics — how the venues actually price

One idea to hold: a “price” on these platforms is not a market-clearing price. It is one of three things — what a single motivated seller will take, a broker’s algorithmic mark, or a fund sponsor’s ask with the fee already baked in. Your job on any listing is to figure out which, then translate it into two comparable numbers: price per share and implied post-money valuation.

EquityZen — set-price funds (dealer offer)

Two products. SPV / Standard Deals: you buy membership in a single-company fund, one line on the cap table, ~$5K minimum, 2.5% placement (2% over $1M), usually no management fee or carry. Direct Share Acquisitions: your name goes on the cap table, $200K+, tiered fees 2.5%→1.5%.

The listed PPS is the sponsor’s ask with the placement fee layered on and often a spread over what the underlying seller took. Read it as a dealer offer, not a market. Multi-company funds are different — 0.75–2.25% annual + 15% carry. Confirm the live schedule in your portal (Morgan Stanley has been cutting EZ fees).

Hiive / Forge — live bid/ask (real discounts show here)

Hiive is the closest thing to an exchange: a live, anonymous order book with real bids and asks, ~$250M+ monthly volume. This is where you actually see a motivated seller undercut the last round — the best single tool for spotting a genuine discount.

Forge (now Schwab-owned, closed early 2026) is more institutional/RFQ and publishes Forge Price (a mark) and the Forge Index. Nasdaq Private Market runs company-sponsored tenders — you transact when a window opens, and the tender price is a clean read on what the company thinks shares are worth.

Caplight / Forge Data — analytics, no execution

Data-only marks you triangulate against, never trade on. Caplight MarketPrice blends rounds, trades and order flow into a daily PPS; buy-side seat shows bid/ask depth. ApeVue (MSCI-owned since Apr 2026) runs give-to-get consensus marks. Notice.co gives a quick algorithmic implied valuation.

Where the true discount surfaces: a live ask on Hiive below the data-provider consensus. That gap — live ask under consensus mark — is the signal. An EZ SPV priced above consensus is the opposite: a markup wearing a discount’s clothes.

The liquidation-preference trap

This is where a “20% discount to the last round” can be worth nothing. The last-round valuation priced senior preferred with downside protection. EZ SPVs usually hold common sourced from employees — the bottom of the stack. Buy common at a 20% discount to that number and you bought a different, worse security at a small discount to a better one.

Preferences are paid off the top at exit; common divides what is left. The trap bites in the mediocre-outcome zone, which is most outcomes. If preferences ahead of common total, say, $2B against a $3B mark, common is thin and the discount is fictional. If the seller can’t show class and stack, price it as common behind an unknown preference — assume the worst tranche.

The per-listing checklist — run this in your own portal

Only your logged-in portal shows the live ask, the fee schedule for your account, and doc access. Pull every field into one line before deciding.

Ask PPS vs last-round PPS. The number on the listing against the most-recent priced-round PPS.
Implied valuation vs last post-money. Implied = ask PPS × fully-diluted shares. Delta = implied / last post-money − 1. That is your premium/discount %.
Share class. Common, or which preferred series? EZ SPVs usually hold common.
Preference stack. Total $ senior to you; 1x or 2x; participating or not; seniority order. Is the discount still real net of the stack?
Minimum. ~$5K SPV vs $200K+ DSA. Is there a qualified-purchaser ($5M+) gate?
All-in fees. Placement % at entry, plus any annual fee + carry (multi-company funds), plus the spread over what the underlying seller took. Compute your breakeven markup — fees push true entry above the headline PPS.
Structure. Direct shares vs SPV vs forward contract. Who holds voting / info / ROFR — you or the manager? Avoid a forward where shares never transfer unless you understand it.
Seller-motivated or fund-markup. Is the ask below the last round AND below Hiive/Caplight/ApeVue consensus (genuine motivated seller), or above consensus (a fund marking it up to you)? This single comparison is the whole game.

Decision rule. A buy is a real discount to a live consensus, in a senior-enough class, in a direct or clean-SPV wrapper, on a name whose round cadence says the next mark is up. Anything priced above consensus in a common-behind-a-big-stack SPV is a markup wearing a discount’s clothes.

Disclaimer. Reconstructed marks are directional and dated (~Aug 2026). Funding facts are from Harmonic; secondary figures are marketplace asks, algorithmic marks, or press estimates — not confirmed closes. Several are flagged where Harmonic lags or double-counts. Verify the live ask, the exact share class, and the full preference stack in your own portal before sizing anything. Nothing here is investment, legal, or tax advice. Markets and method only; no personal data.

Sources: Harmonic API (funding truth). Secondary/valuation: Caplight, Forge, Hiive, Nasdaq Private Market, UpMarket, Notice, PremierAlts, and press (Bloomberg, Reuters, TechCrunch, StatNews, Forbes, PitchBook). For accredited-investor reference use only.