Eridu

June 22, 2026

Public disclaimer: This is an independent public teardown based solely on publicly available information retrieved on 2026-06-22. It is research support, not investment, legal, tax, or financial advice. Company-provided statements are labeled as claims, not verified facts. Eridu did not participate in or review this report. Errors and omissions are possible; final decisions remain with the reader.

The bottleneck Eridu wants to own is already worth fighting over. The company stepped out of stealth with a >$200 million Series A and a serial networking founder, pitching clean-sheet AI data-center networking silicon for the hyperscaler era. That is the kind of setup that makes investors lean forward fast. The harder question is who actually captures the value when the dominant platform can bundle, buy, or wait. Eridu's public story has ambition, capital, and category pressure; what it does not yet show is the outside-the-room proof that turns a hard silicon thesis into an investable standard. Read the teardown before the first call turns prestige into conviction.

What The Company Does

Eridu describes itself as an AI networking company building a clean-sheet network switch across silicon, systems, and software for hyperscalers, neoclouds, sovereign clouds, and large enterprise artificial intelligence data centers (Eridu homepage; Eridu Series A announcement) company claim. The company says its architecture spans custom silicon, optics, packaging, systems, and switch software (Eridu vision article) company claim.

The commercial product is not publicly specified. Eridu has no public pricing page, product data sheet, customer case study, or technical documentation portal. Its public site and press coverage instead describe a development-stage hardware company using the Series A proceeds to complete development (Eridu Series A announcement; Network World) source-backed fact for the development-stage posture.

The business model is therefore unknown from public evidence. The likely model is a contact-led infrastructure sale into data-center buyers, but whether Eridu sells systems, silicon, software, licensing, or a bundle is not publicly disclosed. The homepage and contact page show an enterprise sales path rather than self-serve adoption (Eridu contact page) source-backed fact.

Key Takeaways

Investor reaction
Eridu is a serious first-call candidate because the category is enormous, the round is real, and Drew Perkins has independently corroborated prior networking outcomes. The public record also makes the case unusually dependent on claims the company has not yet validated outside its own surfaces.
Current verdict
hold with medium confidence. The right action is one time-boxed, kill-criteria-gated founder call, not a check and not a broad diligence process.
Why investors might lean in
Eridu announced a >$200 million Series A (TechCrunch; Network World; Crunchbase News) source-backed fact, and Perkins's prior Infinera outcome is independently reported (TechCrunch; Network World) source-backed fact.
Why investors might pull back
No public source shows a named customer, design win, tape-out, product benchmark, pricing model, or capital-to-revenue plan. The closest funded startup comp was absorbed by NVIDIA rather than scaling independently (CNBC; Crunchbase Enfabrica) source-backed fact.
Highest-leverage fix
Bring one independently verifiable adoption signal: a named hyperscaler, neocloud, sovereign-cloud, or large-enterprise evaluation, pilot, letter of intent, or design win that can be referenced outside Eridu's own materials.
Best next move
Take one structured founder call that resolves buyer signal, silicon milestone, team-coverage, and ownership math. If that call produces only investor quotes and company claims, the decision should move toward pass.

What Makes This Potentially Fundable

The fundable version of Eridu is not merely another switch vendor. It is an independent AI fabric layer that becomes designed into large clusters before the incumbent platform can absorb the economics. If Eridu can show a material radix, power, or cost advantage; a standards-interoperable slot buyers can adopt; and durable per-port value capture, the category can support a venture-scale outcome.

The attraction is real because the category ceiling is real. NVIDIA disclosed $31.376 billion in fiscal 2026 Data Center Networking revenue and 71.1% company gross margin (NVIDIA Form 10-K) source-backed fact, while Arista reported $9.006 billion in fiscal 2025 revenue and 64.1% gross margin (Arista Form 10-K) source-backed fact. Those comps prove the prize, not Eridu's claim on it.

The honest gap is equally large. Eridu's 10x radix, 40% capital expenditure savings, 70% networking-power savings, and one-switch-replaces-roughly-30 thesis remain company claim from Eridu-controlled pages (Eridu Series A announcement; Eridu vision article). Until buyer pull or first-silicon evidence exists, the upside is a category option, not an underwritten company outcome.

The Four Holes To Close Before Fundraising

Hole	Investor Fear	What To Bring
Independent adoption is absent.	After a >$200 million raise, investors will fear the traction is fundraising-led rather than buyer-led.	A customer-routable evaluation, pilot, letter of intent, design win, procurement reference, or customer-controlled technical contribution.
Technical proof is still company-controlled.	The home-run economics may be a marketing claim rather than a measured product edge.	Tape-out or first-silicon status, bring-up data, third-party benchmark results, or customer lab evidence.
Value capture may leak to incumbents.	NVIDIA, Broadcom, Arista, and Cisco can bundle, absorb, or out-cadence an independent challenger.	A buyer that wants an independent Eridu layer, standards-interoperability proof, and pricing-power evidence against incumbent bundles.
Ownership and capital path are opaque.	Even technical success may not return a first-check fund if multiple later nine-figure rounds wash out early ownership.	Round structure, cap table, pro-rata access, capital-to-general-availability plan, and a sub-$1 billion absorption return case.

Decision Snapshot

One-sentence company description: Eridu is a Series A AI data-center networking startup that claims to build clean-sheet switch silicon, systems, and software for large AI infrastructure buyers.
Screen: hold.
Confidence: medium.
Suggested next action: One tightly scoped founder call, pre-committed to kill criteria, focused on independent buyer signal, silicon milestone, team depth, and ownership math.
Why this matters now: AI fabric spending is moving into an incumbent-capture window, and pre-GA hardware companies can lose the category before they publicly ship.
Investor-readiness diagnosis: Eridu has enough category and founder signal to earn a call, but not enough independent proof to earn diligence momentum.
Best founder use of this report: Treat it as a proof-packet checklist before investor enthusiasm turns into a customer, benchmark, cap-table, and value-capture interrogation.

IC Disagreement Map

Evidence dossier review was independent across five public partner lenses: the power-law lens tested fund-returning upside, the prepared-mind lens tested category timing, the founder-jockey lens tested team fit, the risk-reduction lens tested critical flaws, and the long-horizon lens tested value capture and ownership durability.

