deepgridSEMI

Pre-Series A · Hyderabad, India

Silicon & ComputeT100

T100 AI Licence

The T100 AI Licence is a pure software offering within the Silicon & Compute segment that licenses trained Indian road perception models, runtime toolchains, and update streams to third-party hardware platforms. It is the only product in the portfolio with no bill of materials, projected by management to generate ₹50 Cr at 94% gross margin in FY2032 as optionality without a separately sized demand pool.

Listed price
₹100.00 L
FY2032 revenue
₹50 Cr
Share of plan
4.4%
Gross margin
94%
First revenue
FY2029
Volume plan
50 @ FY32
Concept render of the DeepGrid SoC2 package on a circuit board, marked 28 nm and 39.3 TOPS
Compute inside the customer’s systemConcept render · T100 domain

From input to outcome

Proposed product architecture. The scene illustrates its application, not a shipped kit.

  1. The customer runs their own hardware and their own sensors

  2. Our models do the perception work inside their runtime

  3. Their system decides and acts

  4. They ship under their own brand, with our perception inside

The T100 Transformer NPU on the SoC2 dieTechnology: six domains, one tapeout

Drag to rotate. Conceptual layout, not a mask.

On the die

T100 AI Licence

Runs on one of the six domains:

  • T100 Transformer NPU

The case in brief

  • The T100 AI Licence is a zero-BOM software offering projected to deliver ₹50.0 Cr at 94% gross margin by FY2032, treated as strategic optionality without a separately sized demand pool.
  • Commercialization relies on licensing proprietary India-road perception models and runtime toolchains to OEMs deploying their own compute hardware and sensors.
  • First revenue begins in FY2029 and depends on establishing an operational track record and securing signed customer licences, neither of which is evidenced yet.

What it is used for

Third-party automotive compute integration

Buyer
OEMs and platform partners who want the India-road stack without our silicon
The job
OEMs seeking perception tuned for Indian traffic conditions, lane discipline, and lighting want to avoid spending three years collecting road datasets to train models themselves.
What DeepGrid supplies
Trained India-road perception models, toolchain and deployment runtime, and continuous model update streams delivered via licence subscription to run on the customer's hardware.

Accountable integrator perception subsystem

Buyer
accountable integrator
The job
Integrators must deploy a certified perception stack within their own system architecture while retaining their own brand, sensors, and compute hardware.
What DeepGrid supplies
A bounded camera-perception subsystem that runs inference inside the customer's runtime, allowing the customer system to decide, act, and ship with DeepGrid perception inside.

The story

A pure-software licence delivering trained India-road perception models and toolchains directly to third-party platforms.

The T100 AI Licence delivers DeepGrid's trained perception models, runtime toolchain, and ongoing update stream without physical hardware. Priced per deployment at ₹100.00 L, it enables customers to run India-specific perception on their own hardware architectures. Management models ₹50 Cr on pure margin with no unit cost, treating the line as portfolio optionality rather than base case.

  • ₹100.00 Lunit price
  • Semiconductorssegment
  • Outdooroperating domain
  • FY2029first revenue
  • Trained India-road perception modelscomponent
  • Toolchain and deployment runtimecomponent

Portfolio deck, slide 59: What it is

Models execute perception inside the customer's runtime while running on their own sensors and hardware.

Customers deploy their own compute platforms and sensor suites, using DeepGrid's models to handle the perception layer. The customer's system retains control over downstream decision and actuation pipelines, shipping finished vehicles under their own brand. Because the product delivers software without a physical board, it carries no bill of materials and prices entirely on value.

  • Customer sensors and hardwareinput stage
  • DeepGrid runtime perceptioncompute stage
  • Customer decision and actiondownstream stage
  • Zero BOMhardware cost
  • Defensible model advantagedependency

Portfolio deck, slide 60: How it works

Automotive OEMs and platform partners seeking localized India-road perception without multi-year data collection.

The offering targets teams that need perception tuned for Indian traffic patterns, lane discipline, and lighting without spending years gathering local training datasets. Sales are negotiated directly per customer engagement to support custom platform integration. No separate demand pool is modelled for this line, leaving its adoption scheduled purely as management optionality.

  • OEMs and platform partnersbuyer
  • Direct licensingchannel
  • Bypassing local dataset collectiontrigger
  • None separately sizeddemand pool

Portfolio deck, slide 61: Who buys it

A 94% gross margin line scaling to ₹50.0 Cr in FY2032 to monetize proprietary dataset investments.

Revenue commences in FY2029 following the establishment of an on-road perception track record, scaling from 2 units to 50 units by FY2032. The product contributes 4.4% of group revenue at a 94% gross margin, representing management projections without an underlying sized demand pool. Commercial delivery depends on converting model performance advantages into signed licensing agreements.

  • 50 unitsFY2032 units projection
  • ₹50.0 CrFY2032 revenue projection
  • 4.4%share of group projection
  • 94%gross margin projection
  • Portfolio optionalitystrategic role
Revenue and units by year (management projection)
  1. ₹0 CrFY270 units
  2. ₹0 CrFY280 units
  3. ₹2 CrFY292 units
  4. ₹6 CrFY306 units
  5. ₹18 CrFY3118 units
  6. ₹50 CrFY3250 units

Portfolio deck, slide 62: How it fits

What has to go right

Commercialization depends on demonstrating a defensible model advantage and securing signed licence agreements, neither of which is evidenced yet. DeepGrid must prove maturity and build an operational track record for its perception stack before the FY2029 revenue start. Management must also manage the risk that licensing the standalone core could undercut its own full-stack systems business, which represents approximately 88% of planned revenue. Finally, sales must close directly with OEMs despite the absence of a separately sized demand pool in the business plan.

Key dependency Maturity of the Indian road perception stack, starts FY29.

Further reading and viewing

Related products

Prices, volumes, revenues and margins are management projections. The text is written from the linked documents, and every figure in it appears in the source it came from.