TinyCubeWayKey findings
- $2.9B raised across 240 disclosed deals: up 4% on Q3 2025, down 17% on Q2 2026.
- AI was the top sector at $635M, with electric vehicles close behind at $583M. Together they took two of every five dollars.
- Bengaluru took 72.5% of all money raised. Series C took the most money; seed did the most deals.
- One in three dollars in the quarter's 22 headline rounds went to companies where AI or data is the product itself.
- In five of those six companies, the moat is not the model. It is owned data, enterprise integrations, silicon, a licence or distribution.
- Only 3% of investors expect data, tooling and evaluation to capture the most AI value. We think that layer is underrated.
1. The quarter in numbers
Funding beat the same quarter last year but slowed for the second quarter running in 2026. A strong September, the third month of the year to cross $1 billion, kept Q3 above Q3 2025.
2. Where the money went
AI and electric vehicles took two of every five dollars raised. Electric-vehicle funding was driven by three large two-wheeler rounds: Simple Energy, River and Ultraviolette.
The year-on-year view shows where momentum moved. AI, cleantech and hard tech roughly tripled, while fintech and ecommerce, long the backbone of Indian venture funding, shrank.
3. Stage and city
Seed rounds made up the largest number of deals, but Series C took the most money: 11 rounds worth $887M, or 30.5% of everything raised. Early-stage deals overall still crossed $1B across 194 rounds.
| Stage | Deals | Money raised | Average cheque |
|---|---|---|---|
| Seed | 88 | $175M | $2.0M |
| Series A | 59 | $689M | $11.7M |
| Series B | 19 | $506M | $26.6M |
| Series C | 11 | $887M | $80.6M |
Source: Entrackr Q3 2026. Average cheque is our calculation (money raised divided by deals). Later rounds and debt are not shown.
4. Where the moat sits: a TinyCubeWay analysis
Funding totals say where money went. They do not say what it bought. So we took the quarter's 22 headline rounds, Entrackr's top 15 growth-stage and top 7 early-stage deals, worth $1.65B or about 57% of all money raised, and classified each company on two questions:
- What is its primary moat? The thing a well-funded competitor would find hardest to copy.
- How much does that moat rest on AI and data? Core (AI or data is the product), enabling (data sharpens a moat built elsewhere) or peripheral (the moat lies elsewhere).
This is a judgement call based on public company descriptions, not reported data. The full classification is in the appendix below.
| Company | What it does | Raised | Role of AI and data |
|---|---|---|---|
| Simple Energy | Electric scooters with in-house motors | $182M | Peripheral |
| SiMa.ai | Chips that run AI at the edge | $150M | Core |
| Emergent | AI app builder for non-coders | $130M | Core |
| River Mobility | Electric scooters | $120M | Peripheral |
| Navi | Lending, insurance and payments | $100M | Enabling |
| Slice | Digital-first small finance bank | $100M | Enabling |
| Pixxel | Hyperspectral satellite imagery | $100M | Core |
| Yulu | Shared electric two-wheeler fleet | $93M | Enabling |
| Ultraviolette | Electric motorcycles | $85M | Peripheral |
| Ema | AI agents for HR, IT and finance work | $77M | Core |
The ten largest rounds of the quarter. Source: Entrackr; role classified by TinyCubeWay.
Most of the money still buys moats you can touch
One in three dollars went to companies where AI or data is the product
Core · $558M · 34%
AI or data is the product itself.
SiMa.ai, Emergent, Pixxel, Ema, Sarvam AI, InRisk Labs
Enabling · $435M · 26%
Data sharpens a moat built elsewhere: credit, fleets, devices.
Navi, Slice, Yulu, Ultrahuman, Airbound, Yuma Energy
Peripheral · $655M · 40%
The moat is hardware, brand or physical capacity.
Simple Energy, River, Ultraviolette, The Pizza Bakery, Nua, Third Wave Coffee, Raghu Vamsi, Centricity, CtrlS, Arboreal
Add the enabling group, and about 60% of the money backed companies whose edge depends heavily on AI or data.
5. The key finding: the moat is rarely the model
In five of the six AI-core companies, the moat is not the model.