Power-law partner
pursuemedium confidence
The category ceiling, Perkins's prior networking outcome, and the Enfabrica transaction justify an option-like first step, even though conviction is not yet underwritten.
Would flip on
A named buyer evaluation, tape-out confirmation, or third-party benchmark would move them toward higher-conviction pursue; a second empty proof pass or slippage would move them toward hold.
Prepared-mind partner
holdmedium confidence
The timing window is real, but Eridu has not shown a chosen battleground, benchmarked product insight, or evidence that independent AI interconnect persists as a durable layer.
Would flip on
A benchmarked slot, standards-interoperability posture, and one independent demand signal would move them toward pursue; proprietary-only architecture or continued buyer silence would move them toward pass.
Founder-jockey partner
pursuelow confidence
Perkins's founder-market fit is genuine, but the public evidence does not verify the technical leadership depth on advanced-node networking silicon.
Would flip on
Verified CTO and systems-architect tape-out history plus a named customer evaluation would move them toward higher-conviction pursue; an empty or adverse technical-leadership check would move them toward hold or worse.
Risk-reduction partner
holdmedium confidence
No fatal flaw is confirmed, but every differentiation claim remains company-controlled and the value-capture risk against NVIDIA and Broadcom is live.
Would flip on
Independent customer proof plus tape-out confirmation would move them toward pursue; a further public-evidence check with no buyer proof, a buyer standardizing on an incumbent bundle, or team credential break would move them toward pass.
Long-horizon partner
holdmedium confidence
Category and founder quality are real, but the company may be structurally un-ownable for a first-check fund across multiple hardware mega-rounds.
Would flip on
A capital-to-general-availability plan, early ownership protection, and one independent design win would move them toward pursue; several more nine-figure rounds with no pro-rata path would move them toward pass.

No partner changed their vote after the evidence-request round. One partner reduced confidence from medium to low because the CTO and systems-architect verification gate remained unresolved; the other four confirmed their votes unchanged. The split was not unanimous, but all five lenses converged on the same proof gates from different reasoning paths, which makes the convergence evidential rather than social.

Scenario Range

Scenario	What The Company Looks Like In 3-5 Years	Falsifiable Trigger To Watch	Earliest Evidence
Strikeout	Eridu burns through development capital, misses tape-out or bring-up milestones, fails to win a referenceable design slot, and is wound down or sold below the entry mark.	A second proof pass shows no buyer evaluation and no silicon milestone while Broadcom, NVIDIA, and Arista continue shipping fabric upgrades.	Founder cannot provide routable buyer references or dated first-silicon evidence on the first call.
Base	Eridu builds useful interconnect technology but becomes a second-source component or strategic acquisition rather than an independent standard.	A buyer adopts Eridu only as an interoperability component, or a strategic acqui-license appears before durable pricing power is proven.	First customer discussions emphasize interop and optionality rather than committed design-win lock-in.
Home run	Eridu becomes a default AI fabric layer for multiple hyperscalers or neoclouds, with design-win lock-in, durable per-port economics, and a software or control-plane moat.	A named buyer reference plus independent benchmark plus confirmed standards-interoperability evidence appears while ownership remains protectable.	Customer-controlled evaluation evidence, a dated tape-out with bring-up data, and a capital plan that preserves pro-rata.

What Investors Will Test

What We Looked For	Current Read	Investor Implication	Founder Prep Priority
Independent proof outside the company's own surfaces.	The round, founder history, market size, and Enfabrica precedent have external support; customer pull, benchmarks, tape-out, and TSMC allocation depth do not.	The first call should be a proof test, not a narrative pitch.	Bring customer-routable, lab, foundry, and cap-table evidence.
A fund-returning path rather than a big-market claim.	Public comps prove the category can be enormous, but Eridu's served slice and value-capture path are unknown.	Investors will not underwrite a category denominator as a company outcome.	Define the beachhead slot, average selling price, design-win sequence, and moat.
Founder-market fit on the hardest build.	Perkins is strongly evidenced; the silicon execution layer remains thin in public sources.	Founder quality is a real lean-in factor but not enough to clear technical diligence.	Document prior tape-outs, shipped chips, packaging ownership, and systems leadership.
Symmetric double-down and quit triggers.	The deal is milestone-gatable, but public evidence does not justify patience without named progress.	A hold should not become an open-ended watch list.	Agree in advance what proof moves the decision to pursue or pass.

Claim Reconciliation: Inconsistencies Investors Will Catch

Claim Or Metric	Where It Appears	Conflicting / Unreconciled Versions	Why Investors Flag It	How To Reconcile
Funding amount and round history.	Eridu Series A announcement TechCrunch Crunchbase Eridu	Public materials support a >$200 million Series A, while TechCrunch reports approximately $230 million total funding and Crunchbase lists one Series A without the same total.	Runway, dilution, ownership, and entry-price math depend on the exact financing history.	Provide a cap table, round-by-round financing schedule, securities filing or exemption explanation, and the composition of the $230 million figure.
TSMC relationship depth.	Eridu Series A announcement Network World TechCrunch	VentureTech Alliance participation is externally reported, but committed foundry allocation and advanced-packaging capacity are not public.	Investors will not treat an investor tie or quote as wafer-start access.	Separate the investor relationship from capacity, process-node, packaging, and non-recurring engineering commitments.
Team coverage for the hardest build.	Eridu leadership page Network World	Leadership titles are visible, but the CTO's detailed public profile was unavailable and advanced-node tape-out history is not public.	The decisive execution risk is not whether the team is senior; it is whether the team has shipped the specific silicon layer.	Bring resumes, patent lists, shipped-product references, foundry milestone history, and a named packaging owner.
Perkins's Eridu role and Mojo Vision board role.	Eridu leadership page TechCrunch Drew Perkins X profile	Public sources consistently show Perkins as Eridu chief executive, while his X profile also lists Mojo Vision board chairman.	This is not a title contradiction, but a capital-intensive tape-out window makes time allocation a fair diligence question.	Explain whether the Mojo role is passive, provide board-duty scope, and show Eridu operating cadence.

