Loading page…
Together.ai has contracted for a significant share of the planned capacity. The CFO’s roughly $750 million investment estimate is not a final project budget.
Listen to this story
A long-term deployment agreement gives Together.ai a committed customer position for a substantial, undisclosed share of L&T’s planned Chennai capacity, but does not establish when or on what terms others can use the facility. L&T estimates that deploying 10,000 GPUs would require roughly $750 million, while explicitly presenting that figure as approximate rather than a final project budget. The January 2027 target is for commissioning, not necessarily a public commercial launch.
L&T is developing the facility at its Chennai campus through wholly owned subsidiary L&T N Compute Private Limited.
The planned GPU count exceeds 10,000 NVIDIA B300s; the final equipment tally has not been disclosed.
Together.ai’s contract covers a significant share of capacity, but its exact share, value, and duration remain undisclosed.
Larsen & Toubro is targeting January 2027 to commission a Chennai AI computing facility with more than 10,000 NVIDIA B300 GPUs. Together.ai has already contracted for a significant share of its capacity, according to Shattered, citing NDTV Profit and CXO Digital Pulse. The project has a named customer, but the target is not a promise of public availability.
The facility is being developed at L&T’s Chennai campus through L&T N Compute Private Limited, a wholly owned subsidiary. L&T calls the site an AI factory: a computing facility built for AI workloads. The planned GPU count is a lower bound, not a final equipment tally.
L&T N Compute’s long-term deployment agreement with Together.ai covers a significant portion of the facility’s GPU capacity. The precise share, contract value and duration have not been disclosed.
The agreement puts a named customer against planned capacity, before the facility has been commissioned.
Chief Financial Officer Parameswaran Ramakrishnan put the investment for 10,000 GPUs at approximately $750 million. His wording—“close to maybe around $750 million US dollars”—makes the uncertainty explicit. The investment estimate gives a rough sense of scale, but it is not a published final cost for the Chennai project.
Treating it as a chip purchase price—or dividing it into a definitive cost per GPU—would imply a precision the disclosure does not provide.
Loading discussion...
Join the conversation
Explain what that commitment reassures you about—and what it leaves unresolved.
Be the first to share a perspective or an experience.
Reader comments
Newest comments first. Replies stay oldest first.