Sarvam AI Raises Nvidia-Backed Series B Extension
Sarvam AI approved a roughly $74 million Series B extension on July 24 that includes Nvidia, according to MCA filings reported by Inc42. The filings put Nvidia's proposed investment at about Rs 237.9 crore, while reports value Sarvam at roughly $1.5 billion post-money; the financing follows the company's announced plan to train a trillion-plus-parameter foundation model on 10,000 Nvidia Blackwell GPUs.
Sarvam AI is raising an approximately $74 million extension to its ongoing Series B round, with Nvidia among the investors, according to Ministry of Corporate Affairs filings reported by Inc42. The Bengaluru-based company approved a resolution at a July 24 extraordinary general meeting to raise about Rs 697.8 crore from roughly 13 investors, Inc42 reported.
The proposed extension follows Sarvam's June Series B tranche of up to $234 million and places its reported post-money valuation near $1.5 billion, according to Business Standard, citing Entrackr. Sarvam co-founder Pratyush Kumar confirmed Nvidia's participation to The Economic Times, as quoted by India Today, and described prior technical work between the companies on Nvidia's software and infrastructure stack.
Funding structure and investors
According to Inc42's review of the filings, Nvidia is slated to invest Rs 237.9 crore, or roughly $25 million. Glade Brook is listed for Rs 190.3 crore, Gaja Capital for Rs 95.1 crore, and Indigo Ventures for Rs 30 crore; the remainder is attributed to high-net-worth investors and family offices.
Free Press Journal reported that the share issuance approved by Sarvam's board consisted of 20,244 Series B preference shares and seven equity shares, each priced at Rs 344,570. Reporting on the transaction differs on whether to characterize the extension as completed or pending: Inc42 and Business Standard describe Sarvam as raising or set to raise the capital through the filing-approved issuance, while Free Press Journal describes the extension as closed.
Kumar told The Economic Times that Sarvam was the first Indian company to receive an Nvidia investment. Nvidia's participation is notable because the company is both a chip supplier and a capital provider to an AI developer whose reported compute roadmap relies on Nvidia hardware.
Trillion-parameter model target
The financing arrived days after Sarvam announced an effort to build a trillion-plus-parameter foundation model from scratch. Business Standard reported that the founders said the model would be trained on a cluster of 10,000 Nvidia Blackwell GPUs and targeted at coding, cybersecurity, scientific research, and enterprise AI applications.
Sarvam develops foundation models, inference infrastructure, and enterprise AI products for Indian languages and use cases, according to Business Standard. The company was selected earlier in 2026 under the IndiaAI Mission to develop India's first sovereign foundation model, the publication reported.
The Head and Tale reported that Sarvam has also appointed Devendra Chaplot, previously a founding member at Mistral AI and a former employee of xAI, as it publicized its model-building target.
What the financing indicates for infrastructure teams
A trillion-parameter training target would place Sarvam in a compute-intensive class of model development where GPU procurement is only one part of the engineering challenge. Comparable efforts typically require distributed-training orchestration, high-bandwidth interconnects, storage throughput, fault tolerance, data curation, evaluation infrastructure, and sustained inference capacity after training.
For ML practitioners, Nvidia's reported investment and Sarvam's declared Blackwell cluster target underscore the increasingly tight relationship between model developers and their infrastructure suppliers. In comparable frontier-model programs, access to accelerators, optimized training software, and systems expertise can materially affect iteration speed and experiment reliability, alongside the availability of capital.
The available reports do not specify the intended model architecture, training-data composition, token budget, interconnect configuration, training duration, or evaluation methodology. Those details would be necessary to assess how Sarvam's proposed model compares technically with other large-scale foundation-model efforts.
Key Points
- 1Sarvam's proposed $74 million extension adds Nvidia as an investor, linking fresh capital with a reported Blackwell-based training infrastructure target.
- 2Corporate filings reported by Inc42 assign Nvidia roughly $25 million of the extension, alongside Glade Brook, Gaja Capital, and Indigo Ventures.
- 3Comparable trillion-parameter programs depend on distributed systems, data pipelines, evaluation, and inference operations, not GPU count alone.
Scoring Rationale
Nvidia's reported investment in a $1.5 billion Indian foundation-model developer is a notable financing and infrastructure ecosystem event. The trillion-parameter target and 10,000-GPU claim make it relevant to practitioners tracking large-scale training capacity, although key technical details remain undisclosed.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