Diligence Findings: Issues To Fix Before You Raise

1 deal-breaker5 high1 medium

No independently referenceable buyer signal is public after a >$200 million Series A.
deal-breaker if unresolved after the first call.
Commercial diligence.
Fix
Provide a routable design win, evaluation, pilot, letter of intent, or procurement reference.
The core performance and economic claims remain company-controlled.
high
Technical diligence.
Fix
Provide third-party benchmark data or customer lab results against named incumbent products.
First-silicon and tape-out status are not public.
high
Technical and financial diligence.
Fix
Bring dated tape-out, bring-up, general-availability, and burn-to-milestone evidence.
The independent value-capture path is not proven against NVIDIA, Broadcom, Arista, and Cisco.
high
Market and strategy diligence.
Fix
Show a buyer that wants Eridu as an independent layer, standards-interoperability evidence, and pricing power.
Funding history, ownership, and pro-rata access are opaque.
high
Financing diligence.
Fix
Provide cap table, round map, valuation, pro-rata terms, and capital-to-revenue model.
CTO and systems-architect advanced-node networking ASIC history is not publicly evidenced.
high
Team diligence.
Fix
Provide resumes, patents, shipped silicon references, foundry and respin history, and reference calls.
Export-control classification and freedom-to-operate are not publicly determinable.
medium
Legal and regulatory diligence.
Fix
Provide export-classification memo, licensable-destination map, patent list, and freedom-to-operate opinion.

Data Room Readiness Checklist

Procurement-readiness score: partial.
Rationale: Eridu targets enterprise-scale, hyperscaler, sovereign-cloud, and large data-center buyers, but public evidence does not show security documentation, export-control classification, supply-chain commitments, product specifications, or customer procurement references.

Data Room Area	What Investors Expect	Current Read	Status	Priority
Company overview & corporate	One-pager, incorporation, good standing, structure	Eridu Corporation is named in legal pages, and public contact details are available (Eridu terms; Eridu contact page).	`partial`	High
Financials	Monthly profit and loss, three-year model, burn, runway	No public financials, burn, runway, or revenue.	`missing`	High
Cap table & funding history	Clean cap table, prior rounds, SAFEs or notes, valuation	>$200 million Series A is externally supported; total funding and ownership details are not.	`partial`	High
Legal & IP	Bylaws, consents, intellectual-property assignments, freedom-to-operate	Legal pages exist, but patent posture and freedom-to-operate are not public.	`partial`	Medium
Product & technology	Roadmap, architecture, benchmarks, security and compliance	Product claims and job-stack signals exist; no public datasheet, tape-out, benchmark, or security package.	`partial`	Critical
Team	Org chart, key agreements, vesting, advisor arrangements	Public leadership page lists 12 executives; private employment, vesting, and role depth are not visible.	`partial`	High
Customers & traction	Retention cohorts, annual recurring revenue bridge, pipeline, references	No public customer or revenue evidence.	`missing`	Critical
Market & competition	Market model, competitive landscape, pricing	Market and incumbent evidence is strong; Eridu-specific pricing and served-market evidence are missing.	`partial`	High

First-Call Agenda For The Startup

Time	Topic	Founder Goal	Evidence To Bring
0-10 minutes	Frame the exact first slot.	Make clear whether Eridu first owns scale-up, scale-out, a systems bundle, or another concrete fabric layer.	Architecture map, beachhead buyer type, and why the slot is defensible.
10-25 minutes	Buyer signal.	Prove there is real adoption pull beyond investor and analyst interest.	Named or routable evaluation, pilot, design-win, letter-of-intent, or procurement contact.
25-40 minutes	Silicon milestone and benchmark.	Convert the technical case from marketing claim to milestone evidence.	Tape-out or first-silicon status, bring-up data, benchmark plan, and comparison set.
40-50 minutes	Team coverage.	Show the hardest build has experienced owners beyond the chief executive.	CTO, chief systems architect, packaging, optics, switch software, and validation resumes and references.
50-60 minutes	Ownership and kill criteria.	Decide whether a first-check investor can own enough upside and when the company should be passed on.	Cap table, pro-rata terms, capital-to-general-availability plan, and agreed quit triggers.

Business Model And Revenue Signals

Public pricing.
No public pricing page found.high
SourceEridu homepage
The company appears contact-led, not self-serve.
Revenue, backlog, or paid pilots.
No public revenue or customer evidence found.high
SourceEridu homepage Network World
Pre-general-availability posture is consistent with absence, but it remains a gate.
Funding to complete development.
source-backed factCompany announcement and reputable press.high
SourceEridu Series A announcement Network World
Indicates product maturity rather than commercial revenue.
Gross-margin benchmark band.
source-backed fact for comps; inference for Eridu.Public-company filings.medium
SourceNVIDIA Form 10-K Arista Form 10-K
Public comps do not prove Eridu's margin.

No public source discloses Eridu revenue, gross margin, backlog, paid pilots, average selling price, bill of materials, or customer concentration. Any revenue model before founder disclosure would be an estimate, not a fact.

Revenue Quality Checklist

Before Eridu can move from hold to pursue, the company should be ready to show contracted annual recurring revenue or hardware backlog if any exists, paid-pilot terms, buyer identity, margin assumptions, average selling price, bill of materials, design-win conversion, renewal or expansion terms, and customer concentration. Investors will rebuild customer acquisition cost, lifetime value, payback, gross revenue retention, net revenue retention, churn, deployment gross margin, and milestone-to-revenue conversion from raw contracts and financials, and those numbers must tie across the deck, model, cap table, and customer documents.

Funding And Ownership Context

Item	Public Read	Evidence Label	Diligence Request
Total raised / rounds	>$200 million Series A is corroborated; approximately $230 million total is single-source press and not fully reconciled.	`source-backed fact` for >$200 million Series A; `company claim` for approximately $230 million total.	Provide round map, cap table, financing documents, and reconciliation to public sources.
Instruments (equity / SAFE / notes)	Not publicly disclosed.	`unknown`	Provide instruments, conversion terms, valuation, and any side letters.
Investors	Publicly named investors include Socratic Partners, John Doerr, Hudson River Trading, Capricorn, Matter, and VentureTech Alliance, with Crunchbase listing 13 investors (Crunchbase Eridu; TechCrunch).	`source-backed fact` for public record contents, with roster caveats.	Provide official investor schedule and lead rights.
Post-money valuation	Not publicly disclosed.	`unknown`	Provide valuation, available allocation, and ownership target.
Cap table / ownership	Not publicly disclosed.	`unknown`	Provide fully diluted cap table, option pool, founder vesting, and pro-rata rights.
Burn & runway	Not publicly disclosed.	`unknown`	Provide monthly burn, cash balance, runway, tape-out spend, and contingency plan.
Next-round plan	Not publicly disclosed.	`unknown`	Provide capital-to-general-availability plan and expected round sizes.

