pi Ventures.
The AI-first fund that got there early (2016) and kept going deeper: Locus sold to IKEA's Ingka, Agnikul seeded, and the current cheques are photonic interconnects, fusion and quantum photonics, but no Fund III three years after Fund II.
GW GP Score
pi was writing AI-first cheques when the term needed explaining (Locus, Wysa, Niramai, SigTuple: 2016-17), and its thesis has migrated honestly: from application AI to the hard infrastructure underneath it (LightSpeed Photonics for AI compute interconnects, Pranos Fusion for power, Quanfluence for quantum photonics). The 2025 Locus sale to Ingka/IKEA validates the earliest vintage. The institutional questions have piled up though: no Fund III announced three years post-Fund II, an MD quietly absent from the team page, and no disclosed economics on either 2025 exit.
Fundraise first: do not underwrite the next vintage until the capital base is visible.
Portfolio quality depends on a narrow theme; the right comp is concentration risk, not headline unicorn count.
No named analyst override yet; current view is generated from the scored dossier and should be refreshed when the next primary source lands.
Fund III: any announcement (none 3 years post-Fund II).
Roopan Aulakh status: unconfirmed departure.
No Fund III = deployment runway question; the 2025-26 cheques may be Fund II's last.
Computed from current dossier sources; analyst override pending.
Right thesis twice in a row, wrapped in a fund whose institutional signals (no Fund III, opaque exits, quiet departures) lag its investing judgment.
Agnikul co-hold, SigTuple, Pixis: plus a fresh infra-AI cohort
Two 2025 events, zero disclosed economics
AI-infrastructure thesis is early again, like 2016
Fund III missing; site data stale
Founder constant; senior bench churn unacknowledged
Portfolio quality 25% · Exit realisation 30% · Sector positioning 15% · Capital velocity 15% · Franchise stability 15%: see the full GP Score methodology · compare this firm →
Pressure catalysts, fund status or evidence gaps outweigh current positives; score support depends on the next verification cycle.
Agnikul Cosmos liquidity clears with credible OFS/block-sale evidence.
Another senior departure or weak attribution on the next fund's named partners.
Fund III: any announcement (none 3 years post-Fund II).
The investing eye is proven; the operation isn't. If a Fund III materialises, demand the Locus/Wysa economics and the team-change story first: both are knowable and both are being left unsaid.
No Fund III = deployment runway question; the 2025-26 cheques may be Fund II's last.
Fund III: any announcement (none 3 years post-Fund II).
Coverage initiated: inaugural GW GP Score.
Roopan Aulakh status: unconfirmed departure.: Directionally confirmed but no press exists: pi Ventures' own team page no longer lists her, and her LinkedIn/Crunchbase/RocketReach profiles now show her as Investment Director at Prosperity7 Ventures. No dated news article covers the move, so treat as GW profile-evidence, not a sourced departure story. E.
Catalyst exposure · latest source pressure
Recent public evidence mapped to this GP. Read this before reading the position table: it is where the score can move next.
Keep the file warm; current events cut both ways or are not yet material enough to move score.
Attribution0 positive catalyst(s), 1 pressure catalyst(s); top driver: India PE-VC value down 5% YoY in H1 CY2026.
The broad private-capital backdrop is selective rather than euphoric; fundraising and mark support now need firm-specific proof.
Score pressureRaises the bar for weak-DPI managers and stale flagship vehicles; rewards funds with fresh closes or visible exits.
Separate broad PE-VC softness from venture AI strength before changing sector-positioning scores.
The funds
~₹927 Cr (~$115M) across two funds across 2 tracked vehicles.
| Vehicle | Vintage | Size | Stage | Note |
|---|---|---|---|---|
| Fund II | 2022Correction onset | ₹702 Cr | Seed/Series A | LPs: BII, SIDBI (₹100 Cr), Premji Invest, Accel, NIDIA, Colruyt |
| Fund I | 2018Expansion | ₹225 Cr | Seed | BII/IFC-backed; Locus, Wysa, Niramai, SigTuple vintage |
Closest booksEndiya Partners (1 shared) · Speciale Invest (1 shared)computed · E
Fund analytics · LP underwriting screen
Computed from the public dossier: vehicle cadence, realised exits, mark drift, concentration and franchise stability. All computed signals are GW estimates.
No fresh flagship close found after the 2021-22 cycle-top window.
Next liquidityAgnikul Cosmos · Pixis
2 visible exit events since Jul 2024.
No fresh flagship close found after the 2021-22 cycle-top window.
3 up / 3 flat / 0 down tracked signals.
Public senior departures disclosed in the dossier.
3 up / 3 flat / 0 down
2 events since Jul 2024
Largest tracked active exposure
Higher spread = less balanced franchise
Repricing began mid-deployment
Rational growth vintages
$500M mark: co-held with Speciale
$209M raised; ~$58M revenue; no round since 2023
At least one cited source is dated 2026.
2 visible events tracked.
8 representative positions tracked.
1 departure signals structured.
Fund-level DPI/TVPI is usually not public; proxy remains estimated.
Peer median 2 · +0
Peer median 2 · +0
Peer median 74 · -14
Peer median 62 · -11
Peer median 0 · +1
Diligence agenda · how this view can change
The live research question, the proof required, and the source standard. This prevents a GP score from becoming a stale label.
Watch is visible in the structured people file; past DPI may not be portable to the next vehicle.
Current · At least one cited source is dated 2026.
Roopan Aulakh status: unconfirmed departure.
Confirm partner roster, board attribution and key-man clauses from firm pages, filings, LP memos or named reporting.