SiMa.ai
Silicon: its own chips for running AI on devices.
Pixxel
An owned data source: its hyperspectral satellites.
Ema
250+ enterprise integrations, with 150+ models it can swap underneath.
InRisk Labs
A reinsurance licence, plus climate-risk data and models.
Emergent
Distribution: reach among non-technical founders building apps.
Sarvam AI · the exception
A model-led bet: Indian-language foundation models.
Ema is the clearest case. It can draw on more than 150 models, which tells you the model is the part it treats as replaceable. What it does not treat as replaceable is its connections into more than 250 business systems. Pixxel's edge is data no one else has. InRisk Labs pairs a licence with climate data and risk models built on it.
6. The investor view
Inc42 surveyed more than 85 institutional investors. 63% see some froth in Indian AI valuations, and about half of those expect a correction within 18 months. On where AI value will be captured, the data layer came last.
7. Six more numbers that defined the quarter
8. What this means for teams building AI on their own data
Capital got pickier
Fewer investors wrote cheques, the median round held at $3M, and the biggest money sat at Series C. Traction now matters more than the story.
AI leads; the model does not
AI was the top sector, yet most AI-core winners built their edge around the model: owned data, silicon, integrations and licences.
The data layer is underrated
Only 3% of investors picked data, tooling and evaluation. Yet owned data or deep integrations were the edge in three of the six AI-core rounds.
Models are becoming interchangeable. What stays hard to copy is what the model is given: which of your systems it can reach, how fresh that data is, and whether it reflects what is true right now. We call this the context layer.
For an AI agent that approves a credit line, reprices stock or answers "where is my order", stale context does not produce a slow answer. It produces a wrong action, with a direct cost.
For Series A to C teams
Taking an AI feature from pilot to production?
TinyCubeWay builds the context layer: streaming pipelines that keep operational data fresh, lakehouse platforms on Databricks and the major clouds, and the retrieval layer between your systems and your AI.
Book a free 30-minute context-layer review. We will look at:
- what data your AI can and cannot see today
- how stale that data is when the AI acts on it
- where the pipeline is most likely to break first
Method and sources
- Totals, sectors, stages, cities, deals, listings, M&A and shutdowns: Entrackr, Q3 2026 quarterly funding report (1 October 2026), full quarter.
- Year-on-year sector change, investor count, unicorns and median round: Inc42, Indian Tech Startup Funding Report Q3 2026 (data to 24 September 2026).
- Investor survey: Inc42, Indian AI startup funding surges 265% YoY in Q3.
- Company details: company announcements, including Ema's Series B and InRisk Labs' Series A.
- Moat and AI and data classification: TinyCubeWay analysis of 22 headline rounds. Figures are rounded; trackers differ in scope and cut-off date, so totals differ by source.
Appendix: how we classified the 22 headline rounds
| Company | $M | Primary moat | AI and data |
|---|---|---|---|
| Simple Energy | 182 | Hardware | Peripheral |
| SiMa.ai | 150 | Deep-tech IP | Core |
| Emergent | 130 | Distribution | Core |
| River Mobility | 120 | Hardware | Peripheral |
| Navi | 100 | Licence | Enabling |
| Slice | 100 | Licence | Enabling |
| Pixxel | 100 | Proprietary data | Core |
| Yulu | 93 | Physical network | Enabling |
| Ultraviolette | 85 | Hardware | Peripheral |
| Ema | 77 | Enterprise integrations | Core |
| Sarvam AI | 74 | Proprietary data | Core |
| Ultrahuman | 70 | Hardware | Enabling |
| The Pizza Bakery | 56 | Distribution | Peripheral |
| Nua | 50 | Distribution | Peripheral |
| Third Wave Coffee | 43 | Distribution | Peripheral |
| Raghu Vamsi | 40 | Hardware | Peripheral |
| Airbound | 37 | Hardware | Enabling |
| Yuma Energy | 35 | Physical network | Enabling |
| Centricity | 30 | Distribution | Peripheral |
| InRisk Labs | 27 | Licence | Core |
| CtrlS | 25 | Physical network | Peripheral |
| Arboreal | 24 | Deep-tech IP | Peripheral |