SEC EDGAR was checked and showed no matching Form D result in the public search. This absence does not disprove the raise, but it means public financing records do not yet support ownership or valuation modeling.

Founder And Team

Drew Perkins
Founder and chief executive; independently reported prior networking founder outcomes.
Eridu leadership page TechCrunch Network World LinkedIn Drew Perkins
Evidencesource-backed fact
Confidencehigh
Omar Hassen
Chief product officer; chip background is reported in press but not independently verified in detail.
Eridu leadership page TechCrunch LinkedIn Omar Hassen
Evidencecompany claim for detailed chip pedigree; source-backed fact for role.
Confidencemedium
Promode Nedungadi
Chief technology officer; identity is public, detailed public profile was unavailable.
Eridu leadership page Network World
Evidencesource-backed fact for role; unknown for track record.
Confidencehigh for role, low for track record.
Mike Capuano
Chief business development officer and chief marketing officer; data-center networking go-to-market signal.
Eridu leadership page LinkedIn Mike Capuano
Evidencesource-backed fact
Confidencehigh
Broader leadership bench
Eridu lists 12 executives and board/advisor names.
Eridu leadership page
Evidencesource-backed fact for roster contents.
Confidencehigh
Hiring scale
Company careers board lists many ASIC, software, validation, and Bangalore roles.
Eridu jobs board LinkedIn Eridu company
Evidencesource-backed fact
Confidencemedium

Founder Competency Coverage

Domain depth
Drew Perkins
evidenced
Omar Hassen
claimed
Promode Nedungadi
claimed
Mike Capuano
evidenced
What The Evidence Is
Perkins's prior networking outcomes are press-corroborated; Omar's and Promode's depth is mainly press and title evidence; Capuano has public data-center networking go-to-market signals.
Technical build capability
Drew Perkins
evidenced
Omar Hassen
unknown
Promode Nedungadi
unknown
Mike Capuano
absent
What The Evidence Is
Perkins has systems-company founder history, but public sources do not show the CTO and chief systems architect's advanced-node tape-out histories.
Product
Drew Perkins
evidenced
Omar Hassen
evidenced
Promode Nedungadi
claimed
Mike Capuano
evidenced
What The Evidence Is
Product roles and quotes exist; independent product performance proof does not.
GTM / sales
Drew Perkins
claimed
Omar Hassen
claimed
Promode Nedungadi
absent
Mike Capuano
evidenced
What The Evidence Is
Capuano covers go-to-market leadership, but no buyer win validates the motion.
Leadership / hiring
Drew Perkins
evidenced
Omar Hassen
evidenced
Promode Nedungadi
evidenced
Mike Capuano
evidenced
What The Evidence Is
Leadership roster, headcount, and open roles support senior team assembly.
Fundraising history
Drew Perkins
evidenced
Omar Hassen
unknown
Promode Nedungadi
unknown
Mike Capuano
unknown
What The Evidence Is
Perkins and the round are visible; individual fundraising roles are not.
Prior founding outcomes
Drew Perkins
evidenced
Omar Hassen
unknown
Promode Nedungadi
unknown
Mike Capuano
unknown
What The Evidence Is
Perkins's prior outcomes are press-corroborated; no equivalent public founding outcomes were found for the others.

Advanced-node networking switch ASIC.
high
Team coverage today
Claimed coverage through CTO and engineering leaders, but prior tape-out evidence is not public.
What would close it
Resumes, patent lists, shipped-product records, foundry references, and prior respin history.
Advanced packaging and chiplet integration.
high
Team coverage today
No named, independently verified owner in public evidence.
What would close it
Named leader with CoWoS-class or equivalent shipped-program evidence.
Hyperscaler-scale systems and switch software.
medium
Team coverage today
Leadership and hiring suggest coverage, including systems and software roles.
What would close it
Reference architecture, software roadmap, and customer validation.
Hyperscaler go-to-market and design-win sales.
high
Team coverage today
Capuano's role is visible, but buyer pull is absent.
What would close it
Named evaluation, pilot, letter of intent, or procurement reference.

Public Professional Footprint

Public technical artifacts, professional social presence, education traces, and credential checks carry low decision weight by design; they corroborate or weaken the competency read above and sharpen founder-call questions, but none alone changes the verdict.

Technical artifacts (repo level)
Drew Perkins
absent
Omar Hassen
absent
What The Public Record Shows
No public repositories were used as competence evidence.
Professional social content
Drew Perkins
evidenced
Omar Hassen
evidenced
What The Public Record Shows
Perkins's public profiles show Eridu and Mojo Vision titles; Omar's profile shows industry activity, but detailed experience fields were thin.
Education / credentials
Drew Perkins
claimed
Omar Hassen
absent
What The Public Record Shows
Perkins's LinkedIn profile includes education history, but degree completion was not independently verified.
Publications / patents / certifications
Drew Perkins
unknown
Omar Hassen
unknown
What The Public Record Shows
Patent and publication checks were not sufficient to evidence the decisive tape-out history.

Net read: the founder footprint supports a first call because Perkins is genuinely category-relevant, but the verdict stays gated because the public footprint does not prove the silicon leadership layer.

Founder-Market Fit Read

What is promising: Perkins has independently reported networking-company outcomes and appears to have assembled a senior hardware and go-to-market bench around a large category.
What is missing: Public proof of advanced-node networking ASIC execution by the CTO, chief systems architect, packaging owner, and validation team.
What to prepare: A role-by-role technical evidence packet, customer-engagement proof, and a clear explanation of Perkins's time allocation given the Mojo Vision board role.