Roopan Aulakh status: unconfirmed departure.
Agnikul Cosmos: next public-market or secondary print
Evidence quality · source file health
How hard the public record is beneath this view: source mix, freshness, low-confidence fields and the next source action.
Low-confidence fields: Performance proxy.
5 total sources · 20% primary
At least one cited source is dated 2026.
YourStory: Pranos Fusion (Mar 2026)
Upgrade Performance proxy evidence before changing the score.
Source tiering follows Gravitywell policy: primary = filing/regulator/company disclosure, secondary = reputable media or research, E = Gravitywell estimate. Evidence score is a GW control metric, not an external rating.
The position book
44 companies over ten years; ~25 active. 8 tracked below.
| Company | Sector | Entered | Entry stage | Status | 12-mo signal | Latest read |
|---|---|---|---|---|---|---|
| Agnikul Cosmos | Deeptech & Space | 2020 | Early | Soonicorn | ▲ | $500M mark: co-held with Speciale |
| SigTuple | Healthcare | 2017 | Seed | Private | → | Edge-AI analyzer with NVIDIA (Feb 2026); co-held with Endiya |
| Pixis | AI | 2020 | Series A | Soonicorn | → | $209M raised; ~$58M revenue; no round since 2023 |
| Niramai | Healthcare | 2017 | Seed | Private | → | AI breast-cancer screening |
| LightSpeed Photonics | Deeptech & Space | 2025 | Pre-Series A led ($6.5M) | Private | ▲ | Optical interconnects for AI compute |
| Pranos Fusion | EV & Climate | 2026 | Seed co-led ($6.8M) | Private | · | Commercial fusion, with Ankur Capital |
| Moonrider | EV & Climate | 2025 | Series A led ($6M) | Private | ▲ | Electric tractors |
| Ottonomy | Deeptech & Space | 2022 | Early | Private | · | Autonomous delivery robots |
Representative tracked positions reconstructed from public disclosures: not the full book. 12-mo signal is a GW estimate of mark direction from round/exit prints (E). Ownership stakes are not shown: not reliably public.
Sector exposure · where the book leans
Share of the tracked active book by normalized sector (GW estimate): read against our sector dossiers for crowding.
New cheques · 2025-26
Thirteen cheques since January 2025, skewing decisively to AI's physical substrate: photonics, fusion, electric farm equipment, quantum. The thesis line 'core AI → disruptive AI → all disruptive deeptech' is visible deal by deal.
Photonic quantum computing
Realisation · are LPs getting paid?
Two 2025 liquidity events from the 2016-17 vintage, both economically opaque. A global strategic (IKEA) buying an Indian AI-logistics company is genuine thesis validation; whether it returned meaningful cash to Fund I is unknowable from outside.
Acquired by Ingka Group (IKEA): terms undisclosed; Locus last marked ~$300M (2021).
Merged with US April Health: no cash economics disclosed.
Strategic validation; terms undisclosed
Institutional re-up from Fund I
3 years post-close; site still says 'current fund $85m'
Aulakh off the team page, unannounced
No DPI, undisclosed exit economics, an unannounced senior change: the institutional hygiene lags the investing. GW read: early-and-right on AI, now early on AI's infrastructure layer; the missing Fund III is the tell to watch. E.
What they're doing
Invest in AI's physical substrate: photonic interconnects, fusion power, quantum photonics: infrastructure before applications.
$250K-3M first cheques; 'Deep India Connect' global startups eligible.
Ride the 2016-17 application-AI vintage to strategic exits (Locus template).
What can break
No Fund III = deployment runway question; the 2025-26 cheques may be Fund II's last.
Hard-tech duration (fusion!) inside a fund structure built for software timelines.
Undisclosed exit economics + quiet team changes erode institutional-LP confidence exactly when a raise needs it.
Fund III: any announcement (none 3 years post-Fund II).
Roopan Aulakh status: unconfirmed departure.
Locus/Wysa exit economics: undisclosed.
Point-in-time discipline: these are the open items we could not verify at the current vintage. They get resolved: confirmed, corrected or dropped, at the next monthly refresh, never silently.
The takes · one screen, three readers
The investing eye is proven; the operation isn't. If a Fund III materialises, demand the Locus/Wysa economics and the team-change story first: both are knowable and both are being left unsaid.
pi's photonics/fusion seeds are the cheapest exposure to AI-infrastructure themes you're seeing at Series B prices elsewhere. Its follow-on constraint is your entry point.
LightSpeed Photonics and Quanfluence map India's answer to the AI-compute bottleneck we flagged in the data-centre dossier: watch whether global strategics repeat the IKEA/Locus pattern here.
Sourcing. Every figure is sourced and dated. We tier provenance: Primary (official, regulatory, exchange or company filings), Secondary (tier-1 industry research and reputable media), and GW estimate (our own reconstruction or opinion, labelled, never presented as external fact). We prefer primary where it exists, reconcile divergent prints to cited ranges, and hold every number point-in-time: dated, and never silently restated; revisions publish as dated changes.
Fact vs opinion. Facts vs opinion: market sizes, official prints, prices, named deals and agency ratings are sourced facts (Primary/Secondary). Scores, grades, purity weights, scenario paths and indicative sparkline points are Gravitywell's analytical opinion (GW estimate): labelled, not presented as external data.
GW GP Scores are research opinions, not investment advice, not a solicitation, and not an assessment under any SEBI regulation. Dossiers are compiled from public sources believed reliable; firms named did not participate and figures marked E are Gravitywell estimates. Point-in-time: dated to the vintage shown and never silently restated.