Eridu Corporation
Legal entity named in terms and privacy pages.
Eridu terms Eridu privacy policy
Evidencesource-backed fact for company-controlled legal pages.
Confidencehigh
Follow-upConfirm incorporation, subsidiaries, and capitalization privately.
Infinera
Prior company tied to Perkins; sale to Nokia reported at roughly $2.3 billion.
TechCrunch Network World
Evidencesource-backed fact
Confidencehigh
Follow-upReference what part of that execution transfers to this silicon layer.
Lightera / Gainspeed / Mojo Vision
Prior founder-associated companies reported in press.
TechCrunch Network World
Evidencesource-backed fact for press-reported association.
Confidencemedium
Follow-upClarify operating role and transferable execution lessons.
Mojo Vision
Perkins lists board chairman role while serving as Eridu chief executive.
Drew Perkins X profile TechCrunch
Evidencesource-backed fact for public titles; needs founder validation for time allocation.
Confidencemedium
Follow-upConfirm the board role is passive and conflict-managed.

Traction

Customer Status Table

No named customer was identified in public evidence.
Eridu names target buyer categories but no customer-controlled logo, pilot, design win, or case study.
No.Claim onlyUnknownUnknownUnknownhigh that no public customer proof was found.

Eridu's homepage, press coverage, Crunchbase profile, LinkedIn company page, and general category searches were checked and showed no public named customer, design-win, or case-study listing. This absence is decision-relevant after a nine-figure raise, but it is not proof that no private evaluations exist.

Traction Signals

>$200 million Series A announced out of stealth.
source-backed facthigh
SourceEridu Series A announcement TechCrunch Network World Crunchbase News
Round structure, valuation, and pro-rata access.
Approximately 100 employees, with 101 shown on LinkedIn and roughly 100 reported by TechCrunch.
source-backed factmedium
SourceLinkedIn Eridu company TechCrunch
Actual full-time headcount, contractor mix, and critical-path ownership.
Hiring across ASIC, software, validation, Saratoga, and Bangalore.
source-backed facthigh
SourceEridu jobs board
Which roles are open because the team is scaling versus missing critical coverage.
Open Compute Project and AI infrastructure event activity appears on company social footprint.
company claim / source-backed fact for post existence.medium
SourceLinkedIn Eridu company
Whether activity corresponds to standards contribution, conformance, or customer adoption.
No public revenue, paid pilot, backlog, or marketplace listing found.
high
SourceEridu homepage Network World
Customer contracts, backlog schedule, or paid evaluation terms.

Product Hunt, app-store listings, and software marketplace checks were not material to this hardware infrastructure company; the relevant traction checks were company pages, reputable press, funding databases, LinkedIn, careers pages, and category searches. The absence of consumer-style listings carries little weight.

Hiring And Org Momentum

Open roles.
source-backed factThe company careers board showed many open roles across ASIC, software, validation, and multiple locations.
SourceEridu jobs board
Hiring mix looks like a build-stage silicon and systems company, not a commercial scaling company.
Headcount.
source-backed factLinkedIn showed 101 employees, and TechCrunch reported approximately 100 employees.
SourceLinkedIn Eridu company TechCrunch
Consistent with substantial team assembly before public product proof.
Executive roster.
source-backed fact for roster contents.Eridu lists 12 executives on its leadership page.
SourceEridu leadership page
Seniority supports seriousness but does not prove the exact technical competency matrix.

LinkedIn Jobs, Indeed, Glassdoor, and the company careers page were checked; the company careers page was the strongest positive listing source, while third-party job boards appeared only as lower-confidence listing mirrors. The exact cross-board role counts carry less weight than the role mix.

Role-mix read: Engineering, ASIC, validation, and switch-software hiring dominate, which supports a pre-production build-stage read.
Momentum read (growth / steady / contraction / unknown): growth, with medium confidence because public role counts do not show hiring conversion or churn.

Traction Quality Read

Traction Dimension	Current Status	Good Enough For First Call?	Needed For Deep Diligence
Funding signal	Strong and externally supported for the Series A.	Yes.	Round structure, valuation, and ownership rights.
Customer proof	No public named customer or design win.	Only if the first call is explicitly gated on resolving it.	Referenceable buyer evidence.
Technical proof	Company claims only; no public benchmark or tape-out.	Only as a question, not as conviction.	Tape-out evidence and independent benchmark.
Team signal	Strong CEO evidence and senior roster; technical-leadership depth not public.	Yes.	Verified technical execution matrix.
Revenue quality	Not public.	No.	Contracts, backlog, pilot terms, margins, and collection history.

Competitive Landscape

Segment	Examples	Customer Alternative	Pressure On Company
Platform incumbent	NVIDIA Spectrum-X and broader networking stack.	Buy the bundled GPU-plus-networking platform from NVIDIA.	NVIDIA can capture value through bundle, roadmap, or acquisition rather than let an independent layer compound.
Merchant silicon incumbent	Broadcom Tomahawk and Thor families; Marvell.	Use known switch or network-interface silicon through existing systems vendors.	Eridu must beat shipping roadmaps and supply trust.
Systems incumbents	Arista and Cisco.	Buy integrated, supported AI Ethernet systems.	Eridu must displace established data-center relationships and support models.
Startup comp	Enfabrica.	Use or wait for strategically absorbed interconnect technology.	The comp suggests strategic value, but also absorption risk.
Status quo / substitutes	Multi-tier Ethernet, InfiniBand, NVLink, hyperscaler in-house designs.	Scale current architectures or build internally.	Eridu must prove enough value to justify qualification risk.

Category Visibility Snapshot

Google Search / United States"AI data center networking switch"

Captured page-one results (2026-06-22): Broadcom, Marvell, HPE, Nokia, Juniper, and other incumbent or informational domains appeared in the retrieved page-one results.
Was the company present?: No.

Google Search / United States"AI data center networking switch startups"

Captured page-one results (2026-06-22): General startup list and infrastructure domains appeared; Eridu was not a page-one result.
Was the company present?: No.

Google Search / United States"NVIDIA Spectrum-X AI networking"

Captured page-one results (2026-06-22): NVIDIA-owned pages dominated the retrieved results, with ecosystem pages also present.
Was the company present?: No.

Google Search / United States"Arista AI networking data center"

Captured page-one results (2026-06-22): Arista-owned pages dominated the retrieved results.
Was the company present?: No.

Visibility read (inference, low confidence): Eridu has company-specific press visibility, but generic category demand is still owned by incumbents and ecosystem pages. That matters less than product proof today, but it reinforces the distribution and credibility gap.

Pricing And Competitive Benchmark

Alternative	What It Offers	Public Price Signal	Price Vs. This Company	Evidence Label
NVIDIA Spectrum-X	End-to-end AI Ethernet platform with switch and SuperNIC positioning.	No public comparable list price in the cited product page.	Eridu has no public price, so relative pricing is `unknown`.	`company claim` for NVIDIA product claims via NVIDIA Spectrum-X.
Broadcom Tomahawk / Thor	Shipping switch and network-interface silicon for very large AI clusters.	No public comparable list price in cited press.	Eridu has no public price, so relative pricing is `unknown`.	`source-backed fact` for product coverage via Network World Tomahawk 6 and Network World Thor Ultra.
Arista AI Etherlink	AI Ethernet systems and lossless fabric portfolio.	No public comparable list price in cited page.	Eridu has no public price, so relative pricing is `unknown`.	`company claim` for Arista product claims via Arista AI Networking.
Cisco AI switching	Silicon One and data-center switch positioning for AI workloads.	No public comparable list price in cited press.	Eridu has no public price, so relative pricing is `unknown`.	`source-backed fact` for coverage via Network World Cisco.

Price positioning read: Eridu's public price position is unknown, and this is a diligence gap because the claim of lower capital expenditure requires a buyer-level total cost of ownership model.
Price claims to correct or substantiate: Any capital expenditure, power, or one-switch-replaces-many claim should be tied to a named cluster configuration, incumbent comparison, and buyer-validated cost model.

Competitive Wedge

Eridu's possible wedge is a clean-sheet hardware and systems architecture that materially collapses network tiers, power draw, and cost for AI clusters. The wedge is potentially defensible only if it produces a benchmarked advantage, lands inside a standards-interoperable buyer slot, and compounds into design-win lock-in. Today, it is still a company claim, while the incumbent response and absorption risk are supported by public evidence.

Risks And Open Questions

Risk	Severity	Evidence	What To Ask
Value capture leaks to NVIDIA, Broadcom, or systems incumbents.	high	CNBC, NVIDIA Form 10-K, and NVIDIA Spectrum-X.	Which buyer wants Eridu independently, and what prevents platform absorption or bundling?
No independent customer proof.	critical	No named customers found on public company pages or reputable coverage.	Can any buyer be referenced independently, even under a controlled diligence process?
No public tape-out or benchmark proof.	high	Eridu Series A announcement and Network World show development-stage posture.	Has first silicon taped out, and what third-party proof exists?
Ownability risk across future hardware rounds.	high	Public financing data does not show valuation, cap table, or pro-rata rights.	Can early ownership survive the capital-to-revenue arc?
Export-control and freedom-to-operate exposure.	medium	CRS and BIS establish the regulatory regime, not Eridu-specific classification.	What export classification, destination map, patent list, and freedom-to-operate work exists?

Pre-Mortem: The Most Likely Obituary

Cause of death (one sentence): Eridu built a better component but failed to become an independent layer, so the value it created was captured by an incumbent platform through bundling or a sub-fund-returning strategic transaction.
The causal chain (3-5 steps from today to the shutdown): Eridu spends the Series A completing development; first silicon or benchmarks slip; buyers wait for NVIDIA, Broadcom, Arista, or Cisco; the company needs another large round before revenue; early ownership is diluted; and a strategic buyer acquires or licenses the technology below the fund-returning threshold.
The earliest observable warning sign: The first founder call produces no routable buyer signal, no dated tape-out evidence, and no path to protect ownership through the next hardware rounds.
The question that defuses this chain today: Which independent buyer is willing to design in Eridu, and what evidence proves the product and ownership path can survive long enough for that decision to matter?

Decision-Critical Unknowns

Unknown	Why It Is Decision-Critical	Best Evidence	Decision Effect
Named buyer evaluation or design win.	It is the strongest evidence that Eridu can beat the absorption prior.	Customer-controlled announcement, procurement reference, or routable evaluation contact.	Confirmation would move the deal toward pursue; continued absence after another proof pass would move it toward pass.
First-silicon or tape-out status.	It determines feasibility, burn, and milestone credibility.	Foundry milestone, bring-up data, validation plan, and dated general-availability roadmap.	On-time evidence would support diligence; slippage would trigger quit logic.
Independent benchmark.	It tests the economic and performance claims.	Third-party or customer benchmark against named incumbents.	Validation would strengthen the home-run case; internal-only evidence would keep the case at hold.
Technical-leadership execution depth.	The hardest build cannot rest only on the chief executive's prior outcomes.	CTO, systems architect, packaging, optics, and validation evidence.	Verified depth would support founder-market fit; empty or adverse evidence would push toward pass.
Capital-to-revenue and ownership plan.	Hardware companies can win technically while becoming un-ownable.	Cap table, valuation, pro-rata, reserve plan, and capital-to-general-availability model.	Fund-returning ownership would support continued work; wash-out math would move the deal toward pass.

Diligence Questions

First Call

Question	Why It Matters	Good Evidence
Which buyer can be independently referenced for an evaluation, pilot, letter of intent, or design win?	This resolves the highest-impact traction and value-capture gap.	Routable buyer contact, customer-controlled material, or procurement-stage proof.
Has first silicon taped out, and what is the dated bring-up and general-availability plan?	This resolves the feasibility and runway gate.	Tape-out confirmation, foundry milestone, bring-up data, and burn plan.
What third-party or customer benchmark supports the headline performance and economics?	This turns the home-run mechanism from claim into evidence.	Independent benchmark against named Broadcom, NVIDIA, Arista, or Cisco alternatives.
What role does Eridu own first, and how does it interoperate with relevant standards?	This tests whether the company is a durable layer or a feature.	Slot map, standards participation, conformance plan, and buyer architecture evidence.

Follow-Up

Question	Why It Matters	Good Evidence
What is the complete round structure, cap table, valuation, and pro-rata path?	This determines whether a technical win can return this fund.	Cap table, financing documents, pro-rata rights, and reserve model.
Who has shipped the relevant networking ASICs and packaging programs before?	This verifies founder-market fit on the hardest build.	Resumes, patents, foundry histories, shipped-product references, and reference calls.
What export-control classification and freedom-to-operate work has been completed?	Sovereign-cloud and advanced-silicon markets carry regulatory and IP risk.	Counsel memo, classification, destination map, patent list, and freedom-to-operate opinion.
What supply allocation exists with TSMC or packaging partners?	Allocation can determine whether product arrives inside the market window.	Capacity commitments, non-recurring engineering terms, wafer-start evidence, or equivalent supply agreements.

Kill Criteria

Kill Criterion	Evidence That Would Trigger It
A second deliberate proof pass returns no independent technical or customer proof.	No buyer reference, no tape-out, no benchmark, and only company-controlled claims after a focused diligence request.
Value-capture impossibility is confirmed.	A target buyer standardizes on NVIDIA, Broadcom, or another incumbent for the same slot, or another startup in the category is absorbed in a way that confirms the pattern.
Silicon-execution leadership cannot be verified.	CTO, chief systems architect, and packaging-owner checks return empty, adverse, or irrelevant histories.
Tape-out or bring-up slips while incumbents keep advancing.	Repeated schedule slippage with no first silicon, while Broadcom, NVIDIA, Arista, or Cisco expands relevant fabric offerings.
Ownership math cannot return the fund.	Several further nine-figure rounds are required with no early pro-rata or available ownership protection.
Integrity or credential issue appears on a founding-team member.	Confirmed material fiduciary, credential, or governance break tied to the team.

Double-Down Criteria

Double-Down Criterion	Evidence That Would Justify More Diligence
Independent buyer proof appears.	A named or routable hyperscaler, neocloud, sovereign-cloud, or enterprise evaluation, pilot, letter of intent, or design win.
First-silicon evidence appears on schedule.	Tape-out confirmation, bring-up data, and a dated general-availability path.
Product advantage is independently measured.	Third-party or customer benchmark validating a material fraction of the claimed performance or cost advantage.
Team coverage is verified on the hardest build.	CTO, systems architect, packaging, optics, and validation leaders show prior relevant shipped silicon.
Ownership remains fund-returning.	Round terms, valuation, and pro-rata preserve a fund-mattering position even in a strategic absorption scenario.

Founder Action Plan

Timeframe	Action	Output
Before the next investor call	Assemble a referenceable customer-engagement list and pick one initial battleground.	Buyer proof packet and slot map.
Before the next investor call	Compile a silicon-execution evidence matrix for technical leaders.	CTO, systems architect, packaging, optics, validation, and software-owner dossiers.
Before the next investor call	Prepare tape-out, bring-up, and general-availability evidence.	Dated milestone plan tied to burn and foundry commitments.
During deep diligence	Surface or commission independent benchmark work.	Third-party or customer proof against named incumbents.
During deep diligence	Provide round structure and capital-to-revenue plan.	Cap table, pro-rata terms, reserve model, and return sensitivity.
During deep diligence	Document standards, export-control, supply-chain, and freedom-to-operate posture.	Interoperability evidence, capacity terms, export classification, and legal risk map.

Decision

Screen: hold.
Confidence: medium.
Rationale: Eridu has a credible category, real founder signal, and a serious financing event, but the public evidence does not yet show independent buyer pull, product proof, technical-leadership depth, or ownable return math.
What would move this to pursue: One independently referenceable buyer signal, confirmed first silicon or third-party benchmark proof, verified technical-leadership coverage, and entry terms that preserve a fund-returning ownership path.
What would move this to pass: A second empty proof pass, confirmed platform absorption risk, unverified technical leadership, repeated tape-out slippage, unownable financing math, or a material integrity issue.
Recommended next step: Take one structured founder call with pre-committed kill criteria and no open-ended watch-list posture.
Founder preparation standard: Bring evidence that a skeptical infrastructure investor can route, verify, and model, not additional prestige or market-size framing.

How We Would Miss This One

The miss scenario: Eridu could be the rare clean-sheet AI networking company where the public silence is normal pre-general-availability noise, the founder's prior operating discipline transfers, and the first customer-controlled design win appears just after cautious investors step away.
Flip conditions: A named buyer evaluation, a dated first-silicon milestone, benchmarked performance, verified technical-leadership history, and protectable ownership would flip the case toward pursue.
Revisit trigger / date: Revisit by 2026-09-22, or immediately if Eridu produces a referenceable buyer signal, tape-out proof, independent benchmark, or capital-to-general-availability plan.

Source Log

Eridu homepage
eridu.ai
Company positioning and absence of public customer logos.
Retrieved 2026-06-22company-controlled pagehigh
Eridu Series A announcement
eridu.ai
Funding claim, target buyers, performance claims, investors, TSMC quote, development-stage posture.
Retrieved 2026-06-22company-controlled announcementmedium
Eridu vision article
eridu.ai
Technical thesis and economic-performance claims.
Retrieved 2026-06-22company-controlled articlemedium
Eridu jobs board
ats.rippling.com
Open roles, hiring mix, locations, and stack signals.
Retrieved 2026-06-22company-controlled careers pagehigh
Eridu leadership page
eridu.ai
Executive roster and leadership titles.
Retrieved 2026-06-22company-controlled pagehigh
Eridu contact page
eridu.ai
Headquarters address, contact path, and social links.
Retrieved 2026-06-22company-controlled pagehigh
Eridu terms of use
eridu.ai
Legal entity and governing-law context.
Retrieved 2026-06-22company-controlled legal pagehigh
Eridu privacy policy
eridu.ai
Legal page, data-practice language, and product-commerce signal.
Retrieved 2026-06-22company-controlled legal pagehigh
Crunchbase Eridu
crunchbase.com
Company category, one Series A record, investor count, employee band, and profile contents.
Retrieved 2026-06-22startup database (Crunchbase)medium
LinkedIn Eridu company
linkedin.com
Employee count and company social footprint.
Retrieved 2026-06-22public profile datasetmedium
LinkedIn Drew Perkins
linkedin.com
Founder identity, public profile, and avatar source.
Retrieved 2026-06-22public profile datasetmedium
LinkedIn Omar Hassen
linkedin.com
CPO identity and profile activity.
Retrieved 2026-06-22public profile datasetmedium
LinkedIn Mike Capuano
linkedin.com
Go-to-market leader identity and public profile activity.
Retrieved 2026-06-22public profile datasetmedium
LinkedIn Ting-Kuang Chiang
linkedin.com
Chief systems architect identity signal.
Retrieved 2026-06-22public profile datasetlow
LinkedIn Brian Kinard
linkedin.com
CFO and chief legal officer identity signal.
Retrieved 2026-06-22public profile datasetmedium
Drew Perkins X profile
x.com
Eridu chief executive and Mojo Vision board chairman self-description.
Retrieved 2026-06-22public profile datasetmedium
TechCrunch Eridu Series A
techcrunch.com
>$200M Series A, approximately $230M total claim, employee count, Perkins history, and VentureTech Alliance participation.
Retrieved 2026-06-22tech press (TechCrunch)high
Network World Eridu coverage
networkworld.com
Funding, incumbents, pre-general-availability posture, TSMC relationship claim, and founder quotes.
Retrieved 2026-06-22technology trade press (Network World)high
Crunchbase News biggest rounds
news.crunchbase.com
Public ranking and corroboration of the >$200M round.
Retrieved 2026-06-22startup funding press (Crunchbase News)high
IDC Ethernet switch market summary
idc.com
FY2025 Ethernet switch and data-center Ethernet switch market figures.
Retrieved 2026-06-22analyst report summary (IDC)high
Network World Gartner switch forecast
networkworld.com
AI fabric timing, high-end operator count, price and lead-time signals.
Retrieved 2026-06-22analyst-via-press report (Network World / Gartner)medium
650 Group AI Ethernet forecast
650group.com
AI scale-out, scale-up, optics, and incumbent-barrier context.
Retrieved 2026-06-22analyst blog (650 Group)medium
650 Group data-center networking demand
650group.com
Broad data-center networking market trajectory.
Retrieved 2026-06-22analyst blog (650 Group)medium
NVIDIA FY2026 Form 10-K
sec.gov
NVIDIA Data Center Networking revenue and gross-margin benchmark.
Retrieved 2026-06-22public-company filing (SEC)high
Arista FY2025 Form 10-K
sec.gov
Arista revenue, gross margin, and customer concentration benchmark.
Retrieved 2026-06-22public-company filing (SEC)high
CNBC Enfabrica transaction coverage
cnbc.com
NVIDIA-Enfabrica acqui-license precedent and value-capture risk.
Retrieved 2026-06-22business press (CNBC)high
Crunchbase Enfabrica
crunchbase.com
Enfabrica closed status and company profile context.
Retrieved 2026-06-22startup database (Crunchbase)medium
Enfabrica Series B announcement
blog.enfabrica.net
Enfabrica positioning and prior funding context.
Retrieved 2026-06-22competitor-controlled announcementmedium
NVIDIA Spectrum-X
nvidia.com
NVIDIA end-to-end AI Ethernet platform positioning.
Retrieved 2026-06-22competitor-controlled product pagemedium
Arista AI Networking
arista.com
Arista AI Ethernet systems positioning.
Retrieved 2026-06-22competitor-controlled product pagemedium
Network World Broadcom Tomahawk 6
networkworld.com
Broadcom 102.4 Tbps switch silicon cadence.
Retrieved 2026-06-22technology trade press (Network World)high
Network World Broadcom Thor Ultra
networkworld.com
Broadcom AI network-interface-card cadence.
Retrieved 2026-06-22technology trade press (Network World)high
Network World Cisco AI switches
networkworld.com
Cisco AI data-center switch positioning.
Retrieved 2026-06-22technology trade press (Network World)high
Congressional Research Service R48642
congress.gov
Advanced semiconductor export-control regime context.
Retrieved 2026-06-22regulatory source (CRS)high
BIS advanced-semiconductor export controls
bis.gov
Export-control and national-security regime context.
Retrieved 2026-06-22regulatory source (BIS)high

Access limitations: Public research could not retrieve Eridu's private customer pipeline, contracts, cap table, valuation, securities filings, export-control classification, first-silicon data, benchmark packets, detailed CTO profile, foundry-capacity terms, or freedom-to-operate analysis. Search-result checks and unavailable profiles were used only to mark gaps, not as client-safe proof. Those limitations mean this report can support a gated first call, but it cannot validate revenue, ownership, technical readiness, or customer adoption without founder-provided documents and reference checks.

Public disclaimer: This is an independent public teardown based solely on publicly available information retrieved on 2026-06-22. It is research support, not investment, legal, tax, or financial advice. Company-provided statements are labeled as claims, not verified facts. Eridu did not participate in or review this report. Errors and omissions are possible; final decisions remain with the reader.

What The Company Does#

Key Takeaways#

What Makes This Potentially Fundable#

The Four Holes To Close Before Fundraising#

Decision Snapshot#

IC Disagreement Map#

Scenario Range#

What Investors Will Test#

Claim Reconciliation: Inconsistencies Investors Will Catch#

Diligence Findings: Issues To Fix Before You Raise#

Data Room Readiness Checklist#

First-Call Agenda For The Startup#

Business Model And Revenue Signals#

Revenue Quality Checklist#

Funding And Ownership Context#

Founder, Team, And Related Entities#

Founder And Team#

Founder Competency Coverage#

Public Professional Footprint#

Founder-Market Fit Read#

Related Entities And Founder-Associated Companies#

Traction#

Customer Status Table#

Traction Signals#

Hiring And Org Momentum#

Traction Quality Read#

Competitive Landscape#

Category Visibility Snapshot#

Pricing And Competitive Benchmark#

Competitive Wedge#

Risks And Open Questions#

Pre-Mortem: The Most Likely Obituary#

Decision-Critical Unknowns#

Diligence Questions#

First Call#

Follow-Up#

Kill Criteria#

Double-Down Criteria#

Founder Action Plan#

Decision#

How We Would Miss This One#

Source Log#